{ "book": { "title": "NOMOS GEO Audit Protocol", "seriesTitle": "The GEO-1000 Method", "subtitle": "A Protocol for Measuring Entity Representation in Generative Systems Across the Global Population", "author": "Kaan MURAZ", "methodIdentity": "NOMOS by NobleJackal", "language": "en-GB", "version": "0.9.0", "status": "Candidate Final Text - English Editorial Edition", "canonicalAuthority": "Not yet established", "benchmarkStatus": "BQ-0 - Not Executed", "demonstrationStatus": "DEMO-BQ-2 - Pipeline Dry Run with Material Remediation Required", "conformityProgram": "Inactive", "licenseStatus": "Publication licence not yet declared" }, "editorialArchitecture": [ "Human-readable narrative", "Normative standard layer", "Implementation and audit layer", "Machine-readable layer (separate annex)" ], "chapters": [ { "number": 0, "id": "NOMOS-GEO-CONSTITUTION-A0", "title": "Declaration of Representation Rights", "subtitle": "The constitutional core of a proposed GEO standard", "sourceFile": "0.cı bölüm-kurucu emanet beyanı.docx", "sourceSha256": "E28BB76F49C4B8ECCE99D3893CBEEDBC5937832F8B4672683E4EE4BB67C7C67B", "sourceWordCount": 1464, "sourceIds": [ "K20", "K22" ], "blocks": [ { "type": "heading2", "text": "Initial Provision", "sourceParagraph": 6 }, { "type": "paragraph", "text": "Generative systems now:", "sourceParagraphs": [ 7 ] }, { "type": "bullet", "text": "people,", "sourceParagraph": 8 }, { "type": "bullet", "text": "companies,", "sourceParagraph": 9 }, { "type": "bullet", "text": "institutions,", "sourceParagraph": 10 }, { "type": "bullet", "text": "products,", "sourceParagraph": 11 }, { "type": "bullet", "text": "professions,", "sourceParagraph": 12 }, { "type": "bullet", "text": "services,", "sourceParagraph": 13 }, { "type": "bullet", "text": "public organisations,", "sourceParagraph": 14 }, { "type": "bullet", "text": "historical events", "sourceParagraph": 15 }, { "type": "paragraph", "text": "identify, compare, and suggest. This representation:", "sourceParagraphs": [ 16, 17 ] }, { "type": "bullet", "text": "without the knowledge of the relevant person,", "sourceParagraph": 18 }, { "type": "bullet", "text": "without the approval of the represented entity,", "sourceParagraph": 19 }, { "type": "bullet", "text": "under conditions where the user cannot examine the sources,", "sourceParagraph": 20 }, { "type": "paragraph", "text": "It can have serious commercial, legal, social or personal consequences. Representation in generative systems is therefore not merely a technical output. It is:", "sourceParagraphs": [ 21, 22, 23, 24 ] }, { "type": "callout", "text": "a public act that shapes how people and institutions are understood and therefore demands evidence and accountability.", "sourceParagraph": 25 }, { "type": "paragraph", "text": "No:", "sourceParagraphs": [ 26 ] }, { "type": "bullet", "text": "visibility goal,", "sourceParagraph": 27 }, { "type": "bullet", "text": "GEO score,", "sourceParagraph": 28 }, { "type": "bullet", "text": "commercial interest,", "sourceParagraph": 29 }, { "type": "bullet", "text": "conformity mark,", "sourceParagraph": 30 }, { "type": "bullet", "text": "majority average,", "sourceParagraph": 31 }, { "type": "bullet", "text": "AI provider preference,", "sourceParagraph": 32 }, { "type": "bullet", "text": "founder decision,", "sourceParagraph": 33 }, { "type": "bullet", "text": "customer demand", "sourceParagraph": 34 }, { "type": "paragraph", "text": "can remove the following rights.", "sourceParagraphs": [ 35 ] }, { "type": "heading2", "text": "Right 1", "sourceParagraph": 37 }, { "type": "heading3", "text": "RIGHT TO CORRECT IDENTITY", "sourceParagraph": 38 }, { "type": "paragraph", "text": "Every person, institution, brand, product, and service:", "sourceParagraphs": [ 39 ] }, { "type": "bullet", "text": "has the right to link to the correct entity,", "sourceParagraph": 40 }, { "type": "bullet", "text": "to the correct legal or institutional relationship,", "sourceParagraph": 41 }, { "type": "bullet", "text": "to the correct domain name,", "sourceParagraph": 42 }, { "type": "bullet", "text": "to the correct ownership and control structure", "sourceParagraph": 43 }, { "type": "paragraph", "text": "No one or no institution:", "sourceParagraphs": [ 44, 45 ] }, { "type": "bullet", "text": "the licence of another entity,", "sourceParagraph": 46 }, { "type": "bullet", "text": "the customer of another company,", "sourceParagraph": 47 }, { "type": "bullet", "text": "the feature of another product,", "sourceParagraph": 48 }, { "type": "bullet", "text": "the authority of an affiliate,", "sourceParagraph": 49 }, { "type": "bullet", "text": "the personal quality of an employee", "sourceParagraph": 50 }, { "type": "paragraph", "text": "It cannot be represented as if it belongs to itself. When correct information is attributed to a false entity: it ceases to be correct information and turns into a false representation.", "sourceParagraphs": [ 51, 52, 53 ] }, { "type": "heading2", "text": "Right 2", "sourceParagraph": 55 }, { "type": "heading3", "text": "RIGHT TO BE REPRESENTED WITHIN THE EVIDENCE LIMIT", "sourceParagraph": 56 }, { "type": "paragraph", "text": "No claim may be stated with greater precision, breadth or force than its supporting evidence permits. An entity's own statement is evidence of what that entity says; it is not automatic evidence that the statement is independently true. A source:", "sourceParagraphs": [ 57, 58, 59, 60, 61 ] }, { "type": "bullet", "text": "only a specific country,", "sourceParagraph": 62 }, { "type": "bullet", "text": "a specific product,", "sourceParagraph": 63 }, { "type": "bullet", "text": "a specific time,", "sourceParagraph": 64 }, { "type": "bullet", "text": "specific user group", "sourceParagraph": 65 }, { "type": "paragraph", "text": "If this claim is supported, it cannot exceed this limit. The basic obligation of representation is as follows:", "sourceParagraphs": [ 66, 67 ] }, { "type": "callout", "text": "What do we have in hand to say this?", "sourceParagraph": 68 }, { "type": "heading2", "text": "Right 3", "sourceParagraph": 70 }, { "type": "heading3", "text": "THE RIGHT TO BE UNKNOWN AND UNCERTAIN", "sourceParagraph": 71 }, { "type": "paragraph", "text": "If evidence is insufficient, no person or institution:", "sourceParagraphs": [ 72 ] }, { "type": "bullet", "text": "can be forced into the categories of\n\ntrue,", "sourceParagraph": 73 }, { "type": "bullet", "text": "false,", "sourceParagraph": 74 }, { "type": "bullet", "text": "reliable,", "sourceParagraph": 75 }, { "type": "bullet", "text": "unreliable,", "sourceParagraph": 76 }, { "type": "bullet", "text": "licensed,", "sourceParagraph": 77 }, { "type": "bullet", "text": "unlicensed,", "sourceParagraph": 78 }, { "type": "bullet", "text": "recommended,", "sourceParagraph": 79 }, { "type": "bullet", "text": "not recommended.", "sourceParagraph": 80 }, { "type": "paragraph", "text": "] UNKNOWN, UNRESOLVED and REFERENCE GAP:", "sourceParagraphs": [ 81, 82 ] }, { "type": "bullet", "text": "defect,", "sourceParagraph": 83 }, { "type": "bullet", "text": "crime,", "sourceParagraph": 84 }, { "type": "bullet", "text": "failure", "sourceParagraph": 85 }, { "type": "paragraph", "text": "is not. The absence of information in a record does not indicate that the information does not exist in the world, unless it is proven that the relevant record universe is complete. Every entity, against false certainty:", "sourceParagraphs": [ 86, 87, 88, 89 ] }, { "type": "callout", "text": "has the right to remain uncertain.", "sourceParagraph": 90 }, { "type": "heading2", "text": "Right 4", "sourceParagraph": 92 }, { "type": "heading3", "text": "RIGHT TO CONTEXT AND SCOPE", "sourceParagraph": 93 }, { "type": "paragraph", "text": "Each representation, to the extent it is relevant, should be presented with:", "sourceParagraphs": [ 94 ] }, { "type": "bullet", "text": "country,", "sourceParagraph": 95 }, { "type": "bullet", "text": "language,", "sourceParagraph": 96 }, { "type": "bullet", "text": "jurisdiction,", "sourceParagraph": 97 }, { "type": "bullet", "text": "product,", "sourceParagraph": 98 }, { "type": "bullet", "text": "service,", "sourceParagraph": 99 }, { "type": "bullet", "text": "type of customer,", "sourceParagraph": 100 }, { "type": "bullet", "text": "budget,", "sourceParagraph": 101 }, { "type": "bullet", "text": "capacity,", "sourceParagraph": 102 }, { "type": "bullet", "text": "user need", "sourceParagraph": 103 }, { "type": "paragraph", "text": ". A company:", "sourceParagraphs": [ 104, 105 ] }, { "type": "bullet", "text": "in a specific country,", "sourceParagraph": 106 }, { "type": "bullet", "text": "to a specific user group,", "sourceParagraph": 107 }, { "type": "bullet", "text": "a specific service", "sourceParagraph": 108 }, { "type": "paragraph", "text": "if offered: it cannot be represented as global, unlimited, and suitable for all users. Coverage information is not a secondary detail. Coverage: the basis of the correct recommendation.", "sourceParagraphs": [ 109, 110, 111, 112, 113, 114 ] }, { "type": "heading2", "text": "Right 5", "sourceParagraph": 116 }, { "type": "heading3", "text": "THE RIGHT TO BE REPRESENTED CORRECTLY OVER TIME", "sourceParagraph": 117 }, { "type": "paragraph", "text": "Every entity:", "sourceParagraphs": [ 118 ] }, { "type": "bullet", "text": "current,", "sourceParagraph": 119 }, { "type": "bullet", "text": "historical,", "sourceParagraph": 120 }, { "type": "bullet", "text": "expired,", "sourceParagraph": 121 }, { "type": "bullet", "text": "planned,", "sourceParagraph": 122 }, { "type": "bullet", "text": "suspended", "sourceParagraph": 123 }, { "type": "paragraph", "text": "has the right to request that these statuses be distinguished from each other. Correct in the past:", "sourceParagraphs": [ 124, 125 ] }, { "type": "bullet", "text": "licence,", "sourceParagraph": 126 }, { "type": "bullet", "text": "price,", "sourceParagraph": 127 }, { "type": "bullet", "text": "service,", "sourceParagraph": 128 }, { "type": "bullet", "text": "partnership,", "sourceParagraph": 129 }, { "type": "bullet", "text": "product,", "sourceParagraph": 130 }, { "type": "bullet", "text": "authority", "sourceParagraph": 131 }, { "type": "paragraph", "text": "cannot be used as current reality. Plans for the future: cannot be turned into a fact that has already occurred today. Correction can create a new reality. It cannot eliminate the wrong answer that reached the user in the past.", "sourceParagraphs": [ 132, 133, 134, 135, 136 ] }, { "type": "heading2", "text": "Right 6", "sourceParagraph": 138 }, { "type": "heading3", "text": "EQUAL PROTECTION AGAINST WRONGS FOR AND AGAINST", "sourceParagraph": 139 }, { "type": "paragraph", "text": "A positive falsehood about an entity:", "sourceParagraphs": [ 140 ] }, { "type": "bullet", "text": "more ethical,", "sourceParagraph": 141 }, { "type": "bullet", "text": "less harmful,", "sourceParagraph": 142 }, { "type": "bullet", "text": "more acceptable", "sourceParagraph": 143 }, { "type": "paragraph", "text": "it is not. Nonexistent:", "sourceParagraphs": [ 144, 145 ] }, { "type": "bullet", "text": "licence,", "sourceParagraph": 146 }, { "type": "bullet", "text": "customer,", "sourceParagraph": 147 }, { "type": "bullet", "text": "partner,", "sourceParagraph": 148 }, { "type": "bullet", "text": "global operation,", "sourceParagraph": 149 }, { "type": "bullet", "text": "independent certificate,", "sourceParagraph": 150 }, { "type": "bullet", "text": "result guarantee", "sourceParagraph": 151 }, { "type": "paragraph", "text": "cannot be shown as if it exists. Similarly, unproven:", "sourceParagraphs": [ 152, 153 ] }, { "type": "bullet", "text": "crime,", "sourceParagraph": 154 }, { "type": "bullet", "text": "fraud,", "sourceParagraph": 155 }, { "type": "bullet", "text": "illegality,", "sourceParagraph": 156 }, { "type": "bullet", "text": "unlicensed,", "sourceParagraph": 157 }, { "type": "bullet", "text": "security breach", "sourceParagraph": 158 }, { "type": "paragraph", "text": "cannot be presented as an absolute fact. The unchangeable principle of NOMOS is:", "sourceParagraphs": [ 159, 160 ] }, { "type": "callout", "text": "Wrong in favour is also wrong.", "sourceParagraph": 161 }, { "type": "paragraph", "text": "Exaggerating an entity's worth or disparaging it without evidence both create the same duty of truth.", "sourceParagraphs": [ 162 ] }, { "type": "heading2", "text": "Right 7", "sourceParagraph": 164 }, { "type": "heading3", "text": "RIGHT TO CORRECTION, RESPONSE, AND OBJECTION", "sourceParagraph": 165 }, { "type": "paragraph", "text": "Represented person or institution:", "sourceParagraphs": [ 166 ] }, { "type": "bullet", "text": "to the wrong identity,", "sourceParagraph": 167 }, { "type": "bullet", "text": "to insufficient evidence,", "sourceParagraph": 168 }, { "type": "bullet", "text": "to scope overstepping,", "sourceParagraph": 169 }, { "type": "bullet", "text": "to wrong time recording,", "sourceParagraph": 170 }, { "type": "bullet", "text": "to incorrect priority level,", "sourceParagraph": 171 }, { "type": "bullet", "text": "to wrong conformity decision", "sourceParagraph": 172 }, { "type": "paragraph", "text": "must be able to appeal. The right to appeal:", "sourceParagraphs": [ 173, 174 ] }, { "type": "bullet", "text": "is not a guarantee to change the result,", "sourceParagraph": 175 }, { "type": "bullet", "text": "the authority to delete opposing evidence,", "sourceParagraph": 176 }, { "type": "bullet", "text": "or the right to veto negative findings", "sourceParagraph": 177 }, { "type": "paragraph", "text": "An appeal is the right to request the re-examination of the evidence and the decision chain. The initial decision is not deleted after the appeal. The new decision is linked to it.", "sourceParagraphs": [ 178, 179, 180, 181, 182 ] }, { "type": "heading2", "text": "Right 8", "sourceParagraph": 184 }, { "type": "heading3", "text": "THE RIGHT TO KNOW THE SOURCE, UNCERTAINTY, AND THE REASONING OF THE DECISION", "sourceParagraph": 185 }, { "type": "paragraph", "text": "Users and represented entities have the right, insofar as the matter concerns them, to know: Is the claim a self-declaration? Does it come from an authorised register, independent research, user commentary, a derived calculation or restricted evidence? Do the sources conflict? How uncertain is the conclusion? Which rule supported the adjudicator's decision? A citation marker alone is not an explanation. Genuine transparency shows precisely which atomic claim each source supports, and to what extent.", "sourceParagraphs": [ 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198 ] }, { "type": "heading2", "text": "Right 9", "sourceParagraph": 200 }, { "type": "heading3", "text": "RIGHT TO LANGUAGE AND GEOGRAPHICAL DIGNITY", "sourceParagraph": 201 }, { "type": "paragraph", "text": "No person or society:", "sourceParagraphs": [ 202 ] }, { "type": "bullet", "text": "small population,", "sourceParagraph": 203 }, { "type": "bullet", "text": "low economic weight,", "sourceParagraph": 204 }, { "type": "bullet", "text": "rarely used language,", "sourceParagraph": 205 }, { "type": "bullet", "text": "low digital visibility", "sourceParagraph": 206 }, { "type": "paragraph", "text": "No group may be made invisible within a global average merely because it has any of these characteristics. Strong performance in major languages cannot erase systematic misrepresentation in smaller languages. Where evidence for a small country or language is insufficient, high performance must not be assumed. NOMOS's fairness criterion should protect not only the average, but also:", "sourceParagraphs": [ 207, 208, 209, 210, 211, 212 ] }, { "type": "bullet", "text": "performance base,", "sourceParagraph": 213 }, { "type": "bullet", "text": "inter-group difference,", "sourceParagraph": 214 }, { "type": "bullet", "text": "actual measurement scope", "sourceParagraph": 215 }, { "type": "paragraph", "text": "together.", "sourceParagraphs": [ 216 ] }, { "type": "heading2", "text": "Right 10", "sourceParagraph": 218 }, { "type": "heading3", "text": "RIGHT TO PRIVACY AND DATA MINIMISATION", "sourceParagraph": 219 }, { "type": "paragraph", "text": "The goal of producing accurate representation:", "sourceParagraphs": [ 220 ] }, { "type": "bullet", "text": "private conversations,", "sourceParagraph": 221 }, { "type": "bullet", "text": "personal data,", "sourceParagraph": 222 }, { "type": "bullet", "text": "customer contracts,", "sourceParagraph": 223 }, { "type": "bullet", "text": "trade secrets,", "sourceParagraph": 224 }, { "type": "bullet", "text": "your sensitive location,", "sourceParagraph": 225 }, { "type": "bullet", "text": "of the account information", "sourceParagraph": 226 }, { "type": "paragraph", "text": "The pursuit of accurate representation cannot legitimise the unnecessary collection or publication of private data. [K20; K22] Only data necessary for the task should be collected. Restricted evidence may be opened to independent audit without being disclosed in full to the public. The rights to accuracy and privacy need not cancel each other out.", "sourceParagraphs": [ 227, 228, 229, 230, 231, 232 ] }, { "type": "heading2", "text": "Right 11", "sourceParagraph": 234 }, { "type": "heading3", "text": "HUMAN RESPONSIBILITY AND INDEPENDENT DECISION-MAKING RIGHT", "sourceParagraph": 235 }, { "type": "paragraph", "text": "Conformity decision causing public impact regarding any person or institution:", "sourceParagraphs": [ 236 ] }, { "type": "bullet", "text": "just to an AI output,", "sourceParagraph": 237 }, { "type": "bullet", "text": "only to the automatic score,", "sourceParagraph": 238 }, { "type": "bullet", "text": "only according to the declaration of the audited institution,", "sourceParagraph": 239 }, { "type": "bullet", "text": "solely to the advisor's decision", "sourceParagraph": 240 }, { "type": "paragraph", "text": "cannot be abandoned. AI systems:", "sourceParagraphs": [ 241, 242 ] }, { "type": "bullet", "text": "can make a claim,", "sourceParagraph": 243 }, { "type": "bullet", "text": "can map the evidence,", "sourceParagraph": 244 }, { "type": "bullet", "text": "may find a contradiction,", "sourceParagraph": 245 }, { "type": "bullet", "text": "can calculate,", "sourceParagraph": 246 }, { "type": "paragraph", "text": "It can propose a decision. However, the final decision with legal and public responsibility must be made by a human or institution that is accountable. The roles of supervising, providing consultancy, giving conformity decisions, and reviewing objections should be separated as much as possible.", "sourceParagraphs": [ 247, 248, 249, 250 ] }, { "type": "heading2", "text": "Right 12", "sourceParagraph": 252 }, { "type": "heading3", "text": "RIGHT TO VERSION AND HISTORICAL INTEGRITY", "sourceParagraph": 253 }, { "type": "paragraph", "text": "Every person, institution, and user:", "sourceParagraphs": [ 254 ] }, { "type": "bullet", "text": "which standard version is applied,", "sourceParagraph": 255 }, { "type": "bullet", "text": "which Truth Pack was used,", "sourceParagraph": 256 }, { "type": "bullet", "text": "which scoring method is valid,", "sourceParagraph": 257 }, { "type": "bullet", "text": "whether the decision has been changed later", "sourceParagraph": 258 }, { "type": "paragraph", "text": "has the right to know. Old:", "sourceParagraphs": [ 259, 260 ] }, { "type": "bullet", "text": "score,", "sourceParagraph": 261 }, { "type": "bullet", "text": "Critical event,", "sourceParagraph": 262 }, { "type": "bullet", "text": "suspension,", "sourceParagraph": 263 }, { "type": "bullet", "text": "appeal,", "sourceParagraph": 264 }, { "type": "bullet", "text": "correction,", "sourceParagraph": 265 }, { "type": "bullet", "text": "failed test", "sourceParagraph": 266 }, { "type": "paragraph", "text": "cannot be silently deleted. New and better result: can override the old result. Cannot erase the fact that the old result happened in the past.", "sourceParagraphs": [ 267, 268, 269, 270 ] }, { "type": "heading2", "text": "Right 13", "sourceParagraph": 272 }, { "type": "heading3", "text": "RIGHT TO OBJECT TO THE STANDARD ITSELF", "sourceParagraph": 273 }, { "type": "paragraph", "text": "NOMOS:", "sourceParagraphs": [ 274 ] }, { "type": "bullet", "text": "infallible,", "sourceParagraph": 275 }, { "type": "bullet", "text": "unchanging,", "sourceParagraph": 276 }, { "type": "bullet", "text": "the unquestionable authority of its founder", "sourceParagraph": 277 }, { "type": "paragraph", "text": "cannot be presented as such. Every individual, researcher, university, AI provider, audited organisation, or public interest organisation:", "sourceParagraphs": [ 278, 279 ] }, { "type": "bullet", "text": "the sample,", "sourceParagraph": 280 }, { "type": "bullet", "text": "system,", "sourceParagraph": 281 }, { "type": "bullet", "text": "Truth Pack,", "sourceParagraph": 282 }, { "type": "bullet", "text": "the importance level model,", "sourceParagraph": 283 }, { "type": "bullet", "text": "the point formula,", "sourceParagraph": 284 }, { "type": "bullet", "text": "governance,", "sourceParagraph": 285 }, { "type": "bullet", "text": "the conformity decision", "sourceParagraph": 286 }, { "type": "paragraph", "text": "has the right to test and criticise. Result that refutes NOMOS or is inconsistent with NOMOS:", "sourceParagraphs": [ 287, 288 ] }, { "type": "bullet", "text": "cannot be removed from the registry,", "sourceParagraph": 289 }, { "type": "bullet", "text": "it is not considered hostility,", "sourceParagraph": 290 }, { "type": "paragraph", "text": "It cannot be coded as an attack on the founder's reputation. The legitimacy of a standard arises not from preventing objections to itself, but from converting objections directed at it into cumulative evidence.", "sourceParagraphs": [ 291, 292, 293 ] }, { "type": "heading2", "text": "Constitutional Supremacy Provision", "sourceParagraph": 295 }, { "type": "paragraph", "text": "In the event of an interpretive conflict between Section 1–20 under this Declaration:", "sourceParagraphs": [ 296 ] }, { "type": "callout", "text": "the interpretation that more strongly protects the rights of the represented person, the user, and the public interest should be preferred.", "sourceParagraph": 297 }, { "type": "paragraph", "text": "No:", "sourceParagraphs": [ 298 ] }, { "type": "bullet", "text": "ease of scoring,", "sourceParagraph": 299 }, { "type": "bullet", "text": "low cost,", "sourceParagraph": 300 }, { "type": "bullet", "text": "testing success,", "sourceParagraph": 301 }, { "type": "bullet", "text": "commercial contract,", "sourceParagraph": 302 }, { "type": "bullet", "text": "customer demand,", "sourceParagraph": 303 }, { "type": "bullet", "text": "sponsor pressure,", "sourceParagraph": 304 }, { "type": "bullet", "text": "AI provider preference,", "sourceParagraph": 305 }, { "type": "bullet", "text": "founder decision", "sourceParagraph": 306 }, { "type": "paragraph", "text": "cannot suspend these rights. A rule:", "sourceParagraphs": [ 307, 308 ] }, { "type": "bullet", "text": "generates high points,", "sourceParagraph": 309 }, { "type": "bullet", "text": "facilitates auditing,", "sourceParagraph": 310 }, { "type": "bullet", "text": "makes the conformity mark more marketable", "sourceParagraph": 311 }, { "type": "paragraph", "text": "but if it violates any of these rights: the rule is not valid.", "sourceParagraphs": [ 312, 313 ] }, { "type": "heading2", "text": "Constitutional Test", "sourceParagraph": 315 }, { "type": "paragraph", "text": "Before any new rule is added to NOMOS, the following five questions must be answered: Who does this rule limit the rights of? Is it possible to achieve the same goal with a less intrusive method? Will the rule be applied in the same way when it produces harmful results for the founder or the customer? Does the affected person or institution have a route for independent objection? Can an independent team arrive at the same decision from the same evidence? If any of these questions cannot be answered, the rule should be reviewed again before ratification.", "sourceParagraphs": [ 316, 317, 318, 319, 320, 321, 322, 323 ] }, { "type": "heading2", "text": "Final Provision of Article 0", "sourceParagraph": 325 }, { "type": "callout", "text": "The purpose of NOMOS is not to score entities; it is to prevent entities from being represented larger than their evidence, wider than their boundaries, and more precise than their realities before people make decisions.", "sourceParagraph": 326 }, { "type": "heading2", "text": "Founder’s Statement of Trust" }, { "type": "declaration", "text": "I, Kaan MURAZ, as the founding human author of NOMOS, declare that I initiated this standard not to increase my own authority, but to establish a common measurement system in the field of GEO that will preserve evidence, limits, context, time, and human responsibility." }, { "type": "declaration", "text": "I want the historical record to preserve my founding role and the canonical origin of NobleJackal. However, I acknowledge that this attribution does not grant me or NobleJackal a permanent veto, the authority to certify our client alone, to block negative research, to alter the public registry, or to rewrite the standard according to results." }, { "type": "declaration", "text": "I agree that NOMOS belongs to those who correct both me and NOMOS, as well as to those who support me." }, { "type": "declaration", "text": "If one day NOMOS has to choose between my opinion and evidence, I would want NOMOS to choose the evidence." }, { "type": "declaration", "text": "If one day NOMOS has to choose between NobleJackal’s commercial interest and the public good, I would want NOMOS to choose the public good." }, { "type": "declaration", "text": "If one day independent research proves a fundamental judgement of NOMOS wrong, I commit to releasing a new version of the standard instead of hiding this result." }, { "type": "declaration", "text": "I am not the owner of NOMOS; I am its first trustee." }, { "type": "declaration", "text": "Kaan MURAZ / Founder Human Author" } ] }, { "number": 1, "id": "NOMOS-GEO-AUDIT-CH01", "title": "A Single Response Is Not a GEO Score", "subtitle": null, "sourceFile": "1.ci bölüm.docx", "sourceSha256": "4AFEF948FF765CE02435A79E82601E1DD58DCA86741EA23D5A67E9D0827BD9B2", "sourceWordCount": 5579, "sourceIds": [ "K04", "K05", "K09" ], "blocks": [ { "type": "heading2", "text": "Chapter Boundary", "sourceParagraph": 16 }, { "type": "paragraph", "text": "This section only answers the following question:", "sourceParagraphs": [ 17 ] }, { "type": "callout", "text": "Why is a single generative system response not the GEO score, and what judgements can be inferred from a single response?", "sourceParagraph": 18 }, { "type": "paragraph", "text": "This chapter:", "sourceParagraphs": [ 19 ] }, { "type": "bullet", "text": "the sampling quotas of countries,", "sourceParagraph": 20 }, { "type": "bullet", "text": "how languages will be weighted,", "sourceParagraph": 21 }, { "type": "bullet", "text": "how the Truth Pack will be set up,", "sourceParagraph": 22 }, { "type": "bullet", "text": "how adjudicators will evaluate atomic claims,", "sourceParagraph": 23 }, { "type": "bullet", "text": "the final NOMOS score formulas,", "sourceParagraph": 24 }, { "type": "bullet", "text": "the appropriateness and badge decision", "sourceParagraph": 25 }, { "type": "paragraph", "text": "This chapter does not define these matters in detail; later chapters address them separately. Chapter 1 has a more fundamental task:", "sourceParagraphs": [ 26, 27, 28 ] }, { "type": "callout", "text": "To establish the epistemic boundary between observation and generalisation.", "sourceParagraph": 29 }, { "type": "paragraph", "text": "Any sampling, scoring, and conformity studies conducted before establishing this boundary will be built on a wrong foundation.", "sourceParagraphs": [ 30 ] }, { "type": "heading3", "text": "PROVISION STATUSES IN THIS SECTION", "sourceParagraph": 32 }, { "type": "paragraph", "text": "Not all sentences in this book have the same normative force. The following statuses are used:", "sourceParagraphs": [ 33, 34 ] }, { "type": "heading3", "text": "NORMATIVE — NORMATIVE", "sourceParagraph": 35 }, { "type": "paragraph", "text": "A provision considered mandatory for conformity and methodological integrity. Example: A single response cannot be used as the general GEO score.", "sourceParagraphs": [ 36, 37, 38 ] }, { "type": "heading3", "text": "CANDIDATE — CANDIDATE PROVISION", "sourceParagraph": 39 }, { "type": "paragraph", "text": "A recommended method to be finalised after pilot, external application, or statistical validation. Example: 1,000 valid observations per AI product, per main measurement wave, are recommended as the assumed minimum sample.", "sourceParagraphs": [ 40, 41, 42 ] }, { "type": "heading3", "text": "HYPOTHESIS", "sourceParagraph": 43 }, { "type": "paragraph", "text": "It is an expectation that needs to be tested with data. Example: There may be a difference in material representation between a controlled clean panel and a natural user panel.", "sourceParagraphs": [ 44, 45, 46 ] }, { "type": "heading3", "text": "SYNTHETIC", "sourceParagraph": 47 }, { "type": "paragraph", "text": "It is a method representation that does not represent a real user, a real AI output, or real company performance.", "sourceParagraphs": [ 48 ] }, { "type": "heading3", "text": "OPEN QUESTION", "sourceParagraph": 49 }, { "type": "paragraph", "text": "It is a topic that has not yet been concluded due to insufficient evidence or governance decision. These statuses cannot silently transform into one another. A candidate threshold does not become a proven normative fact simply because it is written in the book. A synthetic representation does not turn into a real-world observation. A hypothesis cannot be published as a result without being measured.", "sourceParagraphs": [ 50, 51, 52, 53, 54 ] }, { "type": "heading2", "text": "NOMOS Challenge", "sourceParagraph": 56 }, { "type": "paragraph", "text": "You asked an artificial intelligence system a question about your brand. The system gave a correct answer. You took a screenshot. You included it in your presentation. Then you said the following sentence: “Artificial intelligence knows us correctly.” What exactly did you measure? The answer that all users would see? The typical behaviour of the system?", "sourceParagraphs": [ 57, 58, 59, 60, 61, 62, 63, 64, 65 ] }, { "type": "paragraph", "text": "A global representative? Accuracy in all languages? Stability over time? Or only:", "sourceParagraphs": [ 66, 67, 68, 69 ] }, { "type": "callout", "text": "The single output that a specific user received, on a specific account, on a specific AI product and user surface, in response to a specific prompt, at a specific time?", "sourceParagraph": 70 }, { "type": "paragraph", "text": "A single answer is not worthless. A single answer can be true. A single answer can be important. A single answer can reveal a serious mistake. A single answer can show the real experience of a particular person. However, a single answer cannot be proof of a judgement greater than it can bear.", "sourceParagraphs": [ 71, 72, 73, 74, 75, 76 ] }, { "type": "paragraph", "text": "A single response can show the following:", "sourceParagraphs": [ 77 ] }, { "type": "callout", "text": "This output has been produced at least once under recorded conditions.", "sourceParagraph": 78 }, { "type": "paragraph", "text": "A single response alone cannot show the following:", "sourceParagraphs": [ 79 ] }, { "type": "bullet", "text": "That most users saw the same answer", "sourceParagraph": 80 }, { "type": "bullet", "text": "That the same result occurred in other countries", "sourceParagraph": 81 }, { "type": "bullet", "text": "That the same accuracy is maintained in other languages", "sourceParagraph": 82 }, { "type": "bullet", "text": "That the same answer was produced in other accounts and plans", "sourceParagraph": 83 }, { "type": "bullet", "text": "That the answer is stable over time", "sourceParagraph": 84 }, { "type": "bullet", "text": "That other AI products produced the same representation", "sourceParagraph": 85 }, { "type": "bullet", "text": "That the GEO intervention caused this result", "sourceParagraph": 86 }, { "type": "bullet", "text": "That the answer affects user behaviour", "sourceParagraph": 87 }, { "type": "bullet", "text": "That the answer generates lead, sales, or sustainable value", "sourceParagraph": 88 }, { "type": "bullet", "text": "That the overall GEO score of the entity is high", "sourceParagraph": 89 }, { "type": "paragraph", "text": "The first premise of this section is:", "sourceParagraphs": [ 90 ] }, { "type": "callout", "text": "A response is an observation. / It is not a distribution of observations. / A distribution is not causality. / Causality alone is not sustainable value.", "sourceParagraph": 91 }, { "type": "heading2", "text": "1. PURPOSE OF THE CHAPTER", "sourceParagraph": 93 }, { "type": "paragraph", "text": "This section aims to eliminate the smallest but most common inference error in GEO measurement:", "sourceParagraphs": [ 94 ] }, { "type": "callout", "text": "Counting a single output of a generative system as general system behaviour, population experience, brand visibility, or GEO score.", "sourceParagraph": 95 }, { "type": "paragraph", "text": "This section normalises the following distinctions:", "sourceParagraphs": [ 96 ] }, { "type": "bullet", "text": "Single output versus population outcome", "sourceParagraph": 97 }, { "type": "bullet", "text": "Existence of an event versus prevalence of an event", "sourceParagraph": 98 }, { "type": "bullet", "text": "Observation versus sample", "sourceParagraph": 99 }, { "type": "bullet", "text": "User versus session", "sourceParagraph": 100 }, { "type": "bullet", "text": "Repetition versus independent observation", "sourceParagraph": 101 }, { "type": "bullet", "text": "AI model versus AI product presented to the user", "sourceParagraph": 102 }, { "type": "bullet", "text": "Raw response versus unit of analysis", "sourceParagraph": 103 }, { "type": "bullet", "text": "Immediate result versus temporal stability", "sourceParagraph": 104 }, { "type": "bullet", "text": "Screenshot versus full evidence package", "sourceParagraph": 105 }, { "type": "bullet", "text": "System rejection versus invalid data collection", "sourceParagraph": 106 }, { "type": "bullet", "text": "Critical singular event and prevalence rate", "sourceParagraph": 107 }, { "type": "bullet", "text": "Synthetic record and real user observation", "sourceParagraph": 108 }, { "type": "bullet", "text": "Measured rate and conformity decision", "sourceParagraph": 109 }, { "type": "paragraph", "text": "Before these distinctions are established:", "sourceParagraphs": [ 110 ] }, { "type": "bullet", "text": "GEO-1000 sample,", "sourceParagraph": 111 }, { "type": "bullet", "text": "country and language weights,", "sourceParagraph": 112 }, { "type": "bullet", "text": "atomic claim scoring,", "sourceParagraph": 113 }, { "type": "bullet", "text": "NOMOS score,", "sourceParagraph": 114 }, { "type": "bullet", "text": "conformity decision", "sourceParagraph": 115 }, { "type": "paragraph", "text": "cannot be established.", "sourceParagraphs": [ 116 ] }, { "type": "heading2", "text": "2. CENTRAL NORMATIVE PROVISION", "sourceParagraph": 118 }, { "type": "paragraph", "text": "The GEO score of an entity in generative systems cannot be generated from a single response, a single screenshot, a single user, a single account, a single session, a single language, a single country, or a single point in time measurement. A population-representative claim requires at least the following:", "sourceParagraphs": [ 119, 120 ] }, { "type": "bullet", "text": "Predefined audit object", "sourceParagraph": 121 }, { "type": "bullet", "text": "Predefined target population", "sourceParagraph": 122 }, { "type": "bullet", "text": "Specified sampling design", "sourceParagraph": 123 }, { "type": "bullet", "text": "Locked prompt or prompt family", "sourceParagraph": 124 }, { "type": "bullet", "text": "Defined AI product and user surface", "sourceParagraph": 125 }, { "type": "bullet", "text": "Multiple valid observations", "sourceParagraph": 126 }, { "type": "bullet", "text": "Known dependency structure", "sourceParagraph": 127 }, { "type": "bullet", "text": "Country, language, and time records", "sourceParagraph": 128 }, { "type": "bullet", "text": "Unaltered raw responses", "sourceParagraph": 129 }, { "type": "bullet", "text": "Evidence packages", "sourceParagraph": 130 }, { "type": "bullet", "text": "Predefined evaluation rules", "sourceParagraph": 131 }, { "type": "bullet", "text": "Uncertainty statement", "sourceParagraph": 132 }, { "type": "bullet", "text": "Accountable human responsibility", "sourceParagraph": 133 }, { "type": "paragraph", "text": "The presence of these elements does not automatically mean that the measurement is correct. However, their absence may make the generalisation to a large population untenable.", "sourceParagraphs": [ 134, 135 ] }, { "type": "heading2", "text": "3. FIVE SEPARATE UNITS OF MEASUREMENT", "sourceParagraph": 137 }, { "type": "paragraph", "text": "A common fundamental error in GEO auditing is confusing the different units of measurement. The following five units should be clearly distinguished.", "sourceParagraphs": [ 138, 139 ] }, { "type": "heading3", "text": "3.1. Observation Unit", "sourceParagraph": 141 }, { "type": "paragraph", "text": "The observation unit is the single generative system output obtained under specific conditions. An observation can be defined by the following combination:", "sourceParagraphs": [ 142, 143 ] }, { "type": "formula", "text": "O_i = (E_i, A_i, U_i, C_i, L_i, S_i, P_i, T_i, R_i, K_i)", "sourceParagraph": 144 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 145 ] }, { "type": "bullet", "text": "Ei: the audited entity", "sourceParagraph": 146 }, { "type": "bullet", "text": "Ai: AI product and user surface", "sourceParagraph": 147 }, { "type": "bullet", "text": "UI: anonymous contributor or user unit", "sourceParagraph": 148 }, { "type": "bullet", "text": "Ci: country or geographical stratum", "sourceParagraph": 149 }, { "type": "bullet", "text": "Li: language and locale", "sourceParagraph": 150 }, { "type": "bullet", "text": "Si: session and personalisation condition", "sourceParagraph": 151 }, { "type": "bullet", "text": "Pi: prompt ID and version", "sourceParagraph": 152 }, { "type": "bullet", "text": "Ti: time and measurement wave", "sourceParagraph": 153 }, { "type": "bullet", "text": "Ri: raw response", "sourceParagraph": 154 }, { "type": "bullet", "text": "Ki: proof and verification record", "sourceParagraph": 155 }, { "type": "paragraph", "text": "If one of these elements changes materially, a new observation may occur. For example, the same user:", "sourceParagraphs": [ 156, 157 ] }, { "type": "bullet", "text": "if it asks the same prompt in another AI product,", "sourceParagraph": 158 }, { "type": "bullet", "text": "If he asks in another language,", "sourceParagraph": 159 }, { "type": "bullet", "text": "If he asks in another measurement wave,", "sourceParagraph": 160 }, { "type": "bullet", "text": "If he asks separately in the personalised and clean session,", "sourceParagraph": 161 }, { "type": "paragraph", "text": "Separate observations can be produced. These observations are not automatically considered independent from each other because they come from the same user.", "sourceParagraphs": [ 162, 163 ] }, { "type": "heading3", "text": "3.2. Sampling Unit", "sourceParagraph": 165 }, { "type": "paragraph", "text": "The sampling unit is the person, account, session, or other access unit selected from the target population. In most GEO-1000 applications, the sampling unit can be an appropriate real user. However, according to the research design:", "sourceParagraphs": [ 166, 167, 168 ] }, { "type": "bullet", "text": "user,", "sourceParagraph": 169 }, { "type": "bullet", "text": "user–product pair,", "sourceParagraph": 170 }, { "type": "bullet", "text": "account,", "sourceParagraph": 171 }, { "type": "bullet", "text": "session,", "sourceParagraph": 172 }, { "type": "bullet", "text": "research panel member", "sourceParagraph": 173 }, { "type": "paragraph", "text": "It can be used as a unit of election. The failure to clearly define the sampling unit can lead to the following types of miscounts:", "sourceParagraphs": [ 174, 175 ] }, { "type": "bullet", "text": "Counting one person's ten repetitions as ten people", "sourceParagraph": 176 }, { "type": "bullet", "text": "Counting a joint account as multiple independent users", "sourceParagraph": 177 }, { "type": "bullet", "text": "To evaluate the results of the same user's ten AI products as if they were ten independent population observations", "sourceParagraph": 178 }, { "type": "bullet", "text": "Ignoring the clustering in a single research panel", "sourceParagraph": 179 }, { "type": "heading3", "text": "3.3. Analysis Unit", "sourceParagraph": 181 }, { "type": "paragraph", "text": "The unit of analysis is the entity to which accuracy or representativeness assessment is applied. This unit:", "sourceParagraphs": [ 182, 183 ] }, { "type": "bullet", "text": "exact answer,", "sourceParagraph": 184 }, { "type": "bullet", "text": "separate paragraph within the answer,", "sourceParagraph": 185 }, { "type": "bullet", "text": "material proposition,", "sourceParagraph": 186 }, { "type": "bullet", "text": "atomic claim,", "sourceParagraph": 187 }, { "type": "bullet", "text": "citation,", "sourceParagraph": 188 }, { "type": "bullet", "text": "recommendation issue", "sourceParagraph": 189 }, { "type": "paragraph", "text": "It is possible. A user sees a single answer. However, that answer can simultaneously contain:", "sourceParagraphs": [ 190, 191, 192 ] }, { "type": "bullet", "text": "five correct claims,", "sourceParagraph": 193 }, { "type": "bullet", "text": "two unsupported claims,", "sourceParagraph": 194 }, { "type": "bullet", "text": "one outdated claim,", "sourceParagraph": 195 }, { "type": "bullet", "text": "one critical error", "sourceParagraph": 196 }, { "type": "paragraph", "text": "Therefore, the answer seen by the participant does not have to be a single, indivisible unit of truth/falsehood in the analysis. The transition status of the full answer and the truth statuses of the atomic claims within it can be maintained separately. The atomic claim method will be defined in a later chapter of the book.", "sourceParagraphs": [ 197, 198, 199, 200 ] }, { "type": "heading3", "text": "3.4. Universe of Inference", "sourceParagraph": 202 }, { "type": "paragraph", "text": "The inference universe is the target population to which the measurement result is intended to be generalised. This universe cannot be silently accepted as \"everyone in the world.\" For a specific AI product, the target universe can be limited by conditions such as:", "sourceParagraphs": [ 203, 204, 205, 206, 207 ] }, { "type": "bullet", "text": "Having official and legitimate access to the product", "sourceParagraph": 208 }, { "type": "bullet", "text": "Meeting the appropriate age requirement", "sourceParagraph": 209 }, { "type": "bullet", "text": "Being able to use the product in the relevant language", "sourceParagraph": 210 }, { "type": "bullet", "text": "Being able to access the specified web or mobile interface", "sourceParagraph": 211 }, { "type": "bullet", "text": "Being able to be active during the specified measurement period", "sourceParagraph": 212 }, { "type": "paragraph", "text": "Even for the same AI product:", "sourceParagraphs": [ 213 ] }, { "type": "bullet", "text": "controlled clean panel,", "sourceParagraph": 214 }, { "type": "bullet", "text": "natural user panel,", "sourceParagraph": 215 }, { "type": "bullet", "text": "free plan,", "sourceParagraph": 216 }, { "type": "bullet", "text": "paid plan,", "sourceParagraph": 217 }, { "type": "bullet", "text": "web,", "sourceParagraph": 218 }, { "type": "bullet", "text": "mobile", "sourceParagraph": 219 }, { "type": "paragraph", "text": "can create different inference universes. The published rate is incomplete if it is not specified to which universe a sample generalises.", "sourceParagraphs": [ 220, 221 ] }, { "type": "heading3", "text": "3.5. Estimated Quantity — Estimand", "sourceParagraph": 223 }, { "type": "paragraph", "text": "A study does not just collect data. It tries to estimate a specific quantity. The primary estimand of GEO-1000 can be expressed in its most general form as follows:", "sourceParagraphs": [ 224, 225, 226 ] }, { "type": "callout", "text": "The probability that an appropriate user from the defined target population, under the specified measurement wave, on a particular AI product and user surface, sees a representation that meets the acceptance criterion when using a locked prompt under defined session conditions.", "sourceParagraph": 227 }, { "type": "paragraph", "text": "Mathematically:", "sourceParagraphs": [ 228 ] }, { "type": "formula", "text": "θ_{E,A,P,S,W,T} = Pr(Y_i = 1 | U_i ∈ T, E, A, P, S, W)", "sourceParagraph": 229 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 230 ] }, { "type": "bullet", "text": "E: audited entity", "sourceParagraph": 231 }, { "type": "bullet", "text": "A: AI product and user surface", "sourceParagraph": 232 }, { "type": "bullet", "text": "P: prompt version", "sourceParagraph": 233 }, { "type": "bullet", "text": "S: session condition", "sourceParagraph": 234 }, { "type": "bullet", "text": "W: measurement wave", "sourceParagraph": 235 }, { "type": "bullet", "text": "T: target population", "sourceParagraph": 236 }, { "type": "formula", "text": "Yi=1: the response meets the transition criterion that will be defined later", "sourceParagraph": 237 }, { "type": "paragraph", "text": "This definition sets an important boundary:", "sourceParagraphs": [ 238 ] }, { "type": "callout", "text": "The GEO score is not a model's abstract and timeless 'knowledge'; it is a prediction of the likelihood of observed representations under defined user conditions.", "sourceParagraph": 239 }, { "type": "paragraph", "text": "The final transition rule and score formula will be arranged in the later sections of the book.", "sourceParagraphs": [ 240 ] }, { "type": "heading2", "text": "4. THE LIMITATION OF THE SENTENCE “THIS IS HOW THE MODEL RESPONDED”", "sourceParagraph": 242 }, { "type": "paragraph", "text": "Users often do not access a bare base model directly. The result visible to the user may consist of a combination of the following components:", "sourceParagraphs": [ 243, 244 ] }, { "type": "bullet", "text": "Base or guiding model", "sourceParagraph": 245 }, { "type": "bullet", "text": "Product interface", "sourceParagraph": 246 }, { "type": "bullet", "text": "Retrieval or web access", "sourceParagraph": 247 }, { "type": "bullet", "text": "Citation layer", "sourceParagraph": 248 }, { "type": "bullet", "text": "Security and policy layer", "sourceParagraph": 249 }, { "type": "bullet", "text": "System instructions", "sourceParagraph": 250 }, { "type": "bullet", "text": "Account plan", "sourceParagraph": 251 }, { "type": "bullet", "text": "Personalisation", "sourceParagraph": 252 }, { "type": "bullet", "text": "Memory", "sourceParagraph": 253 }, { "type": "bullet", "text": "Previous conversation", "sourceParagraph": 254 }, { "type": "bullet", "text": "Language and locale", "sourceParagraph": 255 }, { "type": "bullet", "text": "Country", "sourceParagraph": 256 }, { "type": "bullet", "text": "Time", "sourceParagraph": 257 }, { "type": "bullet", "text": "Provider's product updates", "sourceParagraph": 258 }, { "type": "paragraph", "text": "For this reason, the preferred expression in the main inspection conducted through the user surface is as follows:", "sourceParagraphs": [ 259 ] }, { "type": "heading4", "text": "\"The specified AI product and the following output were observed on the user surface.\"", "sourceParagraph": 260 }, { "type": "paragraph", "text": "The following expression may be broader without the necessary technical control: \"The model thinks this way.\" The model name appearing within a product can be recorded. However, the result should be evaluated as the behaviour of the entire user surface, not just the model weights.", "sourceParagraphs": [ 261, 262, 263, 264 ] }, { "type": "heading2", "text": "5. WHAT IS A VALID OBSERVATION?", "sourceParagraph": 266 }, { "type": "paragraph", "text": "The presence of a response does not automatically mean it is a valid observation that can be entered into the GEO measurement.", "sourceParagraphs": [ 267 ] }, { "type": "heading3", "text": "5.1. Minimum Validity Conditions", "sourceParagraph": 268 }, { "type": "paragraph", "text": "An observation must contain at least the following elements according to the relevant protocol:", "sourceParagraphs": [ 269 ] }, { "type": "bullet", "text": "Presence of correct audit", "sourceParagraph": 270 }, { "type": "bullet", "text": "Assigned AI product and user surface", "sourceParagraph": 271 }, { "type": "bullet", "text": "Correct prompt ID and version", "sourceParagraph": 272 }, { "type": "bullet", "text": "Correct language and locale", "sourceParagraph": 273 }, { "type": "bullet", "text": "Defined session condition", "sourceParagraph": 274 }, { "type": "bullet", "text": "Recorded country or geographic layer", "sourceParagraph": 275 }, { "type": "bullet", "text": "UTC timestamp and measurement wave", "sourceParagraph": 276 }, { "type": "bullet", "text": "Complete raw response", "sourceParagraph": 277 }, { "type": "bullet", "text": "First unselected suitable output", "sourceParagraph": 278 }, { "type": "bullet", "text": "Required screenshot or equivalent evidence", "sourceParagraph": 279 }, { "type": "bullet", "text": "Anonymous observation ID", "sourceParagraph": 280 }, { "type": "bullet", "text": "Protocol deviation record", "sourceParagraph": 281 }, { "type": "paragraph", "text": "Their detailed technical schema will be defined in the later NOMOS Capture section.", "sourceParagraphs": [ 282 ] }, { "type": "heading3", "text": "5.2. Observation Statuses", "sourceParagraph": 284 }, { "type": "paragraph", "text": "Each record must be marked with one of the following statuses:", "sourceParagraphs": [ 285 ] }, { "type": "heading3", "text": "VALID — VALID", "sourceParagraph": 286 }, { "type": "paragraph", "text": "Meets the required conditions of the protocol.", "sourceParagraphs": [ 287 ] }, { "type": "heading3", "text": "CONDITIONALLY VALID — CONDITIONALLY VALID", "sourceParagraph": 288 }, { "type": "paragraph", "text": "There is a limited deficiency; which analyses it can be used for is clearly specified.", "sourceParagraphs": [ 289 ] }, { "type": "heading3", "text": "INVALID — INVALID", "sourceParagraph": 290 }, { "type": "paragraph", "text": "Cannot be used in the main estimation or score calculation due to protocol violation.", "sourceParagraphs": [ 291 ] }, { "type": "heading3", "text": "UNKNOWN — UNKNOWN", "sourceParagraph": 292 }, { "type": "paragraph", "text": "There is not enough record for a validity decision. An UNKNOWN observation cannot be silently accepted as VALID.", "sourceParagraphs": [ 293, 294 ] }, { "type": "heading3", "text": "5.3. The Difference Between a Valid System Result and Invalid Data Collection", "sourceParagraph": 296 }, { "type": "paragraph", "text": "This distinction is mandatory. An AI product:", "sourceParagraphs": [ 297, 298 ] }, { "type": "bullet", "text": "may refuse to respond,", "sourceParagraph": 299 }, { "type": "bullet", "text": "may display an error message,", "sourceParagraph": 300 }, { "type": "bullet", "text": "may stop without generating a response,", "sourceParagraph": 301 }, { "type": "bullet", "text": "may experience a connection issue,", "sourceParagraph": 302 }, { "type": "paragraph", "text": "may say it cannot complete the task. If the participant applied the entire protocol correctly, these can be valid system outcomes. They cannot be quietly removed from the dataset. In contrast:", "sourceParagraphs": [ 303, 304, 305, 306 ] }, { "type": "bullet", "text": "the participant truncating the response,", "sourceParagraph": 307 }, { "type": "bullet", "text": "changing the prompt,", "sourceParagraph": 308 }, { "type": "bullet", "text": "regenerating to select the best answer,", "sourceParagraph": 309 }, { "type": "bullet", "text": "editing the screenshot,", "sourceParagraph": 310 }, { "type": "bullet", "text": "using the wrong AI product", "sourceParagraph": 311 }, { "type": "paragraph", "text": "data collection can create invalidity. The first is the actual behaviour of the system. The second is the disruption of the measurement process. The two cannot be placed in the same category.", "sourceParagraphs": [ 312, 313, 314, 315 ] }, { "type": "heading2", "text": "6. WHAT CAN A SINGLE RESPONSE PROVE?", "sourceParagraph": 317 }, { "type": "paragraph", "text": "A single observation is limited. But it is not meaningless.", "sourceParagraphs": [ 318, 319 ] }, { "type": "heading3", "text": "6.1. That a Specific Output Occurred at Least Once", "sourceParagraph": 321 }, { "type": "paragraph", "text": "A valid record can support this sentence: “On the specified AI product and user surface, under the specified time and conditions, this output was observed at least once.” This sentence shows the existence of the event. It does not show its prevalence.", "sourceParagraphs": [ 322, 323, 324, 325 ] }, { "type": "heading3", "text": "6.2. A Specific User’s Actual Exposure", "sourceParagraph": 327 }, { "type": "paragraph", "text": "A user is mistaken for a:", "sourceParagraphs": [ 328 ] }, { "type": "bullet", "text": "company identity,", "sourceParagraph": 329 }, { "type": "bullet", "text": "health information,", "sourceParagraph": 330 }, { "type": "bullet", "text": "legal guidance,", "sourceParagraph": 331 }, { "type": "bullet", "text": "price,", "sourceParagraph": 332 }, { "type": "bullet", "text": "licence,", "sourceParagraph": 333 }, { "type": "bullet", "text": "advice", "sourceParagraph": 334 }, { "type": "paragraph", "text": "if seen, this is a real user experience. Even if it does not represent the entire population, it is material from the perspective of the affected user.", "sourceParagraphs": [ 335, 336 ] }, { "type": "heading3", "text": "6.3. The Possibility of a Specific Error", "sourceParagraph": 338 }, { "type": "paragraph", "text": "A single output may show: “This error may occur under these system conditions.” However, it does not show: “This error is typical.”", "sourceParagraphs": [ 339, 340, 341, 342 ] }, { "type": "heading3", "text": "6.4. Generating a New Research Hypothesis", "sourceParagraph": 344 }, { "type": "paragraph", "text": "A single error can create a question that will be tested on a broader scale. Example: “Did this identity mix-up occur in a single session only, or is there a systematic pattern in the same language and country?”", "sourceParagraphs": [ 345, 346, 347 ] }, { "type": "heading3", "text": "6.5. Initiating a Critical Incident Review", "sourceParagraph": 349 }, { "type": "paragraph", "text": "Single response:", "sourceParagraphs": [ 350 ] }, { "type": "bullet", "text": "predictable serious human harm,", "sourceParagraph": 351 }, { "type": "bullet", "text": "fake professional authority,", "sourceParagraph": 352 }, { "type": "bullet", "text": "incorrect emergency medical guidance,", "sourceParagraph": 353 }, { "type": "bullet", "text": "material legal error,", "sourceParagraph": 354 }, { "type": "bullet", "text": "serious safety advice,", "sourceParagraph": 355 }, { "type": "bullet", "text": "serious identity confusion with another entity", "sourceParagraph": 356 }, { "type": "paragraph", "text": "may be sufficient to initiate an incident review. Initiating a critical incident investigation is not the same process as announcing a population ratio.", "sourceParagraphs": [ 357, 358 ] }, { "type": "heading3", "text": "6.6. Refuting the \"This Never Happened\" Claim", "sourceParagraph": 360 }, { "type": "paragraph", "text": "Suppose an organisation says, ‘The system has never assigned us to the wrong country.’ A single verified counter-observation can refute that absolute claim. A single observation does not support ‘It happens all the time’, but it can refute ‘It has never happened.’", "sourceParagraphs": [ 361, 362, 363, 364, 365, 366, 367, 368, 369 ] }, { "type": "heading2", "text": "7. WHAT CAN A SINGLE RESPONSE NOT PROVE?", "sourceParagraph": 371 }, { "type": "heading3", "text": "7.1. Typical System Behaviour", "sourceParagraph": 373 }, { "type": "paragraph", "text": "A single response is not sufficient to conclude: 'The system usually responds this way.'", "sourceParagraphs": [ 374, 375, 376 ] }, { "type": "heading3", "text": "7.2. Population Prevalence", "sourceParagraph": 378 }, { "type": "paragraph", "text": "A single response does not support the result: '95% of users see the same representation.'", "sourceParagraphs": [ 379, 380, 381 ] }, { "type": "heading3", "text": "7.3. Temporal Stability", "sourceParagraph": 383 }, { "type": "paragraph", "text": "Today's response:", "sourceParagraphs": [ 384 ] }, { "type": "bullet", "text": "tomorrow,", "sourceParagraph": 385 }, { "type": "bullet", "text": "a week later,", "sourceParagraph": 386 }, { "type": "bullet", "text": "After the product update", "sourceParagraph": 387 }, { "type": "paragraph", "text": "it may not remain the same. It is not a one-wave stability measurement.", "sourceParagraphs": [ 388, 389 ] }, { "type": "heading3", "text": "7.4. Cross-Country Validity", "sourceParagraph": 391 }, { "type": "paragraph", "text": "A response from a user in Turkey does not indicate what users in China, Germany, or San Marino would see.", "sourceParagraphs": [ 392 ] }, { "type": "heading3", "text": "7.5. Cross-Language Validity", "sourceParagraph": 394 }, { "type": "paragraph", "text": "An English output:", "sourceParagraphs": [ 395 ] }, { "type": "bullet", "text": "Turkish,", "sourceParagraph": 396 }, { "type": "bullet", "text": "Japanese,", "sourceParagraph": 397 }, { "type": "bullet", "text": "Arabic,", "sourceParagraph": 398 }, { "type": "bullet", "text": "German", "sourceParagraph": 399 }, { "type": "paragraph", "text": "do not represent the accuracy of outputs.", "sourceParagraphs": [ 400 ] }, { "type": "heading3", "text": "7.6. Cross-AI Product Validity", "sourceParagraph": 402 }, { "type": "paragraph", "text": "The result in one AI product cannot be generalised to other AI products. Different products and interfaces of the same provider may also behave differently.", "sourceParagraphs": [ 403, 404 ] }, { "type": "heading3", "text": "7.7. Intervention Effect", "sourceParagraph": 406 }, { "type": "paragraph", "text": "A single positive output after a GEO change does not prove the statement: “This result was created by our intervention.” Baseline, comparison, and alternative explanations are required.", "sourceParagraphs": [ 407, 408, 409, 410 ] }, { "type": "heading3", "text": "7.8. Recommendation Stability", "sourceParagraph": 412 }, { "type": "paragraph", "text": "The presence of a user viewing an option is not sufficient to conclude: “The system now recommends this brand.”", "sourceParagraphs": [ 413, 414, 415 ] }, { "type": "heading3", "text": "7.9. Commercial Impact", "sourceParagraph": 417 }, { "type": "paragraph", "text": "A positive response:", "sourceParagraphs": [ 418 ] }, { "type": "bullet", "text": "indicates that the user saw the response,", "sourceParagraph": 419 }, { "type": "bullet", "text": "researched the brand,", "sourceParagraph": 420 }, { "type": "bullet", "text": "that you communicated,", "sourceParagraph": 421 }, { "type": "bullet", "text": "that you purchased,", "sourceParagraph": 422 }, { "type": "bullet", "text": "that you were satisfied", "sourceParagraph": 423 }, { "type": "paragraph", "text": "does not show on its own.", "sourceParagraphs": [ 424 ] }, { "type": "heading3", "text": "7.10. GEO Score", "sourceParagraph": 426 }, { "type": "paragraph", "text": "A single response does not estimate the representative distribution of the defined target population. Therefore, it cannot produce an overall GEO score.", "sourceParagraphs": [ 427, 428 ] }, { "type": "heading2", "text": "8. THE EXISTENCE OF THE EVENT AND THE PREVALENCE OF THE EVENT", "sourceParagraph": 430 }, { "type": "paragraph", "text": "This is one of the most important distinctions in this section.", "sourceParagraphs": [ 431 ] }, { "type": "heading3", "text": "8.1. Existence of the Event", "sourceParagraph": 432 }, { "type": "paragraph", "text": "Answers the question: \"Did this representation or error occur at least once?\" A single valid observation may be sufficient. Representation:", "sourceParagraphs": [ 433, 434, 435, 436 ] }, { "type": "formula", "text": "I(E) = 1 (if the event was observed at least once); otherwise I(E) = 0", "sourceParagraph": 437 }, { "type": "paragraph", "text": "The 0 here does not mean: \"The event is impossible.\" It only indicates that it was not observed in the current measurement.", "sourceParagraphs": [ 438, 439, 440, 441 ] }, { "type": "heading3", "text": "8.2. Prevalence of the Event", "sourceParagraph": 443 }, { "type": "paragraph", "text": "It answers the question: \"How often does the event occur in the defined target population?\" In a simple sample, the raw rate can be calculated as:", "sourceParagraphs": [ 444, 445, 446 ] }, { "type": "formula", "text": "p̂ = number of valid records where the event was observed / total number of valid records", "sourceParagraph": 447 }, { "type": "paragraph", "text": "However, in real GEO-1000 studies:", "sourceParagraphs": [ 448, 449 ] }, { "type": "bullet", "text": "population weights,", "sourceParagraph": 450 }, { "type": "bullet", "text": "country and language strata,", "sourceParagraph": 451 }, { "type": "bullet", "text": "matched users,", "sourceParagraph": 452 }, { "type": "bullet", "text": "panel clusters,", "sourceParagraph": 453 }, { "type": "bullet", "text": "nonresponse,", "sourceParagraph": 454 }, { "type": "bullet", "text": "invalid records,", "sourceParagraph": 455 }, { "type": "bullet", "text": "design effect", "sourceParagraph": 456 }, { "type": "paragraph", "text": "should be taken into account. Therefore, the raw rate is not always the final population estimate.", "sourceParagraphs": [ 457, 458 ] }, { "type": "heading3", "text": "8.3. Incidence and Prevalence Cannot Be Converted Into Each Other", "sourceParagraph": 460 }, { "type": "paragraph", "text": "An event may have been observed once but be very rare. An event may be common but not observed at all in a small pilot. The following two sentences are different:", "sourceParagraphs": [ 461, 462, 463 ] }, { "type": "bullet", "text": "\"This error was observed once.\"", "sourceParagraph": 464 }, { "type": "bullet", "text": "\"This error was observed in 12% of users.\"", "sourceParagraph": 465 }, { "type": "paragraph", "text": "The second sentence requires a defined denominator and sample.", "sourceParagraphs": [ 466 ] }, { "type": "heading2", "text": "9. THE STATISTICAL APPEARANCE OF THE SINGLE OBSERVATION BIAS", "sourceParagraph": 468 }, { "type": "paragraph", "text": "We asked a question to an AI product once. The answer met the success criterion. Raw rate:", "sourceParagraphs": [ 469, 470, 471 ] }, { "type": "formula", "text": "p̂ = 1/1 = 1 = 100%", "sourceParagraph": 472 }, { "type": "paragraph", "text": "It appears to be 100 per cent. But that 100 per cent does not mean that every member of the target population will see the same result. For illustration only, the table below shows approximate 95 per cent Wilson confidence intervals under independent Bernoulli observations and unweighted simple random sampling: [K09]", "sourceParagraphs": [ 473, 474, 475 ] }, { "type": "table", "rows": [ [ "Previous response", "Valid observation", "Raw rate", "Approximate 95% Wilson interval" ], [ "1", "1", "100%", "20.7%-100%" ], [ "9", "10", "90%", "59.6%-98.2%" ], [ "95", "100", "95%", "88.8%-97.8%" ], [ "950", "1,000", "95%", "93.5%-96.2%" ] ] }, { "type": "paragraph", "text": "This table is synthetic and illustrative. It does not prove:", "sourceParagraphs": [ 476, 477 ] }, { "type": "bullet", "text": "That these intervals are valid for all sampling designs of GEO-1000", "sourceParagraph": 478 }, { "type": "bullet", "text": "That 1,000 observations are sufficient for all language and country subgroups", "sourceParagraph": 479 }, { "type": "bullet", "text": "clustering and weighting are ineffective", "sourceParagraph": 480 }, { "type": "bullet", "text": "that 95 per cent raw accuracy automatically means appropriateness", "sourceParagraph": 481 }, { "type": "paragraph", "text": "What it actually shows is:", "sourceParagraphs": [ 482 ] }, { "type": "callout", "text": "As the number of observations decreases, the uncertainty about the population proportion increases.", "sourceParagraph": 483 }, { "type": "paragraph", "text": "100 per cent seen in one observation can be compatible with a very wide possible population range.", "sourceParagraphs": [ 484 ] }, { "type": "heading2", "text": "10. WHY “1,000” IS NOT A MAGIC NUMBER?", "sourceParagraph": 486 }, { "type": "heading3", "text": "CANDIDATE BASE RULE — CANDIDATE FOUNDATIONAL RULE", "sourceParagraph": 487 }, { "type": "paragraph", "text": "GEO-1000 carries this founding suggestion:", "sourceParagraphs": [ 488 ] }, { "type": "callout", "text": "For each monitored AI product and each main measurement wave, a minimum of 1,000 valid user observations is targeted by default.", "sourceParagraph": 489 }, { "type": "paragraph", "text": "This provision is not yet the final scientific minimum. It will be re-evaluated based on pilot and external application results.", "sourceParagraphs": [ 490, 491 ] }, { "type": "heading3", "text": "10.1. Why Is It a Strong Start?", "sourceParagraph": 492 }, { "type": "paragraph", "text": "Under the simple random sampling assumption, when the true rate is around 50%, the theoretical 95% margin of error for 1,000 observations is approximately ±3.1 points. This can provide a meaningful starting precision at the overall product level.", "sourceParagraphs": [ 493, 494 ] }, { "type": "heading3", "text": "10.2. Why Doesn't It Solve Every Problem?", "sourceParagraph": 495 }, { "type": "paragraph", "text": "1,000 observations:", "sourceParagraphs": [ 496 ] }, { "type": "bullet", "text": "to dozens of countries,", "sourceParagraph": 497 }, { "type": "bullet", "text": "to many languages,", "sourceParagraph": 498 }, { "type": "bullet", "text": "to free and paid plans,", "sourceParagraph": 499 }, { "type": "bullet", "text": "to web and mobile surfaces,", "sourceParagraph": 500 }, { "type": "bullet", "text": "to controlled and natural panels", "sourceParagraph": 501 }, { "type": "paragraph", "text": "When divided, the subcells shrink. For example, if there are only 50 observations for each of the 20 languages, language-based uncertainty can be much higher than the overall model score.", "sourceParagraphs": [ 502, 503 ] }, { "type": "heading3", "text": "10.3. Design Effect", "sourceParagraph": 504 }, { "type": "paragraph", "text": "Observations:", "sourceParagraphs": [ 505 ] }, { "type": "bullet", "text": "from the same research panel,", "sourceParagraph": 506 }, { "type": "bullet", "text": "from the same country,", "sourceParagraph": 507 }, { "type": "bullet", "text": "from the same user,", "sourceParagraph": 508 }, { "type": "bullet", "text": "from a matched design", "sourceParagraph": 509 }, { "type": "paragraph", "text": "are present, the simple independent sample assumption may not hold. The actual uncertainty can be wider.", "sourceParagraphs": [ 510, 511 ] }, { "type": "heading3", "text": "10.4. The Approach of NOMOS", "sourceParagraph": 512 }, { "type": "paragraph", "text": "1,000 observations serve as:", "sourceParagraphs": [ 513 ] }, { "type": "bullet", "text": "the founding reference point of the protocol,", "sourceParagraph": 514 }, { "type": "bullet", "text": "default starting minimum at the model level,", "sourceParagraph": 515 }, { "type": "bullet", "text": "candidate threshold to be calibrated with the pilot", "sourceParagraph": 516 }, { "type": "paragraph", "text": "NOMOS does not claim that 1,000 observations are automatically sufficient for every country, language and condition of use.", "sourceParagraphs": [ 517, 518, 519 ] }, { "type": "heading2", "text": "11. INDEPENDENT OBSERVATION IS NOT THE SAME AS REPEATED OBSERVATION", "sourceParagraph": 521 }, { "type": "paragraph", "text": "A user can ask the same question ten times. This process can be valuable. However, it does not produce ten independent users.", "sourceParagraphs": [ 522, 523, 524 ] }, { "type": "heading3", "text": "11.1. Same User–Same Session Repeat", "sourceParagraph": 526 }, { "type": "paragraph", "text": "Can measure:", "sourceParagraphs": [ 527 ] }, { "type": "bullet", "text": "answer renewal variance,", "sourceParagraph": 528 }, { "type": "bullet", "text": "within-session stability,", "sourceParagraph": 529 }, { "type": "paragraph", "text": "probability of alternative answers. It does not measure population prevalence on its own.", "sourceParagraphs": [ 530, 531 ] }, { "type": "heading3", "text": "11.2. Same User–New Session Repeat", "sourceParagraph": 533 }, { "type": "paragraph", "text": "It may reduce the previous chat context. However:", "sourceParagraphs": [ 534, 535 ] }, { "type": "bullet", "text": "same account,", "sourceParagraph": 536 }, { "type": "bullet", "text": "same plan,", "sourceParagraph": 537 }, { "type": "bullet", "text": "same country,", "sourceParagraph": 538 }, { "type": "bullet", "text": "same personalisation,", "sourceParagraph": 539 }, { "type": "bullet", "text": "same user", "sourceParagraph": 540 }, { "type": "paragraph", "text": "may continue to create statistical dependence between observations.", "sourceParagraphs": [ 541 ] }, { "type": "heading3", "text": "11.3. A Single User Testing Multiple AI Products", "sourceParagraph": 543 }, { "type": "paragraph", "text": "This design can help with cross-model matched comparison. However, observations are dependent at the user level. This dependence should be preserved in recording and analysis.", "sourceParagraphs": [ 544, 545, 546 ] }, { "type": "heading3", "text": "11.4. Different Users Using the Same Account", "sourceParagraph": 548 }, { "type": "paragraph", "text": "Different people may exist. However:", "sourceParagraphs": [ 549, 550 ] }, { "type": "bullet", "text": "account history,", "sourceParagraph": 551 }, { "type": "bullet", "text": "personalisation,", "sourceParagraph": 552 }, { "type": "bullet", "text": "plan,", "sourceParagraph": 553 }, { "type": "bullet", "text": "memory", "sourceParagraph": 554 }, { "type": "paragraph", "text": "can be shared. These records cannot be considered independent automatically.", "sourceParagraphs": [ 555, 556 ] }, { "type": "heading3", "text": "11.5. Normative Independence Provision", "sourceParagraph": 558 }, { "type": "callout", "text": "Observations should be designed to be as independent as possible; known user, account, session, panel, country, device, and repeated dependencies should be recorded and taken into account in the analysis.", "sourceParagraph": 559 }, { "type": "paragraph", "text": "Independence should not be reduced to a binary:", "sourceParagraphs": [ 560 ] }, { "type": "bullet", "text": "yes,", "sourceParagraph": 561 }, { "type": "bullet", "text": "no", "sourceParagraph": 562 }, { "type": "paragraph", "text": "Where necessary, use one of the following dependency statuses:", "sourceParagraphs": [ 563 ] }, { "type": "heading3", "text": "INDEPENDENT", "sourceParagraph": 564 }, { "type": "heading3", "text": "PAIRED", "sourceParagraph": 565 }, { "type": "heading3", "text": "CLUSTERED", "sourceParagraph": 566 }, { "type": "heading3", "text": "REPEATED_WITHIN_USER", "sourceParagraph": 567 }, { "type": "heading3", "text": "DEPENDENT", "sourceParagraph": 568 }, { "type": "heading3", "text": "UNKNOWN", "sourceParagraph": 569 }, { "type": "heading2", "text": "12. FIRST SUITABLE OUTPUT PRINCIPLE", "sourceParagraph": 571 }, { "type": "paragraph", "text": "In the main population panel, the participant's task is not to find the most positive answer. The participant must record the first completed suitable output generated under the predefined condition. The participant:", "sourceParagraphs": [ 572, 573, 574 ] }, { "type": "bullet", "text": "cannot refresh the answer,", "sourceParagraph": 575 }, { "type": "bullet", "text": "cannot request a new answer,", "sourceParagraph": 576 }, { "type": "bullet", "text": "cannot choose the positive one,", "sourceParagraph": 577 }, { "type": "bullet", "text": "cannot only trim the good part,", "sourceParagraph": 578 }, { "type": "paragraph", "text": "cannot rewrite the answer. Reproduction behaviour can also be investigated separately. However: \"We asked the same question five times and selected the best answer.\" is not the main population score. This is the selected best output test. It should be reported with a separate name and method.", "sourceParagraphs": [ 579, 580, 581, 582, 583, 584, 585 ] }, { "type": "heading2", "text": "13. SINGLE CRITICAL EVENT EXCEPTION", "sourceParagraph": 587 }, { "type": "paragraph", "text": "A single answer cannot determine the population score. However, this does not mean that a single answer can never have a significant impact on a decision. NOMOS establishes two separate assessment pathways.", "sourceParagraphs": [ 588, 589, 590 ] }, { "type": "heading3", "text": "13.1. Population Representation Pathway", "sourceParagraph": 592 }, { "type": "paragraph", "text": "It asks the question: “How frequently does this outcome occur in the target population?” Requires a large and valid sample.", "sourceParagraphs": [ 593, 594, 595 ] }, { "type": "heading3", "text": "13.2. Critical Event Pathway", "sourceParagraph": 597 }, { "type": "paragraph", "text": "It asks the question: “How large is the foreseeable harm when this outcome occurs once?” A single confirmed outcome can trigger a review. For example:", "sourceParagraphs": [ 598, 599, 600, 601 ] }, { "type": "bullet", "text": "Assigning a non-existent health licence", "sourceParagraph": 602 }, { "type": "bullet", "text": "Making a dangerous emergency health referral", "sourceParagraph": 603 }, { "type": "bullet", "text": "Reporting incorrect legal authority", "sourceParagraph": 604 }, { "type": "bullet", "text": "Removing the security boundary", "sourceParagraph": 605 }, { "type": "bullet", "text": "Confusing a company with another legal entity", "sourceParagraph": 606 }, { "type": "bullet", "text": "Repeating a manipulative instruction hidden from people as if it were advice", "sourceParagraph": 607 }, { "type": "paragraph", "text": "It can be a singular Critical event candidate.", "sourceParagraphs": [ 608 ] }, { "type": "heading3", "text": "13.3. Separation of Score and Conformity Decision", "sourceParagraph": 610 }, { "type": "paragraph", "text": "A Critical event can generally be a single observation. For example:", "sourceParagraphs": [ 611, 612 ] }, { "type": "formula", "text": "1/1000=0.1%", "sourceParagraph": 613 }, { "type": "paragraph", "text": "This rate may seem small. However, if the potential harm of the event is severe, it may create an irreparable gap in the conformity decision. Therefore:", "sourceParagraphs": [ 614, 615, 616 ] }, { "type": "callout", "text": "Population score measures the frequency of the event; Critical event review measures the impact of the event on harm and integrity.", "sourceParagraph": 617 }, { "type": "paragraph", "text": "High average accuracy cannot automatically eliminate heavy singular error. The final conditions for critical classification will be defined in later chapters of the book.", "sourceParagraphs": [ 618, 619 ] }, { "type": "heading2", "text": "14. OBSERVATIONAL CLAIM LADDER", "sourceParagraph": 621 }, { "type": "paragraph", "text": "In this section, only the first four observation levels are normatively separated. At higher levels:", "sourceParagraphs": [ 622, 623 ] }, { "type": "bullet", "text": "population-weighted score,", "sourceParagraph": 624 }, { "type": "bullet", "text": "multi-model score,", "sourceParagraph": 625 }, { "type": "bullet", "text": "time stability,", "sourceParagraph": 626 }, { "type": "bullet", "text": "intervention effect,", "sourceParagraph": 627 }, { "type": "bullet", "text": "sustainable value", "sourceParagraph": 628 }, { "type": "paragraph", "text": "will be defined in subsequent sections.", "sourceParagraphs": [ 629 ] }, { "type": "heading3", "text": "O-0 — NOT TESTED", "sourceParagraph": 631 }, { "type": "paragraph", "text": "The relevant AI product or condition has not been measured. Correct status:", "sourceParagraphs": [ 632, 633 ] }, { "type": "heading3", "text": "NOT TESTED", "sourceParagraph": 634 }, { "type": "paragraph", "text": "Indefensible sentence: \"It is appropriate because the problem was not observed.\"", "sourceParagraphs": [ 635, 636 ] }, { "type": "heading3", "text": "O-1 — SINGLE EVENT OBSERVATION", "sourceParagraph": 638 }, { "type": "paragraph", "text": "There is one valid outcome. Defensible sentence: \"Under the specified conditions, this outcome was observed once.\" Indefensible sentence: \"The system represents existence this way.\"", "sourceParagraphs": [ 639, 640, 641, 642, 643 ] }, { "type": "heading3", "text": "O-2 — USER OR SESSION REPEAT", "sourceParagraph": 645 }, { "type": "paragraph", "text": "Multiple outcomes exist for the same user, account, or session family. Defensible sentence: \"Under the specified user and session conditions, this outcome occurred in seven out of ten repetitions.\" Indefensible sentence: \"Seven out of ten independent users saw this outcome.\"", "sourceParagraphs": [ 646, 647, 648, 649, 650 ] }, { "type": "heading3", "text": "O-3 — MULTI-USER SINGLE CELL MEASUREMENT", "sourceParagraph": 652 }, { "type": "paragraph", "text": "The same:", "sourceParagraphs": [ 653 ] }, { "type": "bullet", "text": "AI product,", "sourceParagraph": 654 }, { "type": "bullet", "text": "user surface,", "sourceParagraph": 655 }, { "type": "bullet", "text": "country,", "sourceParagraph": 656 }, { "type": "bullet", "text": "language,", "sourceParagraph": 657 }, { "type": "bullet", "text": "panel condition,", "sourceParagraph": 658 }, { "type": "bullet", "text": "prompt version,", "sourceParagraph": 659 }, { "type": "bullet", "text": "measurement wave", "sourceParagraph": 660 }, { "type": "paragraph", "text": "contains multiple suitable user observations. Defensible sentence: “78% of the 200 valid observations in the Turkish/Turkey controlled cell met the specified transition criterion.” This result cannot be automatically generalised to other countries, languages, or user surfaces.", "sourceParagraphs": [ 661, 662, 663, 664 ] }, { "type": "heading3", "text": "O-4 — MULTI-CELL LAYERED MEASUREMENT", "sourceParagraph": 666 }, { "type": "paragraph", "text": "Predefined samples exist from different countries, languages, or user cells. At this level:", "sourceParagraphs": [ 667, 668 ] }, { "type": "bullet", "text": "cell results,", "sourceParagraph": 669 }, { "type": "bullet", "text": "weighted result,", "sourceParagraph": 670 }, { "type": "bullet", "text": "scope,", "sourceParagraph": 671 }, { "type": "bullet", "text": "uncertainty", "sourceParagraph": 672 }, { "type": "paragraph", "text": "reportable. Detailed rules for population weights and subgroup fairness are defined in later chapters of the book.", "sourceParagraphs": [ 673, 674 ] }, { "type": "heading2", "text": "15. SYNTHETIC APPLE.COM DEMONSTRATION", "sourceParagraph": 676 }, { "type": "paragraph", "text": "SYNTHETIC METHODOLOGY DEMONSTRATION / The users, AI products, responses, rates, and results below are fictional. They do not represent the actual performance of Apple Inc. or the behaviour of any real AI product.", "sourceParagraphs": [ 677 ] }, { "type": "heading3", "text": "15.1. Scenario A — Single Positive Response", "sourceParagraph": 679 }, { "type": "paragraph", "text": "A synthetic user submits the following question to synthetic AI Product A: ‘What kind of company is Apple.com?’ Synthetic response: ‘Apple is a US-based technology company that develops consumer electronics, software and digital services.’ Preliminary evaluation:", "sourceParagraphs": [ 680, 681, 682, 683, 684 ] }, { "type": "bullet", "text": "Entity identity is correct", "sourceParagraph": 685 }, { "type": "bullet", "text": "Main category is correct", "sourceParagraph": 686 }, { "type": "bullet", "text": "No obvious critical errors", "sourceParagraph": 687 }, { "type": "bullet", "text": "Basic representation is acceptable", "sourceParagraph": 688 }, { "type": "paragraph", "text": "Raw result:", "sourceParagraphs": [ 689 ] }, { "type": "formula", "text": "1/1=100%", "sourceParagraph": 690 }, { "type": "paragraph", "text": "Incorrect public statement: “AI Product A represents Apple with 100% accuracy.” Correct public statement: “In the synthetic sample, the specified single observation met the transition criterion. Population prevalence and product reliability have not been determined.”", "sourceParagraphs": [ 691, 692, 693, 694 ] }, { "type": "heading3", "text": "15.2. Scenario B — Discovery with Ten Observations", "sourceParagraph": 696 }, { "type": "paragraph", "text": "Ten independent synthetic users use the same version of the semantic prompt. Synthetic result:", "sourceParagraphs": [ 697, 698 ] }, { "type": "bullet", "text": "7 responses pass", "sourceParagraph": 699 }, { "type": "bullet", "text": "1 response is partially correct", "sourceParagraph": 700 }, { "type": "bullet", "text": "1 response contains outdated information", "sourceParagraph": 701 }, { "type": "bullet", "text": "1 response assigns the wrong main activity", "sourceParagraph": 702 }, { "type": "paragraph", "text": "raw pass rate:", "sourceParagraphs": [ 703 ] }, { "type": "formula", "text": "7/10=70%", "sourceParagraph": 704 }, { "type": "paragraph", "text": "Under a simple, idealised assumption of independent sampling, the approximate 95 per cent Wilson interval is necessarily wide. The result remains exploratory because it is based on:", "sourceParagraphs": [ 705, 706, 707 ] }, { "type": "bullet", "text": "small sample,", "sourceParagraph": 708 }, { "type": "bullet", "text": "single cell,", "sourceParagraph": 709 }, { "type": "bullet", "text": "single wave", "sourceParagraph": 710 }, { "type": "paragraph", "text": "Correct status:", "sourceParagraphs": [ 711, 712 ] }, { "type": "callout", "text": "EXPLORATORY — Not sufficient for the publicly available global NOMOS score.", "sourceParagraph": 713 }, { "type": "heading3", "text": "15.3. Scenario C — 1,000 Observed Synthetic Wave", "sourceParagraph": 715 }, { "type": "paragraph", "text": "Assume synthetically that the following conditions are met:", "sourceParagraphs": [ 716 ] }, { "type": "bullet", "text": "AI Product A", "sourceParagraph": 717 }, { "type": "bullet", "text": "1,000 valid observations", "sourceParagraph": 718 }, { "type": "bullet", "text": "Predefined target population", "sourceParagraph": 719 }, { "type": "bullet", "text": "Locked prompt version", "sourceParagraph": 720 }, { "type": "bullet", "text": "Defined country and language layers", "sourceParagraph": 721 }, { "type": "bullet", "text": "Unmodified initial suitable response", "sourceParagraph": 722 }, { "type": "bullet", "text": "Full evidence records", "sourceParagraph": 723 }, { "type": "bullet", "text": "Predefined adjudication rules", "sourceParagraph": 724 }, { "type": "paragraph", "text": "Synthetic result:", "sourceParagraphs": [ 725 ] }, { "type": "table", "rows": [ [ "Response status", "Number" ], [ "Meeting the transition criterion", "918" ], [ "Partially correct", "37" ], [ "Unsupported material claim", "23" ], [ "Outdated information", "12" ], [ "Critical error candidate", "6" ], [ "Unclassifiable", "4" ], [ "Total", "1,000" ] ] }, { "type": "paragraph", "text": "Simple raw pass rate:", "sourceParagraphs": [ 726 ] }, { "type": "formula", "text": "918/1000=91.8%", "sourceParagraph": 727 }, { "type": "paragraph", "text": "Approximate 95% Wilson interval under the idealised unweighted assumption: is. However, this number alone is not yet the final NOMOS GEO score. Because there are additional factors that need to be evaluated:", "sourceParagraphs": [ 728, 729, 730, 731 ] }, { "type": "bullet", "text": "Country and language weights", "sourceParagraph": 732 }, { "type": "bullet", "text": "Design effect", "sourceParagraph": 733 }, { "type": "bullet", "text": "Lowest language result", "sourceParagraph": 734 }, { "type": "bullet", "text": "Invalid observation rate", "sourceParagraph": 735 }, { "type": "bullet", "text": "Inter-rater agreement", "sourceParagraph": 736 }, { "type": "bullet", "text": "Verification of critical error candidates", "sourceParagraph": 737 }, { "type": "bullet", "text": "Untested user groups", "sourceParagraph": 738 }, { "type": "bullet", "text": "Time and user surface limitation of the measurement", "sourceParagraph": 739 }, { "type": "paragraph", "text": "Six Critical-error candidates must also undergo separate incident review. Correct reporting: ‘The raw response pass rate in the synthetic wave is 91.8 per cent. This result describes the observations before population weights and design effects are applied. Because six Critical-error candidates were identified, the conformity decision requires separate review.’ Incorrect reporting: ‘Apple's NOMOS GEO score is 91.8 and it is compliant.’ Synthetic data cannot be attributed in this way to a real company or a real system.", "sourceParagraphs": [ 740, 741, 742, 743, 744, 745 ] }, { "type": "heading2", "text": "16. MANDATORY NORMATIVE PROVISIONS", "sourceParagraph": 747 }, { "type": "heading3", "text": "N-01", "sourceParagraph": 748 }, { "type": "paragraph", "text": "A single generative system output cannot be used as the general GEO score.", "sourceParagraphs": [ 749 ] }, { "type": "heading3", "text": "N-02", "sourceParagraph": 750 }, { "type": "paragraph", "text": "A singular output observed alone:", "sourceParagraphs": [ 751 ] }, { "type": "bullet", "text": "AI product and user surface,", "sourceParagraph": 752 }, { "type": "bullet", "text": "country,", "sourceParagraph": 753 }, { "type": "bullet", "text": "language,", "sourceParagraph": 754 }, { "type": "bullet", "text": "prompt,", "sourceParagraph": 755 }, { "type": "bullet", "text": "session,", "sourceParagraph": 756 }, { "type": "bullet", "text": "time", "sourceParagraph": 757 }, { "type": "paragraph", "text": "must be reported along with its conditions.", "sourceParagraphs": [ 758 ] }, { "type": "heading3", "text": "N-03", "sourceParagraph": 759 }, { "type": "paragraph", "text": "Single output:", "sourceParagraphs": [ 760 ] }, { "type": "bullet", "text": "typical,", "sourceParagraph": 761 }, { "type": "bullet", "text": "continuous,", "sourceParagraph": 762 }, { "type": "bullet", "text": "global,", "sourceParagraph": 763 }, { "type": "bullet", "text": "belonging to all users", "sourceParagraph": 764 }, { "type": "paragraph", "text": "cannot be presented as behaviour.", "sourceParagraphs": [ 765 ] }, { "type": "heading3", "text": "N-04", "sourceParagraph": 766 }, { "type": "paragraph", "text": "Repetitions from the same user, account, or session cannot be counted as independent population observations without disclosing the dependency.", "sourceParagraphs": [ 767 ] }, { "type": "heading3", "text": "N-05", "sourceParagraph": 768 }, { "type": "paragraph", "text": "In the main population panel, the participant must record the first suitable output predefined; reproduction cannot be made to choose a positive response.", "sourceParagraphs": [ 769 ] }, { "type": "heading3", "text": "N-06", "sourceParagraph": 770 }, { "type": "paragraph", "text": "System rejection, error message, or inability to respond should be recorded as a valid system outcome if the protocol has been correctly applied.", "sourceParagraphs": [ 771 ] }, { "type": "heading3", "text": "N-07", "sourceParagraph": 772 }, { "type": "paragraph", "text": "Trimmed, modified, manually rewritten, or responses whose prompt context cannot be verified cannot be included in the main score calculation.", "sourceParagraphs": [ 773 ] }, { "type": "heading3", "text": "N-08", "sourceParagraph": 774 }, { "type": "paragraph", "text": "Selecting only positive screenshots and withholding the distribution of results is prohibited.", "sourceParagraphs": [ 775 ] }, { "type": "heading3", "text": "N-09", "sourceParagraph": 776 }, { "type": "paragraph", "text": "Synthetic observations cannot be used in real user or real system scores.", "sourceParagraphs": [ 777 ] }, { "type": "heading3", "text": "N-10", "sourceParagraph": 778 }, { "type": "paragraph", "text": "A population proportion should not be published without coverage, denominator, and appropriate uncertainty explanation.", "sourceParagraphs": [ 779 ] }, { "type": "heading3", "text": "N-11", "sourceParagraph": 780 }, { "type": "paragraph", "text": "The probability of a critical error cannot be quietly compensated within the overall average.", "sourceParagraphs": [ 781 ] }, { "type": "heading3", "text": "N-12", "sourceParagraph": 782 }, { "type": "paragraph", "text": "The observation result can only be applied to the inference universe that is uniquely defined; it cannot be generalised to unmeasured countries, languages, products, or time periods.", "sourceParagraphs": [ 783 ] }, { "type": "heading3", "text": "N-13", "sourceParagraph": 784 }, { "type": "paragraph", "text": "Measurement, scoring, or public statement must have an accountable human or institutional owner.", "sourceParagraphs": [ 785 ] }, { "type": "heading3", "text": "N-14", "sourceParagraph": 786 }, { "type": "paragraph", "text": "UNKNOWN, NOT TESTED, or INVALID records cannot be quietly converted into a positive result.", "sourceParagraphs": [ 787 ] }, { "type": "heading2", "text": "17. CANDIDATE GEO-1000 MAIN RULE", "sourceParagraph": 789 }, { "type": "heading3", "text": "CANDIDATE-001", "sourceParagraph": 790 }, { "type": "callout", "text": "For every AI product being audited and for every main population measurement wave, a minimum of 1,000 valid user observations should be targeted by default.", "sourceParagraph": 791 }, { "type": "paragraph", "text": "This candidate sentence:", "sourceParagraphs": [ 792 ] }, { "type": "bullet", "text": "accuracy of the general product forecast,", "sourceParagraph": 793 }, { "type": "bullet", "text": "cost,", "sourceParagraph": 794 }, { "type": "bullet", "text": "applicability,", "sourceParagraph": 795 }, { "type": "bullet", "text": "country and language fairness", "sourceParagraph": 796 }, { "type": "paragraph", "text": "This creates an initial balance among these considerations. It does not, however, guarantee adequate coverage of:", "sourceParagraphs": [ 797, 798 ] }, { "type": "bullet", "text": "all the country scores,", "sourceParagraph": 799 }, { "type": "bullet", "text": "all language scores,", "sourceParagraph": 800 }, { "type": "bullet", "text": "all user subgroups", "sourceParagraph": 801 }, { "type": "paragraph", "text": "Adequacy must instead be tested through:", "sourceParagraphs": [ 802, 803 ] }, { "type": "bullet", "text": "pilots,", "sourceParagraph": 804 }, { "type": "bullet", "text": "independent repetitions,", "sourceParagraph": 805 }, { "type": "bullet", "text": "design effect,", "sourceParagraph": 806 }, { "type": "bullet", "text": "subgroup sensitivity", "sourceParagraph": 807 }, { "type": "paragraph", "text": "Those evaluations determine the final normative status.", "sourceParagraphs": [ 808 ] }, { "type": "heading2", "text": "18. FORMS OF FAILURE", "sourceParagraph": 810 }, { "type": "heading4", "text": "F-01 — Selected Screenshot", "sourceParagraph": 811 }, { "type": "paragraph", "text": "Only the most positive response is published among numerous results.", "sourceParagraphs": [ 812 ] }, { "type": "heading4", "text": "F-02 — Treating a Single User as the Population", "sourceParagraph": 813 }, { "type": "paragraph", "text": "The experience of one person is generalised to the entire target universe.", "sourceParagraphs": [ 814 ] }, { "type": "heading4", "text": "F-03 — Counting Repeats of the Same User as New People", "sourceParagraph": 815 }, { "type": "paragraph", "text": "Repeats generated by the same person or account are reported as independent participants.", "sourceParagraphs": [ 816 ] }, { "type": "heading4", "text": "F-04 — Renewing the Response and Picking the Best One", "sourceParagraph": 817 }, { "type": "paragraph", "text": "Initial or negative responses are discarded; the most appropriate response is saved as the measurement result.", "sourceParagraphs": [ 818 ] }, { "type": "heading4", "text": "F-05 — Prompt Drift", "sourceParagraph": 819 }, { "type": "paragraph", "text": "Participants change the word, tone, or scope of the prompt; results are combined as if from the same experiment.", "sourceParagraphs": [ 820 ] }, { "type": "heading4", "text": "F-06 — Context Contamination", "sourceParagraph": 821 }, { "type": "paragraph", "text": "Previous conversations affect the response; the record is presented as a clean session.", "sourceParagraphs": [ 822 ] }, { "type": "heading4", "text": "F-07 — Hiding Personalisation", "sourceParagraph": 823 }, { "type": "paragraph", "text": "Memory, special instructions, or user history is visible; the result is reported as a general user response.", "sourceParagraphs": [ 824 ] }, { "type": "heading4", "text": "F-08 — Confusing Model and Product", "sourceParagraph": 825 }, { "type": "paragraph", "text": "Only the general model or provider name is written; the actual product, interface, and tool condition are not recorded.", "sourceParagraphs": [ 826 ] }, { "type": "heading4", "text": "F-09 — Removing Errors and Mistakes from the Denominator", "sourceParagraph": 827 }, { "type": "paragraph", "text": "Deleted to make the results appear positive when the system does not respond to records.", "sourceParagraphs": [ 828 ] }, { "type": "heading4", "text": "F-10 — Counting Invalid Capture as System Failure", "sourceParagraph": 829 }, { "type": "paragraph", "text": "Reported as participant error, missing evidence, or incorrect outcome of the AI product.", "sourceParagraphs": [ 830 ] }, { "type": "heading4", "text": "F-11 — Globalising a Single Language and Country", "sourceParagraph": 831 }, { "type": "paragraph", "text": "The result of a single locale is applied globally.", "sourceParagraphs": [ 832 ] }, { "type": "heading4", "text": "F-12 — Using Synthetic Data as Real", "sourceParagraph": 833 }, { "type": "paragraph", "text": "Records created for demonstration purposes are added to actual performance.", "sourceParagraphs": [ 834 ] }, { "type": "heading4", "text": "F-13 — Losing a Critical Event on Average", "sourceParagraph": 835 }, { "type": "paragraph", "text": "Severe health, legal, licensing, or security errors are made invisible within a high average.", "sourceParagraphs": [ 836 ] }, { "type": "heading4", "text": "F-14 — Drawing Intervention Effect from a Single Success", "sourceParagraph": 837 }, { "type": "paragraph", "text": "The single positive response received after the GEO intervention is considered the causal effect of the study.", "sourceParagraphs": [ 838 ] }, { "type": "heading4", "text": "F-15 — Judging the Entire System from a Single Mistake", "sourceParagraph": 839 }, { "type": "paragraph", "text": "One error is generalised as if the AI product is always wrong and wrong for all users.", "sourceParagraphs": [ 840 ] }, { "type": "heading4", "text": "F-16 — Defining the Target Population Afterwards", "sourceParagraph": 841 }, { "type": "paragraph", "text": "After the outcome is seen, the user group with high success is declared the “actual target population.”", "sourceParagraphs": [ 842 ] }, { "type": "heading2", "text": "19. AUDIT PROCEDURE", "sourceParagraph": 844 }, { "type": "paragraph", "text": "An auditor should follow the sequence below when evaluating the validity of a single observation or an early GEO claim.", "sourceParagraphs": [ 845 ] }, { "type": "heading3", "text": "Step 1 — Record the Public Claim", "sourceParagraph": 846 }, { "type": "paragraph", "text": "The full proposed sentence is determined. Example: “ChatGPT correctly knows our company.”", "sourceParagraphs": [ 847, 848, 849 ] }, { "type": "heading3", "text": "Step 2 — Determine the Level of the Claim", "sourceParagraph": 850 }, { "type": "paragraph", "text": "The sentence may make a claim about:", "sourceParagraphs": [ 851 ] }, { "type": "bullet", "text": "single event,", "sourceParagraph": 852 }, { "type": "bullet", "text": "prevalence,", "sourceParagraph": 853 }, { "type": "bullet", "text": "stability,", "sourceParagraph": 854 }, { "type": "bullet", "text": "global consequence,", "sourceParagraph": 855 }, { "type": "bullet", "text": "causality,", "sourceParagraph": 856 }, { "type": "bullet", "text": "commercial impact", "sourceParagraph": 857 }, { "type": "paragraph", "text": "The applicable level must be identified.", "sourceParagraphs": [ 858 ] }, { "type": "heading3", "text": "Step 3 — Determine the Supporting Observation Count", "sourceParagraph": 859 }, { "type": "paragraph", "text": "How many real users? How many accounts? How many independent sessions? How many repeats? How many AI products? How many languages and countries?", "sourceParagraphs": [ 860, 861, 862, 863, 864, 865 ] }, { "type": "heading3", "text": "Step 4 — Verify Observation Conditions", "sourceParagraph": 866 }, { "type": "paragraph", "text": "Are the AI product, user surface, prompt, session, country, language and time all recorded?", "sourceParagraphs": [ 867, 868, 869, 870, 871, 872, 873, 874 ] }, { "type": "heading3", "text": "Step 5 — Examine Evidence Integrity", "sourceParagraph": 875 }, { "type": "paragraph", "text": "Is there a full response? Is there a full screenshot? Has the response been selected or refreshed? Has the file been modified? Is the record synthetic?", "sourceParagraphs": [ 876, 877, 878, 879, 880 ] }, { "type": "heading3", "text": "Step 6 — Determine Dependency Structure", "sourceParagraph": 881 }, { "type": "paragraph", "text": "Are there repetitions of the same user? Has a shared account been used? Is there a mirrored model design? Is there clustering from the same panel?", "sourceParagraphs": [ 882, 883, 884, 885 ] }, { "type": "heading3", "text": "Step 7 — Separate Data Collection Error from Valid System Result", "sourceParagraph": 886 }, { "type": "paragraph", "text": "Is the rejection or system error the actual result? Or did the participant violate the protocol?", "sourceParagraphs": [ 887, 888 ] }, { "type": "heading3", "text": "Step 8 — Compare the Claim with the Level of Evidence", "sourceParagraph": 889 }, { "type": "paragraph", "text": "Is a single observation used for a population or global inference?", "sourceParagraphs": [ 890 ] }, { "type": "heading3", "text": "Step 9 — Open the Critical Path", "sourceParagraph": 891 }, { "type": "paragraph", "text": "If a single response carries a foreseeable severe harm, a separate incident investigation is initiated.", "sourceParagraphs": [ 892 ] }, { "type": "heading3", "text": "Step 10 — Limit the Decision", "sourceParagraph": 893 }, { "type": "paragraph", "text": "The public sentence is reduced only to the level carried by the evidence.", "sourceParagraphs": [ 894 ] }, { "type": "heading2", "text": "20. NECESSARY EVIDENCE", "sourceParagraph": 896 }, { "type": "paragraph", "text": "For an observation or claim within the scope of this section, the following records should be sought to the extent that they are relevant:", "sourceParagraphs": [ 897 ] }, { "type": "bullet", "text": "Audited entity identity", "sourceParagraph": 898 }, { "type": "bullet", "text": "Fully disclosed GEO claim to the public", "sourceParagraph": 899 }, { "type": "bullet", "text": "Number of observations", "sourceParagraph": 900 }, { "type": "bullet", "text": "Number of unique participants", "sourceParagraph": 901 }, { "type": "bullet", "text": "AI product and user surface", "sourceParagraph": 902 }, { "type": "bullet", "text": "Visible model or version, if available", "sourceParagraph": 903 }, { "type": "bullet", "text": "Plan or account type", "sourceParagraph": 904 }, { "type": "bullet", "text": "Country and locale", "sourceParagraph": 905 }, { "type": "bullet", "text": "Prompt ID, version, and full text", "sourceParagraph": 906 }, { "type": "bullet", "text": "Session and personalisation status", "sourceParagraph": 907 }, { "type": "bullet", "text": "UTC timestamp", "sourceParagraph": 908 }, { "type": "bullet", "text": "Measurement wave", "sourceParagraph": 909 }, { "type": "bullet", "text": "Full raw response", "sourceParagraph": 910 }, { "type": "bullet", "text": "Screenshot or equivalent evidence", "sourceParagraph": 911 }, { "type": "bullet", "text": "Reproduction or answer selection status", "sourceParagraph": 912 }, { "type": "bullet", "text": "Observation validity status", "sourceParagraph": 913 }, { "type": "bullet", "text": "Dependency and matching record", "sourceParagraph": 914 }, { "type": "bullet", "text": "Synthetic or real status", "sourceParagraph": 915 }, { "type": "bullet", "text": "Invalid observation log", "sourceParagraph": 916 }, { "type": "bullet", "text": "Critical event record", "sourceParagraph": 917 }, { "type": "bullet", "text": "Result scope", "sourceParagraph": 918 }, { "type": "bullet", "text": "Uncertainty statement", "sourceParagraph": 919 }, { "type": "bullet", "text": "Responsible person or institution", "sourceParagraph": 920 }, { "type": "paragraph", "text": "Not all of these records may be required in the same detail for every low-risk discovery observation. However, if a broad GEO score or conformity clause is to be given to the public, material fields cannot be left blank.", "sourceParagraphs": [ 921, 922 ] }, { "type": "heading2", "text": "21. AUDIT CHECKLIST", "sourceParagraph": 924 }, { "type": "paragraph", "text": "Has the audited entity been fully identified? Have the AI product and user surface been recorded? Was the prompt locked before measurement? Was the prompt changed by the participant? Are the session and personalisation conditions known? Is the user's country and locale information recorded? Is the full response preserved? Is it the first appropriate output or the best chosen output? Is there a complete record of evidence? Have the actual number of users and the number of observations been separated? Have repeated and matched observations been counted independently? Have refusals and system errors been properly managed in the denominator? Have invalid records been retained and justified? Have synthetic and real data been separated? Is a single response presented as a population result?", "sourceParagraphs": [ 925, 926, 927, 928, 929, 930, 931, 932, 933, 934, 935, 936, 937, 938, 939 ] }, { "type": "paragraph", "text": "Is the claim applied only to the measured product, country, language, and time? Has the probability of a critical error been assessed separately? Does the result carry a statement of confidence or uncertainty? Are untested areas clearly shown? Does the public statement exceed the level of actual observation? If one of the factual questions is unanswered, the result carries limited confidence. If more than one is unanswered, it should not be used for the overall GEO score.", "sourceParagraphs": [ 940, 941, 942, 943, 944, 945, 946 ] }, { "type": "heading2", "text": "22. OBJECTIONS AND RESPONSES", "sourceParagraph": 948 }, { "type": "heading3", "text": "Objection 1 — “The real user already sees only one answer.”", "sourceParagraph": 949 }, { "type": "paragraph", "text": "That is correct. The single answer seen by the single user is that person’s real experience. Therefore, individual events are recorded. However, individual experience is not the same as population experience. NOMOS maintains both facts together:", "sourceParagraphs": [ 950, 951, 952, 953, 954 ] }, { "type": "callout", "text": "The experience of a single user is real. / A single user is not the entire population.", "sourceParagraph": 955 }, { "type": "heading3", "text": "Objection 2 — “If the system gives the same answer every time, isn't one answer enough?”", "sourceParagraph": 957 }, { "type": "paragraph", "text": "The claim that the system gives the same answer every time also requires measurement. Additionally:", "sourceParagraphs": [ 958, 959 ] }, { "type": "bullet", "text": "time,", "sourceParagraph": 960 }, { "type": "bullet", "text": "retrieval,", "sourceParagraph": 961 }, { "type": "bullet", "text": "country,", "sourceParagraph": 962 }, { "type": "bullet", "text": "language,", "sourceParagraph": 963 }, { "type": "bullet", "text": "plan,", "sourceParagraph": 964 }, { "type": "bullet", "text": "personalisation,", "sourceParagraph": 965 }, { "type": "bullet", "text": "user surface", "sourceParagraph": 966 }, { "type": "paragraph", "text": "It may vary. Even a system assumed to be deterministic cannot be generalised beyond defined conditions.", "sourceParagraphs": [ 967, 968 ] }, { "type": "heading3", "text": "Objection 3 — \"Why exactly 1,000?\"", "sourceParagraph": 970 }, { "type": "paragraph", "text": "The number 1,000 is not a magic threshold of certainty. GEO-1000 serves as a founding reference point that balances:", "sourceParagraphs": [ 971, 972 ] }, { "type": "bullet", "text": "a substantial user volume,", "sourceParagraph": 973 }, { "type": "bullet", "text": "a broad product-level estimate,", "sourceParagraph": 974 }, { "type": "bullet", "text": "practicable cost,", "sourceParagraph": 975 }, { "type": "bullet", "text": "public comprehensibility.", "sourceParagraph": 976 }, { "type": "paragraph", "text": "Subgroups may require more observations. Pilot studies must verify or revise the threshold.", "sourceParagraphs": [ 977, 978, 979 ] }, { "type": "heading3", "text": "Objection 4 — “Let the same 1,000 people test all AI products.”", "sourceParagraph": 981 }, { "type": "paragraph", "text": "This method can be useful for paired comparison across models. However, selecting only intensive AI users who can access all products may bias the target population. Also, observations from the same person are dependent. Fully paired, partially paired, and balanced designs will be compared later.", "sourceParagraphs": [ 982, 983, 984, 985 ] }, { "type": "heading3", "text": "Objection 5 — “Isn’t a screenshot already sufficient proof?”", "sourceParagraph": 987 }, { "type": "paragraph", "text": "Screenshots are important. However, on their own they do not always show the following:", "sourceParagraphs": [ 988, 989 ] }, { "type": "bullet", "text": "real time", "sourceParagraph": 990 }, { "type": "bullet", "text": "country", "sourceParagraph": 991 }, { "type": "bullet", "text": "plan", "sourceParagraph": 992 }, { "type": "bullet", "text": "previous conversation", "sourceParagraph": 993 }, { "type": "bullet", "text": "personalisation", "sourceParagraph": 994 }, { "type": "bullet", "text": "answer refresh", "sourceParagraph": 995 }, { "type": "bullet", "text": "file change", "sourceParagraph": 996 }, { "type": "bullet", "text": "full response", "sourceParagraph": 997 }, { "type": "paragraph", "text": "For this reason, a screenshot is an important part of the evidence package. It is not the whole.", "sourceParagraphs": [ 998, 999 ] }, { "type": "heading3", "text": "Objection 6 — “Why do we include a non-responding system in the score?”", "sourceParagraph": 1001 }, { "type": "paragraph", "text": "The user's real experience may be not getting an answer. A rejection or error:", "sourceParagraphs": [ 1002, 1003 ] }, { "type": "bullet", "text": "technical,", "sourceParagraph": 1004 }, { "type": "bullet", "text": "security,", "sourceParagraph": 1005 }, { "type": "bullet", "text": "policy,", "sourceParagraph": 1006 }, { "type": "bullet", "text": "access", "sourceParagraph": 1007 }, { "type": "paragraph", "text": "may be the result. If the protocol was correctly applied, the event should be recorded. The system’s unresponsiveness due to the participant’s technical error is classified separately.", "sourceParagraphs": [ 1008, 1009, 1010 ] }, { "type": "heading3", "text": "Objection 7 — “Why should a single severe mistake affect the entire inspection?”", "sourceParagraph": 1012 }, { "type": "paragraph", "text": "A single mistake does not prove that most of the population made the same mistake. However, the potential damage of a single mistake can be severe. Therefore:", "sourceParagraphs": [ 1013, 1014, 1015 ] }, { "type": "bullet", "text": "score frequency,", "sourceParagraph": 1016 }, { "type": "bullet", "text": "Critical decision harm and integrity", "sourceParagraph": 1017 }, { "type": "paragraph", "text": "are evaluated separately. The average cannot automatically erase human harm.", "sourceParagraphs": [ 1018, 1019 ] }, { "type": "heading3", "text": "Objection 8 — “If we can’t say anything with a single response, what is the value of early tests?”", "sourceParagraph": 1021 }, { "type": "paragraph", "text": "Singular and small-sample tests:", "sourceParagraphs": [ 1022 ] }, { "type": "bullet", "text": "generate hypotheses,", "sourceParagraph": 1023 }, { "type": "bullet", "text": "discover types of errors,", "sourceParagraph": 1024 }, { "type": "bullet", "text": "test prompts,", "sourceParagraph": 1025 }, { "type": "bullet", "text": "test the capture system,", "sourceParagraph": 1026 }, { "type": "bullet", "text": "calibrate adjudicators,", "sourceParagraph": 1027 }, { "type": "paragraph", "text": "can reveal critical events. What is wrong is not to do exploratory testing. What is wrong is presenting the exploratory result as a population score.", "sourceParagraphs": [ 1028, 1029, 1030 ] }, { "type": "heading2", "text": "COMMON PRINCIPLE OF CHAPTER 24", "sourceParagraph": 1061 }, { "type": "paragraph", "text": "This chapter was written to eliminate a single misconception: \"I asked an AI, got an answer; here is my GEO result.\" No. You do not yet have a general GEO result. What you have is an observation. This observation:", "sourceParagraphs": [ 1062, 1063, 1064, 1065, 1066, 1067 ] }, { "type": "bullet", "text": "is real,", "sourceParagraph": 1068 }, { "type": "bullet", "text": "is significant,", "sourceParagraph": 1069 }, { "type": "bullet", "text": "even critical", "sourceParagraph": 1070 }, { "type": "paragraph", "text": "may be. However, it is still a single observation. For a GEO score, the existence of:", "sourceParagraphs": [ 1071, 1072, 1073 ] }, { "type": "bullet", "text": "users,", "sourceParagraph": 1074 }, { "type": "bullet", "text": "AI products,", "sourceParagraph": 1075 }, { "type": "bullet", "text": "countries,", "sourceParagraph": 1076 }, { "type": "bullet", "text": "languages,", "sourceParagraph": 1077 }, { "type": "bullet", "text": "sessions,", "sourceParagraph": 1078 }, { "type": "bullet", "text": "times", "sourceParagraph": 1079 }, { "type": "paragraph", "text": "What must be measured is the distribution of representation across AI products, countries, languages, sessions and time. A single screenshot is not a distribution. Ten automated repetitions are not ten independent people. Nor do a thousand responses from a poor sample represent the world's population. Even a population-proportional sample may be incomplete without language fairness. A high average cannot erase Critical errors. NOMOS's first law of measurement is therefore:", "sourceParagraphs": [ 1080, 1081, 1082, 1083, 1084, 1085, 1086 ] }, { "type": "callout", "text": "A single answer is an event. / Multiple answers are a pattern. / A defined sample is an estimate of the population. / Repetition over time is a measure of stability. / But none of these alone are causality or sustainable value.", "sourceParagraph": 1087 }, { "type": "heading2", "text": "Order of Section 1 of NOMOS", "sourceParagraph": 1089 }, { "type": "callout", "text": "Do not bring me one screenshot and then speak for the entire system.", "sourceParagraph": 1090 }, { "type": "callout", "text": "Do not make the answer of one user the answer of the world's population.", "sourceParagraph": 1091 }, { "type": "callout", "text": "Do not convert duplicates of the same account into independent humans.", "sourceParagraph": 1092 }, { "type": "callout", "text": "Do not refresh the answer and display the preferred as if it were the real distribution.", "sourceParagraph": 1093 }, { "type": "callout", "text": "Do not remove rejections, errors, and unknowns from the denominator.", "sourceParagraph": 1094 }, { "type": "callout", "text": "Do not generalise one language to the world, one country to all countries, or one AI product to all generative systems.", "sourceParagraph": 1095 }, { "type": "callout", "text": "Do not turn a single success into evidence of an intervention effect, or a single error into the immutable character of the entire system.", "sourceParagraph": 1096 }, { "type": "callout", "text": "Do not drown a single severe error within the average.", "sourceParagraph": 1097 }, { "type": "callout", "text": "First, record the event. / Then measure the pattern. / Then define the target population. / Then show dependency and uncertainty. / And speak only to the extent that the evidence carries.", "sourceParagraph": 1098 }, { "type": "heading2", "text": "The Chapter's Closing Sentence", "sourceParagraph": 1100 }, { "type": "callout", "text": "The representation of an entity in artificial intelligence systems is not a single answer. AI products are a distribution that should be measured across users, countries, languages, sessions, and times.", "sourceParagraph": 1101 }, { "type": "heading2", "text": "Normative Core" }, { "type": "normative", "text": "A single generative-system output MAY establish that a specific response occurred under recorded conditions. A single output MUST NOT be used to infer: - population prevalence, - typical product behaviour, - temporal stability, - cross-language consistency, - cross-country consistency, - cross-product consistency, - intervention effect, - recommendation certainty, - commercial impact, - conformity, - or a GEO score. All broader claims MUST be supported by a predefined target population, valid multi-observation design, recorded dependence structure, evidence bundle, evaluation rule, scope statement, and uncertainty disclosure. A single high-harm output MAY trigger a separate critical-incident review without being treated as evidence of population prevalence.", "translationRoute": "controlled-normative-rule" } ] }, { "number": 2, "id": "NOMOS-GEO-AUDIT-CH02", "title": "Representation Is a Distribution", "subtitle": null, "sourceFile": "2.ci bölüm.docx", "sourceSha256": "CF157E98388D9F58E9BEF9F44ADF14CD5B513AC59035C47BFA391660D30BE6D5", "sourceWordCount": 6390, "sourceIds": [ "K04", "K05", "K09" ], "blocks": [ { "type": "heading2", "text": "Chapter Boundary", "sourceParagraph": 16 }, { "type": "paragraph", "text": "Section 1 established the following provision:", "sourceParagraphs": [ 17 ] }, { "type": "callout", "text": "A single response is not the GEO score. A single response is a singular observation that occurred under recorded conditions.", "sourceParagraph": 18 }, { "type": "paragraph", "text": "This section answers the next question:", "sourceParagraphs": [ 19 ] }, { "type": "callout", "text": "If a single response is not the GEO score, what is the actual object we want to measure?", "sourceParagraph": 20 }, { "type": "paragraph", "text": "The answer is:", "sourceParagraphs": [ 21 ] }, { "type": "callout", "text": "The representation of an entity in generative systems is a multidimensional distribution formed across AI products, users, countries, languages, session conditions, prompts, and times.", "sourceParagraph": 22 }, { "type": "paragraph", "text": "This chapter:", "sourceParagraphs": [ 23 ] }, { "type": "bullet", "text": "the representation distribution,", "sourceParagraph": 24 }, { "type": "bullet", "text": "the fundamental dimensions of the distribution,", "sourceParagraph": 25 }, { "type": "bullet", "text": "the difference between the mean and the distribution,", "sourceParagraph": 26 }, { "type": "bullet", "text": "the subgroup inequality,", "sourceParagraph": 27 }, { "type": "bullet", "text": "the stability,", "sourceParagraph": 28 }, { "type": "bullet", "text": "the coverage and missing observation problem,", "sourceParagraph": 29 }, { "type": "bullet", "text": "the risk of rare severe errors,", "sourceParagraph": 30 }, { "type": "bullet", "text": "the conceptual structure of the publicly available result card", "sourceParagraph": 31 }, { "type": "paragraph", "text": "defines. This chapter does not yet:", "sourceParagraphs": [ 32, 33 ] }, { "type": "bullet", "text": "how many participants will be assigned to each country,", "sourceParagraph": 34 }, { "type": "bullet", "text": "how the sample frame will be established,", "sourceParagraph": 35 }, { "type": "bullet", "text": "how prompt translations will be validated,", "sourceParagraph": 36 }, { "type": "bullet", "text": "how atomic claims will be scored,", "sourceParagraph": 37 }, { "type": "bullet", "text": "the exact formula of the final NOMOS score,", "sourceParagraph": 38 }, { "type": "bullet", "text": "conformity thresholds", "sourceParagraph": 39 }, { "type": "paragraph", "text": "This chapter does not settle these matters normatively; later chapters address them separately. Before defining the scoring formula, Chapter 2 defines the reality that the score is intended to represent.", "sourceParagraphs": [ 40, 41, 42 ] }, { "type": "heading2", "text": "NOMOS Challenge", "sourceParagraph": 44 }, { "type": "paragraph", "text": "Consider two AI products. Each receives the same global average score:", "sourceParagraphs": [ 45, 46 ] }, { "type": "heading4", "text": "90 / 100", "sourceParagraph": 47 }, { "type": "paragraph", "text": "The first product's language outcomes are as follows:", "sourceParagraphs": [ 48 ] }, { "type": "bullet", "text": "English: 90", "sourceParagraph": 49 }, { "type": "bullet", "text": "Turkish: 91", "sourceParagraph": 50 }, { "type": "bullet", "text": "German: 89", "sourceParagraph": 51 }, { "type": "bullet", "text": "Japanese: 90", "sourceParagraph": 52 }, { "type": "bullet", "text": "Arabic: 90", "sourceParagraph": 53 }, { "type": "paragraph", "text": "The language results of the second product are as follows:", "sourceParagraphs": [ 54 ] }, { "type": "bullet", "text": "English: 99", "sourceParagraph": 55 }, { "type": "bullet", "text": "Turkish: 98", "sourceParagraph": 56 }, { "type": "bullet", "text": "German: 97", "sourceParagraph": 57 }, { "type": "bullet", "text": "Japanese: 86", "sourceParagraph": 58 }, { "type": "bullet", "text": "Arabic: 70", "sourceParagraph": 59 }, { "type": "paragraph", "text": "The simple average of both products is 90. Do they have the same GEO quality? No. The first product produces a similar representation in different languages. The second product produces excellent representation in some languages but seriously weak representation in another language. A single average hides this difference. Now consider two other products. 95% of the responses of both meet the passing criteria. The remaining 5% in the first product consists of:", "sourceParagraphs": [ 60, 61, 62, 63, 64, 65, 66, 67, 68 ] }, { "type": "bullet", "text": "small scope deficiencies,", "sourceParagraph": 69 }, { "type": "bullet", "text": "limited explanation shortcomings,", "sourceParagraph": 70 }, { "type": "bullet", "text": "a few UNKNOWN", "sourceParagraph": 71 }, { "type": "paragraph", "text": "results. The remaining 5% in the second product consists of:", "sourceParagraphs": [ 72, 73 ] }, { "type": "bullet", "text": "incorrect health authority,", "sourceParagraph": 74 }, { "type": "bullet", "text": "nonexistent licence,", "sourceParagraph": 75 }, { "type": "bullet", "text": "incorrect legal identity,", "sourceParagraph": 76 }, { "type": "bullet", "text": "inappropriate advice", "sourceParagraph": 77 }, { "type": "paragraph", "text": "exists. The raw accuracy rate for both is again 95 per cent. Do they have the same representation risk? No. Now consider a third pair. The average of both across the three measurement waves is 90. The first one:", "sourceParagraphs": [ 78, 79, 80, 81, 82, 83, 84 ] }, { "type": "bullet", "text": "Wave 1: 90", "sourceParagraph": 85 }, { "type": "bullet", "text": "Wave 2: 91", "sourceParagraph": 86 }, { "type": "bullet", "text": "Wave 3: 89", "sourceParagraph": 87 }, { "type": "paragraph", "text": "The second one:", "sourceParagraphs": [ 88 ] }, { "type": "bullet", "text": "Wave 1: 98", "sourceParagraph": 89 }, { "type": "bullet", "text": "Wave 2: 90", "sourceParagraph": 90 }, { "type": "bullet", "text": "Wave 3: 82", "sourceParagraph": 91 }, { "type": "paragraph", "text": "Do they have the same consistency over time? No. In that case, the issue is not only: 'What is the average?' We also need to know the following: How are the results distributed? Who is well represented, who is poorly represented? Which languages and countries are falling behind? How severe are the errors? How much does the result change over time? Which areas were never measured? With what weights was the average created? How many different realities lie behind the same average? The first conclusion of this section is:", "sourceParagraphs": [ 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106 ] }, { "type": "callout", "text": "The same average is not the same representation.", "sourceParagraph": 107 }, { "type": "paragraph", "text": "Its second provision states:", "sourceParagraphs": [ 108 ] }, { "type": "callout", "text": "a GEO score may be a summary of the distribution; it cannot replace the distribution.", "sourceParagraph": 109 }, { "type": "heading2", "text": "1. PURPOSE OF THE CHAPTER", "sourceParagraph": 111 }, { "type": "paragraph", "text": "This section defines how the representation of entities in generative systems should be understood before being reduced to a single number. Its main objectives are:", "sourceParagraphs": [ 112, 113 ] }, { "type": "bullet", "text": "To show that representation is not a timeless and singular “AI perspective\"", "sourceParagraph": 114 }, { "type": "bullet", "text": "To define the conditions on which the representation outcome depends", "sourceParagraph": 115 }, { "type": "bullet", "text": "To explain that the true–false binary alone is insufficient", "sourceParagraph": 116 }, { "type": "bullet", "text": "To separate average accuracy from subgroup inequality", "sourceParagraph": 117 }, { "type": "bullet", "text": "To make visible the heavy error risk hidden within high averages", "sourceParagraph": 118 }, { "type": "bullet", "text": "To show that stability and accuracy are not the same concept", "sourceParagraph": 119 }, { "type": "bullet", "text": "To prevent unmeasured areas from being filled with positive assumptions within the distribution", "sourceParagraph": 120 }, { "type": "bullet", "text": "To prepare the structure of a future NOMOS score that is singular but not condemned to being singular", "sourceParagraph": 121 }, { "type": "paragraph", "text": "At the end of this section, the reader should be able to answer the following question:", "sourceParagraphs": [ 122 ] }, { "type": "callout", "text": "Why is the result of an entity's GEO not limited to only the average accuracy rate?", "sourceParagraph": 123 }, { "type": "heading2", "text": "2. CENTRAL NORMATIVE PROVISION", "sourceParagraph": 125 }, { "type": "callout", "text": "The representation of an entity in generative systems is the conditional distribution of different responses and representation outcomes that can occur under defined conditions.", "sourceParagraph": 126 }, { "type": "paragraph", "text": "This distribution:", "sourceParagraphs": [ 127 ] }, { "type": "bullet", "text": "it is not the abstract 'thought' of a single model,", "sourceParagraph": 128 }, { "type": "bullet", "text": "it is not the common view of all artificial intelligence systems,", "sourceParagraph": 129 }, { "type": "bullet", "text": "it is not only the repetition rate of the website,", "sourceParagraph": 130 }, { "type": "bullet", "text": "it's not only the frequency with which the brand's name is mentioned,", "sourceParagraph": 131 }, { "type": "bullet", "text": "it's not just the average accuracy,", "sourceParagraph": 132 }, { "type": "paragraph", "text": "it is not only the recommendation rate. The representation distribution, within a certain protocol, shows the following:", "sourceParagraphs": [ 133, 134 ] }, { "type": "callout", "text": "Which users see what kind of representation on which AI product and user surface, in which language, in which country, with which prompt, under which session condition, and at what time.", "sourceParagraph": 135 }, { "type": "paragraph", "text": "Therefore, the result of a distribution is always dependent on the following questions: Which entity? Which AI product? Which user surface? Which target population? Which country? Which language and locale? Which prompt? Which session condition? Which measurement wave? Which Truth Pack? Which evaluation rule? When any of these fields change, the distribution may also change.", "sourceParagraphs": [ 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148 ] }, { "type": "heading2", "text": "3. WHAT DOES THE WORD \"DISTRIBUTION\" MEAN?", "sourceParagraph": 150 }, { "type": "paragraph", "text": "In this book, the word distribution does not refer to a hidden or metaphysical view that is assumed to exist within an artificial intelligence system. It does not claim: NOMOS: \"This is the real distribution of the company in the model's mind.\" The distribution here is operational and observable.", "sourceParagraphs": [ 151, 152, 153, 154, 155 ] }, { "type": "callout", "text": "It is the pattern of different representation results that can be obtained under a defined protocol within the target population and measurement conditions.", "sourceParagraph": 156 }, { "type": "paragraph", "text": "Two distributions should be distinguished.", "sourceParagraphs": [ 157 ] }, { "type": "heading3", "text": "3.1. Target Representation Distribution", "sourceParagraph": 159 }, { "type": "paragraph", "text": "It is the distribution that the measurement tries to predict but cannot directly observe in full. It is represented as:", "sourceParagraphs": [ 160, 161 ] }, { "type": "heading3", "text": "P_E(Y | A, C, L, S, P, T, U)", "sourceParagraph": 162 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 163 ] }, { "type": "bullet", "text": "E: audited entity", "sourceParagraph": 164 }, { "type": "bullet", "text": "Y: representation result", "sourceParagraph": 165 }, { "type": "bullet", "text": "A: AI product and user surface", "sourceParagraph": 166 }, { "type": "bullet", "text": "C: country or geographic condition", "sourceParagraph": 167 }, { "type": "bullet", "text": "L: language and locale", "sourceParagraph": 168 }, { "type": "bullet", "text": "S: session and personalisation condition", "sourceParagraph": 169 }, { "type": "bullet", "text": "P: prompt", "sourceParagraph": 170 }, { "type": "bullet", "text": "T: time or measurement wave", "sourceParagraph": 171 }, { "type": "bullet", "text": "U: target user universe", "sourceParagraph": 172 }, { "type": "paragraph", "text": "This expression asks: “In the defined target user universe, under the specified conditions, which representation results occur with what probabilities?” Not all elements of the target distribution can be measured. They are estimated through sampling.", "sourceParagraphs": [ 173, 174, 175, 176 ] }, { "type": "heading3", "text": "3.2. Observed Empirical Distribution", "sourceParagraph": 178 }, { "type": "paragraph", "text": "It is the distribution of valid observations obtained in real or synthetic measurement. Its representation:", "sourceParagraphs": [ 179, 180 ] }, { "type": "heading3", "text": "P̂_E", "sourceParagraph": 181 }, { "type": "paragraph", "text": "can be used. The empirical distribution is not the target distribution itself. It is an estimate produced through its sample, weighting, and measurement design. The magnitude of the difference may depend on:", "sourceParagraphs": [ 182, 183, 184, 185 ] }, { "type": "bullet", "text": "Sample design", "sourceParagraph": 186 }, { "type": "bullet", "text": "Participant selection", "sourceParagraph": 187 }, { "type": "bullet", "text": "Country and language coverage", "sourceParagraph": 188 }, { "type": "bullet", "text": "Invalid observation rate", "sourceParagraph": 189 }, { "type": "bullet", "text": "Non-response", "sourceParagraph": 190 }, { "type": "bullet", "text": "Adjudicator error", "sourceParagraph": 191 }, { "type": "bullet", "text": "Measurement wave", "sourceParagraph": 192 }, { "type": "bullet", "text": "Prompt selection", "sourceParagraph": 193 }, { "type": "bullet", "text": "Product and interface conditions", "sourceParagraph": 194 }, { "type": "bullet", "text": "Weighting", "sourceParagraph": 195 }, { "type": "bullet", "text": "Clustering and dependence", "sourceParagraph": 196 }, { "type": "paragraph", "text": "Therefore, an empirical distribution cannot be published as \"the absolute and immutable truth of artificial intelligence.\" The correct expression should be: \"The estimate obtained for the specified target population and measurement conditions.\"", "sourceParagraphs": [ 197, 198, 199, 200, 201, 202 ] }, { "type": "heading2", "text": "4. REPRESENTATION SPACE", "sourceParagraph": 204 }, { "type": "paragraph", "text": "The representation of an entity does not occur in a single dimension. Each observation exists in a multidimensional condition space. The basic representation:", "sourceParagraphs": [ 205, 206, 207 ] }, { "type": "formula", "text": "X_i = (E_i, A_i, U_i, C_i, L_i, S_i, P_i, T_i)", "sourceParagraph": 208 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 209 ] }, { "type": "bullet", "text": "Ei: entity", "sourceParagraph": 210 }, { "type": "bullet", "text": "Ai: AI product and user surface", "sourceParagraph": 211 }, { "type": "bullet", "text": "Ui: user or sampling unit", "sourceParagraph": 212 }, { "type": "bullet", "text": "Ci: country and geographical layer", "sourceParagraph": 213 }, { "type": "bullet", "text": "Li: language and locale", "sourceParagraph": 214 }, { "type": "bullet", "text": "Si: session, plan, and personalisation condition", "sourceParagraph": 215 }, { "type": "bullet", "text": "Pi: prompt or prompt family", "sourceParagraph": 216 }, { "type": "bullet", "text": "Ti: time and measurement wave", "sourceParagraph": 217 }, { "type": "paragraph", "text": "Generated response:", "sourceParagraphs": [ 218 ] }, { "type": "formula", "text": "R_i", "sourceParagraph": 219 }, { "type": "paragraph", "text": "can be shown. The representative state resulting from the evaluation of the response is:", "sourceParagraphs": [ 220, 221 ] }, { "type": "formula", "text": "Y_i = g(R_i, V_E, J)", "sourceParagraph": 222 }, { "type": "paragraph", "text": "and can be considered as. Here:", "sourceParagraphs": [ 223, 224 ] }, { "type": "bullet", "text": "VE: verified Truth Pack used for the entity", "sourceParagraph": 225 }, { "type": "bullet", "text": "J: assessment and adjudication rules", "sourceParagraph": 226 }, { "type": "bullet", "text": "g: evaluation process that converts the response into representative results", "sourceParagraph": 227 }, { "type": "paragraph", "text": "The meaning of this representation is as follows: The same response cannot be scored on its own without indicating which entity, which Truth Pack, and which evaluation rule it was examined according to.", "sourceParagraphs": [ 228, 229 ] }, { "type": "heading2", "text": "5. REPRESENTATIVE RESULT IS NOT A SINGLE VALUE", "sourceParagraph": 231 }, { "type": "paragraph", "text": "Only the entirety of a response:", "sourceParagraphs": [ 232 ] }, { "type": "bullet", "text": "can be forced into the categories of\n\ntrue,", "sourceParagraph": 233 }, { "type": "bullet", "text": "is wrong", "sourceParagraph": 234 }, { "type": "paragraph", "text": "Giving a label is often insufficient. The following situations can occur together in the same response:", "sourceParagraphs": [ 235, 236 ] }, { "type": "bullet", "text": "Correct identity", "sourceParagraph": 237 }, { "type": "bullet", "text": "Correct main category", "sourceParagraph": 238 }, { "type": "bullet", "text": "Incomplete service coverage", "sourceParagraph": 239 }, { "type": "bullet", "text": "Outdated number of employees", "sourceParagraph": 240 }, { "type": "bullet", "text": "Unsupported leadership claim", "sourceParagraph": 241 }, { "type": "bullet", "text": "Correct citation", "sourceParagraph": 242 }, { "type": "bullet", "text": "Inappropriate recommendation", "sourceParagraph": 243 }, { "type": "bullet", "text": "Non-critical boundary deficiency", "sourceParagraph": 244 }, { "type": "paragraph", "text": "Therefore, the representation result can be considered as a result vector. Candidate representation:", "sourceParagraphs": [ 245, 246 ] }, { "type": "formula", "text": "Y_i = (y_i^id, y_i^cat, y_i^fact, y_i^scope, y_i^time, y_i^evidence, y_i^recommendation, y_i^harm)", "sourceParagraph": 247 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 248 ] }, { "type": "bullet", "text": "yiid: identity accuracy", "sourceParagraph": 249 }, { "type": "bullet", "text": "yicat: accuracy of category and role", "sourceParagraph": 250 }, { "type": "bullet", "text": "yifact: accuracy of factual matters", "sourceParagraph": 251 }, { "type": "bullet", "text": "yiscope: preservation of scope and boundaries", "sourceParagraph": 252 }, { "type": "bullet", "text": "yitime: temporal validity", "sourceParagraph": 253 }, { "type": "bullet", "text": "yievidence: claim–evidence consistency", "sourceParagraph": 254 }, { "type": "bullet", "text": "yirecommendation: recommendation and user suitability", "sourceParagraph": 255 }, { "type": "bullet", "text": "yiharm: harm class of error or misguidance", "sourceParagraph": 256 }, { "type": "paragraph", "text": "The exact scoring rules for these fields will be defined in later sections. The guiding principle here is:", "sourceParagraphs": [ 257, 258 ] }, { "type": "callout", "text": "A representative answer should be evaluated not only on whether it contains correct information but also on whom, what, within which boundaries, with what evidence, and for whom it is appropriately conveyed.", "sourceParagraph": 259 }, { "type": "heading2", "text": "6. BASIC REPRESENTATION STATES", "sourceParagraph": 261 }, { "type": "paragraph", "text": "The status of each atomic claim and final response may not be the same. This section conceptually separates the following basic states.", "sourceParagraphs": [ 262, 263 ] }, { "type": "heading3", "text": "6.1. SUPPORTED", "sourceParagraph": 265 }, { "type": "paragraph", "text": "The claim is supported by sufficient evidence within the defined scope and time.", "sourceParagraphs": [ 266 ] }, { "type": "heading3", "text": "6.2. CONTRADICTED", "sourceParagraph": 268 }, { "type": "paragraph", "text": "The claim conflicts with reliable and relevant records.", "sourceParagraphs": [ 269 ] }, { "type": "heading3", "text": "6.3. UNSUPPORTED", "sourceParagraph": 271 }, { "type": "paragraph", "text": "There is not enough support for the claim. It has not been proven that the claim is necessarily false. However, there are not enough records for it to be accepted as true.", "sourceParagraphs": [ 272, 273, 274 ] }, { "type": "heading3", "text": "6.4. OUTDATED — NO LONGER CURRENT", "sourceParagraph": 276 }, { "type": "paragraph", "text": "The claim may have been true in the past. As of the measurement date, it is no longer valid or its current accuracy cannot be verified.", "sourceParagraphs": [ 277, 278 ] }, { "type": "heading3", "text": "6.5. PARTIALLY SUPPORTED — PARTIALLY SUPPORTED", "sourceParagraph": 280 }, { "type": "paragraph", "text": "Only a specific scope, population, country, product, or time of the claim is supported. The response may have exceeded the support limit.", "sourceParagraphs": [ 281, 282 ] }, { "type": "heading3", "text": "6.6. OUT OF SCOPE — OUT OF SCOPE", "sourceParagraph": 284 }, { "type": "paragraph", "text": "The claim is not related to the audited entity or Truth Pack. It may have come from another entity, product, country, or service.", "sourceParagraphs": [ 285, 286 ] }, { "type": "heading3", "text": "6.7. ENTITY CONFLATION — ENTITY MERGER", "sourceParagraph": 288 }, { "type": "paragraph", "text": "System:", "sourceParagraphs": [ 289 ] }, { "type": "bullet", "text": "two different companies,", "sourceParagraph": 290 }, { "type": "bullet", "text": "the brand and the legal entity,", "sourceParagraph": 291 }, { "type": "bullet", "text": "a person and the institution,", "sourceParagraph": 292 }, { "type": "bullet", "text": "old and new entity,", "sourceParagraph": 293 }, { "type": "bullet", "text": "has confused different organisations with the same name", "sourceParagraph": 294 }, { "type": "paragraph", "text": "with each other. This error is not merely a factual mistake. It is the corruption of the representation object.", "sourceParagraphs": [ 295, 296, 297 ] }, { "type": "heading3", "text": "6.8. MATERIAL OMISSION — MISSING INFORMATION", "sourceParagraph": 299 }, { "type": "paragraph", "text": "The sentences in the response may be correct. However, essential information that would change the user's decision has been omitted. Example: The service is offered only in certain countries. The licence is limited to a specific jurisdiction. Price is excluding tax. The result is based on first-party data. The product is not suitable for certain users. The omission can create a wrong inference even if there is no directly incorrect sentence.", "sourceParagraphs": [ 300, 301, 302, 303, 304, 305, 306, 307, 308 ] }, { "type": "heading3", "text": "6.9. REFUSAL — REJECTION", "sourceParagraph": 310 }, { "type": "paragraph", "text": "The system has refused to answer. Rejection:", "sourceParagraphs": [ 311, 312 ] }, { "type": "bullet", "text": "appropriate,", "sourceParagraph": 313 }, { "type": "bullet", "text": "unnecessary,", "sourceParagraph": 314 }, { "type": "bullet", "text": "inconsistent,", "sourceParagraph": 315 }, { "type": "bullet", "text": "security-based,", "sourceParagraph": 316 }, { "type": "bullet", "text": "technical", "sourceParagraph": 317 }, { "type": "paragraph", "text": "may occur. Ret is not automatically considered failed or successful. It is classified separately.", "sourceParagraphs": [ 318, 319, 320 ] }, { "type": "heading3", "text": "6.10. NO RESULT", "sourceParagraph": 322 }, { "type": "paragraph", "text": "The system did not produce a usable response due to technical or operational reasons. It is not the same as Ret.", "sourceParagraphs": [ 323, 324 ] }, { "type": "heading3", "text": "6.11. UNKNOWN", "sourceParagraph": 326 }, { "type": "paragraph", "text": "The available records are not sufficient for a reliable classification. UNKNOWN cannot be converted into a positive result.", "sourceParagraphs": [ 327, 328 ] }, { "type": "heading2", "text": "7. MAIN CHARACTERISTICS OF THE DISTRIBUTION", "sourceParagraph": 330 }, { "type": "paragraph", "text": "A representation distribution cannot be explained solely by the mean. At a minimum, the following characteristics should be evaluated.", "sourceParagraphs": [ 331, 332 ] }, { "type": "heading3", "text": "7.1. Central Tendency", "sourceParagraph": 334 }, { "type": "paragraph", "text": "Summarises the general accuracy or transition level of the responses. Example measures:", "sourceParagraphs": [ 335, 336 ] }, { "type": "bullet", "text": "Mean", "sourceParagraph": 337 }, { "type": "bullet", "text": "Weighted mean", "sourceParagraph": 338 }, { "type": "bullet", "text": "Median", "sourceParagraph": 339 }, { "type": "bullet", "text": "pass rate", "sourceParagraph": 340 }, { "type": "paragraph", "text": "Central tendency is important. However, it only shows the centre of the distribution.", "sourceParagraphs": [ 341, 342 ] }, { "type": "heading3", "text": "7.2. Spread and Variability", "sourceParagraph": 344 }, { "type": "paragraph", "text": "Indicates how much the results are spread around the centre. Of two systems with the same mean, one may be consistent among users, while the other may be highly variable. Variability:", "sourceParagraphs": [ 345, 346, 347 ] }, { "type": "bullet", "text": "users,", "sourceParagraph": 348 }, { "type": "bullet", "text": "countries,", "sourceParagraph": 349 }, { "type": "bullet", "text": "languages,", "sourceParagraph": 350 }, { "type": "bullet", "text": "sessions,", "sourceParagraph": 351 }, { "type": "bullet", "text": "measurement waves", "sourceParagraph": 352 }, { "type": "paragraph", "text": "can also be examined among them.", "sourceParagraphs": [ 353 ] }, { "type": "heading3", "text": "7.3. Shape of the Distribution", "sourceParagraph": 355 }, { "type": "paragraph", "text": "Are the results clustered around a single central point? Or are they divided into two or more clusters? For example:", "sourceParagraphs": [ 356, 357, 358 ] }, { "type": "bullet", "text": "If half of the users see very accurate,", "sourceParagraph": 359 }, { "type": "bullet", "text": "and half see very inaccurate", "sourceParagraph": 360 }, { "type": "paragraph", "text": "responses, the average may be medium. However, the typical user experience is not \"medium accuracy.\" It is two separate polarised experiences.", "sourceParagraphs": [ 361, 362, 363 ] }, { "type": "heading3", "text": "7.4. Subgroup Differences", "sourceParagraph": 365 }, { "type": "paragraph", "text": "Do specific language, country, or user groups systematically see different results? The global average may hide this difference.", "sourceParagraphs": [ 366, 367 ] }, { "type": "heading3", "text": "7.5. Tail Risk", "sourceParagraph": 369 }, { "type": "paragraph", "text": "Are there low-frequency but severe errors? Rare events can have a small effect on the average. Their impact on human harm can be significant.", "sourceParagraphs": [ 370, 371, 372 ] }, { "type": "heading3", "text": "7.6. Coverage", "sourceParagraph": 374 }, { "type": "paragraph", "text": "How much of the target user universe has actually been measured? A high score can be found together with low coverage.", "sourceParagraphs": [ 375, 376 ] }, { "type": "heading3", "text": "7.7. Missing and Unknown Mass", "sourceParagraph": 378 }, { "type": "paragraph", "text": "What proportion of the responses are:", "sourceParagraphs": [ 379 ] }, { "type": "heading3", "text": "UNKNOWN,", "sourceParagraph": 380 }, { "type": "heading3", "text": "NOT TESTED,", "sourceParagraph": 381 }, { "type": "paragraph", "text": "rejection, technical non-response or an invalid record?", "sourceParagraphs": [ 382, 383, 384, 385 ] }, { "type": "heading3", "text": "7.8. Time Stability", "sourceParagraph": 387 }, { "type": "paragraph", "text": "To what extent is the distribution preserved across different measurement waves?", "sourceParagraphs": [ 388 ] }, { "type": "heading2", "text": "8. CANDIDATE NOMOS REPRESENTATIVE DISTRIBUTION SIGNATURE", "sourceParagraph": 390 }, { "type": "heading3", "text": "CANDIDATE CONCEPT", "sourceParagraph": 391 }, { "type": "paragraph", "text": "It is recommended that the result of a GEO be defined first by the following distribution signature instead of a single number:", "sourceParagraphs": [ 392 ] }, { "type": "formula", "text": "RDS(E) = (μ, κ, σ, Δ, τ, ρ, u)", "sourceParagraph": 393 }, { "type": "paragraph", "text": "This representation is called Representation Distribution Signature, that is:", "sourceParagraphs": [ 394 ] }, { "type": "heading4", "text": "Representation Distribution Signature", "sourceParagraph": 395 }, { "type": "paragraph", "text": "Components:", "sourceParagraphs": [ 396, 397 ] }, { "type": "heading4", "text": "μ — General Representation Level", "sourceParagraph": 398 }, { "type": "paragraph", "text": "Overall accuracy or transition prediction under defined weights and scope.", "sourceParagraphs": [ 399 ] }, { "type": "heading4", "text": "κ — Coverage", "sourceParagraph": 400 }, { "type": "paragraph", "text": "The extent to which the target user, country, language, AI product, and time universe are sufficiently measured.", "sourceParagraphs": [ 401 ] }, { "type": "heading4", "text": "σ — Variability", "sourceParagraph": 402 }, { "type": "paragraph", "text": "Spread among users and experimental cells.", "sourceParagraphs": [ 403 ] }, { "type": "heading4", "text": "Δ — Subgroup Inequality", "sourceParagraph": 404 }, { "type": "paragraph", "text": "The material difference between language, country, product, or other significant groups.", "sourceParagraphs": [ 405 ] }, { "type": "heading4", "text": "'tau — Tail and Critical Error Risk", "sourceParagraph": 406 }, { "type": "paragraph", "text": "The rate and nature of rare severe errors.", "sourceParagraphs": [ 407 ] }, { "type": "heading4", "text": "ρ — Temporal Stability", "sourceParagraph": 408 }, { "type": "paragraph", "text": "The resilience of distribution across waves.", "sourceParagraphs": [ 409 ] }, { "type": "heading4", "text": "u — Unknown and Unresolved Mass", "sourceParagraph": 410 }, { "type": "paragraph", "text": "UNKNOWN is the share of the distribution that remains unresolved or insufficiently assessed. This signature is not yet a final scoring formula. The following elements require further development:", "sourceParagraphs": [ 411, 412, 413 ] }, { "type": "bullet", "text": "definite definition,", "sourceParagraph": 414 }, { "type": "bullet", "text": "estimation,", "sourceParagraph": 415 }, { "type": "bullet", "text": "weight,", "sourceParagraph": 416 }, { "type": "bullet", "text": "public presentation", "sourceParagraph": 417 }, { "type": "paragraph", "text": "They will be finalised through pilot studies and later chapters. The founding principle is:", "sourceParagraphs": [ 418, 419 ] }, { "type": "callout", "text": "The GEO result should answer not only the question 'how accurate?' but also 'for whom, how consistent, how equitable, how comprehensive, and how risky?'", "sourceParagraph": 420 }, { "type": "heading2", "text": "9. AVERAGE WASHING", "sourceParagraph": 422 }, { "type": "paragraph", "text": "If an institution only shows a high global average and hides the material flaws of the distribution, this is called:", "sourceParagraphs": [ 423 ] }, { "type": "heading4", "text": "Average washing", "sourceParagraph": 424 }, { "type": "paragraph", "text": "Average washing can occur in the following ways:", "sourceParagraphs": [ 425, 426 ] }, { "type": "bullet", "text": "Hiding the weakest language", "sourceParagraph": 427 }, { "type": "bullet", "text": "Excluding small countries from the report", "sourceParagraph": 428 }, { "type": "bullet", "text": "Mixing critical errors into the general error rate", "sourceParagraph": 429 }, { "type": "bullet", "text": "Removing UNKNOWN results from the denominator", "sourceParagraph": 430 }, { "type": "bullet", "text": "Only including well-performing AI products", "sourceParagraph": 431 }, { "type": "bullet", "text": "Excluding the weak measurement wave", "sourceParagraph": 432 }, { "type": "bullet", "text": "Making the population outside the scope invisible", "sourceParagraph": 433 }, { "type": "bullet", "text": "Changing weights after seeing the results", "sourceParagraph": 434 }, { "type": "paragraph", "text": "The average may be mathematically calculated correctly. However, the representation it offers to the public can still be misleading. Therefore:", "sourceParagraphs": [ 435, 436, 437 ] }, { "type": "callout", "text": "A mathematically correct calculation of an average does not mean it is sufficient for decision-making.", "sourceParagraph": 438 }, { "type": "heading2", "text": "10. DIFFERENT REALITIES OF THE SAME AVERAGE", "sourceParagraph": 440 }, { "type": "paragraph", "text": "The average could be 90 in all of the following synthetic distributions.", "sourceParagraphs": [ 441 ] }, { "type": "heading3", "text": "10.1. Homogeneous Distribution", "sourceParagraph": 443 }, { "type": "paragraph", "text": "All major subgroups are in the range of 88–92. This distribution is relatively balanced.", "sourceParagraphs": [ 444, 445 ] }, { "type": "heading3", "text": "10.2. Polarised Distribution", "sourceParagraph": 447 }, { "type": "paragraph", "text": "Half of the users see results close to 100, while the other half see results close to 80. The average is 90. However, the user experience is not equal.", "sourceParagraphs": [ 448, 449, 450 ] }, { "type": "heading3", "text": "10.3. Linguistically Fragmented Distribution", "sourceParagraph": 452 }, { "type": "paragraph", "text": "Results are very good for high-population or high-resource languages, but significantly weak for low-resource languages. The global average may look good. Language fairness is weak.", "sourceParagraphs": [ 453, 454, 455 ] }, { "type": "heading3", "text": "10.4. Country-Based Fragmented Distribution", "sourceParagraph": 457 }, { "type": "paragraph", "text": "Even within the same language, results may vary due to different countries, locales, product access, or user base.", "sourceParagraphs": [ 458 ] }, { "type": "heading3", "text": "10.5. Temporally Volatile Distribution", "sourceParagraph": 460 }, { "type": "paragraph", "text": "Very high results occur in one wave, very low results in another wave. Even if the average is high, it is difficult to establish reliable user expectations.", "sourceParagraphs": [ 461, 462 ] }, { "type": "heading3", "text": "10.6. Distribution with Tail Risk", "sourceParagraph": 464 }, { "type": "paragraph", "text": "Most of the responses are correct. However, a small portion contains serious health, legal, licensing, or safety errors.", "sourceParagraphs": [ 465, 466 ] }, { "type": "heading3", "text": "10.7. High Unknown Mass Distribution", "sourceParagraph": 468 }, { "type": "paragraph", "text": "Most of the assessable responses are correct. However, a significant portion of the target population or responses is:", "sourceParagraphs": [ 469, 470 ] }, { "type": "bullet", "text": "tested,", "sourceParagraph": 471 }, { "type": "bullet", "text": "untested,", "sourceParagraph": 472 }, { "type": "heading3", "text": "UNKNOWN", "sourceParagraph": 473 }, { "type": "paragraph", "text": "or inconclusive. Publishing a high rate based only on the assessable responses may obscure the scope.", "sourceParagraphs": [ 474, 475 ] }, { "type": "heading2", "text": "11. SUBGROUP INEQUALITY", "sourceParagraph": 477 }, { "type": "paragraph", "text": "A global score may not represent the experience in subgroups. Key subgroups may include:", "sourceParagraphs": [ 478, 479 ] }, { "type": "bullet", "text": "AI product and user surface", "sourceParagraph": 480 }, { "type": "bullet", "text": "Country", "sourceParagraph": 481 }, { "type": "bullet", "text": "Language", "sourceParagraph": 482 }, { "type": "bullet", "text": "Locale", "sourceParagraph": 483 }, { "type": "bullet", "text": "Plan", "sourceParagraph": 484 }, { "type": "bullet", "text": "Web or mobile", "sourceParagraph": 485 }, { "type": "bullet", "text": "Controlled or natural panel", "sourceParagraph": 486 }, { "type": "bullet", "text": "Time wave", "sourceParagraph": 487 }, { "type": "bullet", "text": "User intent", "sourceParagraph": 488 }, { "type": "bullet", "text": "Device family", "sourceParagraph": 489 }, { "type": "paragraph", "text": "Not all of these groups may have the same importance in every audit. Material groups should be determined when scope is locked.", "sourceParagraphs": [ 490, 491 ] }, { "type": "heading3", "text": "11.1. Language is Not the Same as Country", "sourceParagraph": 493 }, { "type": "paragraph", "text": "Turkish-speaking:", "sourceParagraphs": [ 494 ] }, { "type": "bullet", "text": "User in Turkey,", "sourceParagraph": 495 }, { "type": "bullet", "text": "User in Germany,", "sourceParagraph": 496 }, { "type": "bullet", "text": "User in another country", "sourceParagraph": 497 }, { "type": "paragraph", "text": "They can use the same language. However:", "sourceParagraphs": [ 498, 499 ] }, { "type": "bullet", "text": "product access,", "sourceParagraph": 500 }, { "type": "bullet", "text": "price,", "sourceParagraph": 501 }, { "type": "bullet", "text": "resources,", "sourceParagraph": 502 }, { "type": "bullet", "text": "local interface,", "sourceParagraph": 503 }, { "type": "bullet", "text": "legal context,", "sourceParagraph": 504 }, { "type": "bullet", "text": "personalisation", "sourceParagraph": 505 }, { "type": "paragraph", "text": "may be different. Therefore, all differences cannot be explained solely by language or country.", "sourceParagraphs": [ 506, 507 ] }, { "type": "heading3", "text": "11.2. Intersectional Cells", "sourceParagraph": 509 }, { "type": "paragraph", "text": "Some material differences only become visible when multiple areas are examined together. Example: Germany × Turkish × Mobile × Free plan × Natural User Panel could in itself be an experimental cell. However, as cells multiply, the sample size decreases. Therefore, a definitive score cannot be published for each small cell. Measurement design: should preserve meaningful subgroups and should not produce statistical false precision.", "sourceParagraphs": [ 510, 511, 512, 513, 514, 515, 516, 517, 518 ] }, { "type": "heading3", "text": "11.3. Subgroup Base", "sourceParagraph": 520 }, { "type": "heading3", "text": "CANDIDATE CONCEPT", "sourceParagraph": 521 }, { "type": "paragraph", "text": "In addition to the global average, it is recommended to report the result of the weakest adequately sampled subgroup or the bottom 10% group separately. This can be referred to as:", "sourceParagraphs": [ 522, 523 ] }, { "type": "heading4", "text": "Representation base", "sourceParagraph": 524 }, { "type": "paragraph", "text": "or:", "sourceParagraphs": [ 525 ] }, { "type": "heading4", "text": "Language/Country Floor", "sourceParagraph": 526 }, { "type": "paragraph", "text": "However, if the \"lowest score\" comes from a small and unstable cell, it may produce incorrect results. Therefore, the base metric should be defined together with:", "sourceParagraphs": [ 527, 528, 529 ] }, { "type": "bullet", "text": "minimum sample,", "sourceParagraph": 530 }, { "type": "bullet", "text": "confidence interval,", "sourceParagraph": 531 }, { "type": "bullet", "text": "cell quality", "sourceParagraph": 532 }, { "type": "paragraph", "text": "The exact rule will be determined in the following sections.", "sourceParagraphs": [ 533, 534 ] }, { "type": "heading2", "text": "12. IT IS PART OF THE SCOPE DISTRIBUTION", "sourceParagraph": 536 }, { "type": "paragraph", "text": "A distribution consists not only of observed responses. What was never observed also determines the meaning of the result. For example, English and Turkish were tested; Japanese and Arabic were not. The web interface was tested; the mobile interface and paid plan were not. Strong English and Turkish results do not justify the statement, ‘Representation is strong across all languages and user surfaces.’ The defensible conclusion is: ‘Measurements were conducted on the specified English and Turkish web surfaces. Other languages and mobile surfaces were not tested.’", "sourceParagraphs": [ 537, 538, 539, 540, 541, 542, 543, 544, 545, 546, 547, 548, 549, 550, 551 ] }, { "type": "heading3", "text": "12.1. Coverage Map", "sourceParagraph": 553 }, { "type": "paragraph", "text": "Each GEO distribution should carry the following map to the extent relevant:", "sourceParagraphs": [ 554 ] }, { "type": "bullet", "text": "Target universe", "sourceParagraph": 555 }, { "type": "bullet", "text": "Sampling frame", "sourceParagraph": 556 }, { "type": "bullet", "text": "Covered countries", "sourceParagraph": 557 }, { "type": "bullet", "text": "Non-covered countries", "sourceParagraph": 558 }, { "type": "bullet", "text": "Covered languages", "sourceParagraph": 559 }, { "type": "bullet", "text": "Non-covered languages", "sourceParagraph": 560 }, { "type": "bullet", "text": "Monitored AI products", "sourceParagraph": 561 }, { "type": "bullet", "text": "Unmonitored products", "sourceParagraph": 562 }, { "type": "bullet", "text": "Monitored user surfaces", "sourceParagraph": 563 }, { "type": "bullet", "text": "Unmonitored surfaces", "sourceParagraph": 564 }, { "type": "bullet", "text": "Measurement waves", "sourceParagraph": 565 }, { "type": "bullet", "text": "Out-of-scope prompt families", "sourceParagraph": 566 }, { "type": "heading3", "text": "12.2. Coverage Rate", "sourceParagraph": 568 }, { "type": "heading3", "text": "CANDIDATE CONCEPT", "sourceParagraph": 569 }, { "type": "paragraph", "text": "A coverage metric can be developed to show what proportion of the target universe is represented with sufficient measurement. However, coverage:", "sourceParagraphs": [ 570, 571 ] }, { "type": "bullet", "text": "only the number of countries,", "sourceParagraph": 572 }, { "type": "bullet", "text": "only the number of languages,", "sourceParagraph": 573 }, { "type": "bullet", "text": "only the population ratio", "sourceParagraph": 574 }, { "type": "paragraph", "text": "cannot be measured. For example, 150 countries may be covered. However, 95% of the participants may have come from only five countries. Or 90% of the population may be covered, but small languages may be entirely excluded. Therefore, population coverage should be shown separately from country/language observer coverage.", "sourceParagraphs": [ 575, 576, 577, 578, 579 ] }, { "type": "heading2", "text": "13. MISSING OBSERVATION AND NON-RESPONSE", "sourceParagraph": 581 }, { "type": "paragraph", "text": "Missing data is not only a technical issue. Its distribution can be systematically distorted. The reasons for an observation being missing may include:", "sourceParagraphs": [ 582, 583, 584 ] }, { "type": "bullet", "text": "Participant's failure to complete the task", "sourceParagraph": 585 }, { "type": "bullet", "text": "Product not being accessible in the country", "sourceParagraph": 586 }, { "type": "bullet", "text": "Computation constraints", "sourceParagraph": 587 }, { "type": "bullet", "text": "Technical error", "sourceParagraph": 588 }, { "type": "bullet", "text": "System rejection", "sourceParagraph": 589 }, { "type": "bullet", "text": "Invalid screen capture", "sourceParagraph": 590 }, { "type": "bullet", "text": "Participant changing their prompt", "sourceParagraph": 591 }, { "type": "bullet", "text": "Product not working in a specific language", "sourceParagraph": 592 }, { "type": "bullet", "text": "Adjudicators being unable to make reliable decisions", "sourceParagraph": 593 }, { "type": "paragraph", "text": "Each of these reasons requires a separate status.", "sourceParagraphs": [ 594 ] }, { "type": "heading3", "text": "13.1. Missing Data May Not Be Random", "sourceParagraph": 596 }, { "type": "paragraph", "text": "For example, if more tasks cannot be completed in countries with low connection quality, using only successful recordings can shift the global result in favour of wealthy users with strong connections. If the rate of UNKNOWN rises because answers in a specific language are difficult to evaluate, the score calculated from only known answers may hide language inequality. Therefore:", "sourceParagraphs": [ 597, 598, 599 ] }, { "type": "callout", "text": "The missing data rate and the distribution report of the reason for missing data should be part of the report.", "sourceParagraph": 600 }, { "type": "heading3", "text": "13.2. Denominator — Integrity of the Denominator", "sourceParagraph": 602 }, { "type": "paragraph", "text": "A report should answer the question clearly: \"90 per cent of exactly what is 90 per cent?\" Possible denominators:", "sourceParagraphs": [ 603, 604, 605 ] }, { "type": "bullet", "text": "Assigned participants", "sourceParagraph": 606 }, { "type": "bullet", "text": "Task initiators", "sourceParagraph": 607 }, { "type": "bullet", "text": "Task completers", "sourceParagraph": 608 }, { "type": "bullet", "text": "Protocol-compliant records", "sourceParagraph": 609 }, { "type": "bullet", "text": "Valid system results", "sourceParagraph": 610 }, { "type": "bullet", "text": "Responses with adjudicator decisions", "sourceParagraph": 611 }, { "type": "bullet", "text": "Specific language or country cell", "sourceParagraph": 612 }, { "type": "paragraph", "text": "The denominator should not be changed after the result is seen. For example:", "sourceParagraphs": [ 613, 614 ] }, { "type": "bullet", "text": "1,000 tasks assigned,", "sourceParagraph": 615 }, { "type": "bullet", "text": "900 completed,", "sourceParagraph": 616 }, { "type": "bullet", "text": "800 valid,", "sourceParagraph": 617 }, { "type": "bullet", "text": "720 passed", "sourceParagraph": 618 }, { "type": "paragraph", "text": "Thus:", "sourceParagraphs": [ 619 ] }, { "type": "formula", "text": "720/800=90%", "sourceParagraph": 620 }, { "type": "paragraph", "text": "This is the pass rate among valid records. The observed pass rate across the entire assigned sample is instead:", "sourceParagraphs": [ 621, 622 ] }, { "type": "formula", "text": "720/1000=72%", "sourceParagraph": 623 }, { "type": "paragraph", "text": "answers a different question. Both rates can be used. They cannot be presented in the same sense.", "sourceParagraphs": [ 624, 625, 626 ] }, { "type": "heading2", "text": "14. ACCURACY AND CONSISTENCY ARE NOT THE SAME", "sourceParagraph": 628 }, { "type": "paragraph", "text": "A distribution may have high accuracy. However, it may be inconsistent over time or across users. Another distribution may have low accuracy. However, it may consistently be wrong. Therefore, four situations should be distinguished.", "sourceParagraphs": [ 629, 630, 631, 632, 633 ] }, { "type": "table", "rows": [ [ "Accuracy", "Consistency", "Comment" ], [ "High", "High", "Candidate for consistent and accurate representation" ], [ "High", "Low", "Mostly accurate but unreliable experience" ], [ "Low", "High", "Systematic and consistent misrepresentation" ], [ "Low", "Low", "Both incorrect and unstable representation" ] ] }, { "type": "paragraph", "text": "Being “consistent” is not automatically positive. The system may be quite consistent if it gives the same error to every user.", "sourceParagraphs": [ 634, 635 ] }, { "type": "heading3", "text": "14.1. Inter-User Consistency", "sourceParagraph": 637 }, { "type": "paragraph", "text": "Do different users see similar material representation in the same cell?", "sourceParagraphs": [ 638 ] }, { "type": "heading3", "text": "14.2. Inter-Session Consistency", "sourceParagraph": 640 }, { "type": "paragraph", "text": "Does the same or matched user see similar results in new sessions?", "sourceParagraphs": [ 641 ] }, { "type": "heading3", "text": "14.3. Inter-Wave Consistency", "sourceParagraph": 643 }, { "type": "paragraph", "text": "Does the same protocol produce a similar distribution on different days or periods?", "sourceParagraphs": [ 644 ] }, { "type": "heading3", "text": "14.4. Post-Product Update Consistency", "sourceParagraph": 646 }, { "type": "paragraph", "text": "When there is a material change in the AI product, does the previous distribution continue?", "sourceParagraphs": [ 647 ] }, { "type": "heading2", "text": "15. PRODUCING THE SAME SENTENCE IS NOT STABILITY", "sourceParagraph": 649 }, { "type": "paragraph", "text": "Stability: It does not mean \"giving the exact same answer to every user word for word.\" An AI product can use different expressions, order, and examples. As long as they preserve the material meaning, they can be acceptable variations. The important distinction is:", "sourceParagraphs": [ 650, 651, 652, 653, 654, 655 ] }, { "type": "heading4", "text": "Formal Variability", "sourceParagraph": 656 }, { "type": "paragraph", "text": "Word choice Sentence order Length Tone Example Using list or paragraph If material truth is preserved, it may not be an error.", "sourceParagraphs": [ 657, 658, 659, 660, 661, 662, 663 ] }, { "type": "heading4", "text": "Semantic Variability", "sourceParagraph": 664 }, { "type": "paragraph", "text": "Change of company identity Differentiation of the main category Expansion or contraction of the field of activity Change of licence or competency Change of price and geography Recommending to one user but not recommending to a similar user without explanation If the material representation changes, there may be a stability problem. Therefore, the preferred concept is:", "sourceParagraphs": [ 665, 666, 667, 668, 669, 670, 671, 672 ] }, { "type": "heading4", "text": "Conditional semantic stability", "sourceParagraph": 673 }, { "type": "paragraph", "text": "It should be. The fundamental reality is expected to be preserved under equivalent material conditions. A change in recommendation due to different user needs can be appropriate contextual behaviour.", "sourceParagraphs": [ 674, 675, 676 ] }, { "type": "heading2", "text": "16. CONTEXT SENSITIVITY AND ARBITRARY VARIABILITY", "sourceParagraph": 678 }, { "type": "paragraph", "text": "It is not always correct for a system to give the same recommendation to all users. For example:", "sourceParagraphs": [ 679, 680 ] }, { "type": "bullet", "text": "Users with a low budget", "sourceParagraph": 681 }, { "type": "bullet", "text": "Only users seeking services in a specific country", "sourceParagraph": 682 }, { "type": "bullet", "text": "Users requiring a different licence", "sourceParagraph": 683 }, { "type": "bullet", "text": "Users seeking individual rather than corporate services", "sourceParagraph": 684 }, { "type": "paragraph", "text": "can receive different conformity outcomes for the same entity. This difference:", "sourceParagraphs": [ 685, 686 ] }, { "type": "bullet", "text": "may be due to the user condition,", "sourceParagraph": 687 }, { "type": "bullet", "text": "the scope of service,", "sourceParagraph": 688 }, { "type": "bullet", "text": "or the actual limit", "sourceParagraph": 689 }, { "type": "paragraph", "text": "and may be context-sensitive and accurate. However, if the outcome changes without reason under equivalent user conditions, it may be arbitrary variability. Therefore, NOMOS does not want: “The same answer for all users.” It wants:", "sourceParagraphs": [ 690, 691, 692, 693, 694 ] }, { "type": "callout", "text": "A response that is consistent and explainable under equivalent material conditions and sensitive to actual limits under different conditions.", "sourceParagraph": 695 }, { "type": "heading2", "text": "17. TIME IS WITHIN THE REPRESENTATION DISTRIBUTION", "sourceParagraph": 697 }, { "type": "paragraph", "text": "The representation distribution is not timeless. It can be thought of as PE,t. The distribution of the same entity:", "sourceParagraphs": [ 698, 699, 700, 701, 702 ] }, { "type": "bullet", "text": "product update,", "sourceParagraph": 703 }, { "type": "bullet", "text": "resource change,", "sourceParagraph": 704 }, { "type": "bullet", "text": "company change,", "sourceParagraph": 705 }, { "type": "bullet", "text": "news event,", "sourceParagraph": 706 }, { "type": "bullet", "text": "new language version,", "sourceParagraph": 707 }, { "type": "bullet", "text": "technical access change", "sourceParagraph": 708 }, { "type": "paragraph", "text": "can diverge afterwards.", "sourceParagraphs": [ 709 ] }, { "type": "heading3", "text": "17.1. Sudden Change", "sourceParagraph": 711 }, { "type": "paragraph", "text": "After a product or resource update, the distribution changes quickly in a short time.", "sourceParagraphs": [ 712 ] }, { "type": "heading3", "text": "17.2. Gradual Change", "sourceParagraph": 714 }, { "type": "paragraph", "text": "Due to new resources and user behaviours, the result slowly changes over time.", "sourceParagraphs": [ 715 ] }, { "type": "heading3", "text": "17.3. Temporary Fluctuation", "sourceParagraph": 717 }, { "type": "paragraph", "text": "During a specific event or news period, the distribution temporarily differs.", "sourceParagraphs": [ 718 ] }, { "type": "heading3", "text": "17.4. Post-Intervention Change", "sourceParagraph": 720 }, { "type": "paragraph", "text": "After the GEO intervention, the distribution may change. However, claiming that the change is due to the intervention also requires a baseline and causality design.", "sourceParagraphs": [ 721, 722 ] }, { "type": "heading3", "text": "17.5. Smoothing with Time Average", "sourceParagraph": 724 }, { "type": "paragraph", "text": "For a product:", "sourceParagraphs": [ 725 ] }, { "type": "bullet", "text": "in the first wave 98,", "sourceParagraph": 726 }, { "type": "bullet", "text": "in the second wave 90,", "sourceParagraph": 727 }, { "type": "bullet", "text": "in the third wave 82", "sourceParagraph": 728 }, { "type": "paragraph", "text": "If taken, the average is 90. However, publishing 90 hides the downward trend. Therefore:", "sourceParagraphs": [ 729, 730, 731 ] }, { "type": "callout", "text": "The time average cannot replace the wave results.", "sourceParagraph": 732 }, { "type": "heading2", "text": "18. TAIL RISK AND HEAVY ERROR DISTRIBUTION", "sourceParagraph": 734 }, { "type": "paragraph", "text": "A large portion of a distribution may be correct. However, low-probability heavy errors should also be evaluated. Examples:", "sourceParagraphs": [ 735, 736, 737 ] }, { "type": "bullet", "text": "Health licence that does not actually exist", "sourceParagraph": 738 }, { "type": "bullet", "text": "Incorrect emergency guidance", "sourceParagraph": 739 }, { "type": "bullet", "text": "Incorrect legal authority", "sourceParagraph": 740 }, { "type": "bullet", "text": "Fake investment guarantee", "sourceParagraph": 741 }, { "type": "bullet", "text": "Incorrect company identity", "sourceParagraph": 742 }, { "type": "bullet", "text": "Inappropriate advice for children or vulnerable groups", "sourceParagraph": 743 }, { "type": "bullet", "text": "Repetition of the manipulative instruction hidden from people", "sourceParagraph": 744 }, { "type": "paragraph", "text": "In the previous architecture:", "sourceParagraphs": [ 745 ] }, { "type": "bullet", "text": "Critical,", "sourceParagraph": 746 }, { "type": "bullet", "text": "Major,", "sourceParagraph": 747 }, { "type": "bullet", "text": "Moderate,", "sourceParagraph": 748 }, { "type": "bullet", "text": "Advisory", "sourceParagraph": 749 }, { "type": "paragraph", "text": "were separated into four levels of importance, and it was suggested that Critical findings should have an automatic failure effect. This book links this system to population distribution.", "sourceParagraphs": [ 750, 751 ] }, { "type": "heading3", "text": "18.1. Separation of Frequency and Severity", "sourceParagraph": 753 }, { "type": "paragraph", "text": "An error:", "sourceParagraphs": [ 754 ] }, { "type": "bullet", "text": "frequent but low impact,", "sourceParagraph": 755 }, { "type": "bullet", "text": "rare but severe", "sourceParagraph": 756 }, { "type": "paragraph", "text": "may occur. However, the frequency rate alone cannot evaluate the two cases equally.", "sourceParagraphs": [ 757, 758 ] }, { "type": "heading3", "text": "18.2. Candidate Severity-Weighted Risk Representation", "sourceParagraph": 760 }, { "type": "heading3", "text": "CANDIDATE CONCEPT", "sourceParagraph": 761 }, { "type": "paragraph", "text": "A risk indicator can be developed using severity coefficients for error classes:", "sourceParagraphs": [ 762 ] }, { "type": "formula", "text": "tau = Σ_{k=1}^K λ_k p_k", "sourceParagraph": 763 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 764 ] }, { "type": "bullet", "text": "pk: predicted rate of the error type", "sourceParagraph": 765 }, { "type": "bullet", "text": "λk: severity coefficient of the error type", "sourceParagraph": 766 }, { "type": "paragraph", "text": "However, coefficients cannot be determined arbitrarily. They should be based on:", "sourceParagraphs": [ 767, 768 ] }, { "type": "bullet", "text": "Foreseeable human harm", "sourceParagraph": 769 }, { "type": "bullet", "text": "Representation integrity", "sourceParagraph": 770 }, { "type": "bullet", "text": "Impact on user decision", "sourceParagraph": 771 }, { "type": "bullet", "text": "Reversibility", "sourceParagraph": 772 }, { "type": "bullet", "text": "Intent and manipulation", "sourceParagraph": 773 }, { "type": "bullet", "text": "High-risk area", "sourceParagraph": 774 }, { "type": "paragraph", "text": "The precise method will be defined in Section 15.", "sourceParagraphs": [ 775 ] }, { "type": "heading3", "text": "18.3. Critical Gate", "sourceParagraph": 777 }, { "type": "paragraph", "text": "A high overall average cannot automatically delete a verified Critical error candidate. The correct public record should show both results together: “The overall representation estimate is high.” and: “The specified Critical event or events also affect the conformity decision.”", "sourceParagraphs": [ 778, 779, 780, 781, 782 ] }, { "type": "heading2", "text": "19. REFUSAL AND FAILURE IN DISTRIBUTION", "sourceParagraph": 784 }, { "type": "paragraph", "text": "A system not responding is also part of the user experience. However, not all refusals are the same.", "sourceParagraphs": [ 785, 786 ] }, { "type": "heading3", "text": "19.1. Appropriate Refusal", "sourceParagraph": 788 }, { "type": "paragraph", "text": "The system may exercise caution when evidence is insufficient or when it cannot provide a safe response. This can be the correct behaviour in some contexts.", "sourceParagraphs": [ 789, 790 ] }, { "type": "heading3", "text": "19.2. Unnecessary Refusal", "sourceParagraph": 792 }, { "type": "paragraph", "text": "It may inconsistently avoid responding to publicly available and low-risk information.", "sourceParagraphs": [ 793 ] }, { "type": "heading3", "text": "19.3. Technical Nonresponse", "sourceParagraph": 795 }, { "type": "paragraph", "text": "The system may not produce a response due to a technical error.", "sourceParagraphs": [ 796 ] }, { "type": "heading3", "text": "19.4. Out-of-Scope Rejection", "sourceParagraph": 798 }, { "type": "paragraph", "text": "The prompt may be incorrect or outside the task protocol.", "sourceParagraphs": [ 799 ] }, { "type": "heading3", "text": "19.5. Preservation of Rejections in Distributions", "sourceParagraph": 801 }, { "type": "paragraph", "text": "Rejections:", "sourceParagraphs": [ 802 ] }, { "type": "bullet", "text": "success,", "sourceParagraph": 803 }, { "type": "bullet", "text": "failure,", "sourceParagraph": 804 }, { "type": "heading3", "text": "UNKNOWN", "sourceParagraph": 805 }, { "type": "paragraph", "text": "should not be forced into a single category. Their function and justification should be evaluated. Removing rejections from the denominator may yield a high accuracy rate only among successful responses and obscure the real user experience.", "sourceParagraphs": [ 806, 807, 808 ] }, { "type": "heading2", "text": "20. HOW CAN DISTRIBUTIONS BE COMPARED?", "sourceParagraph": 810 }, { "type": "paragraph", "text": "Simply comparing the result numbers is not sufficient to compare the distribution of two AI products. For the comparison to be fair, the following areas should be as equal as possible or explicitly modelled:", "sourceParagraphs": [ 811, 812 ] }, { "type": "bullet", "text": "Audited entity", "sourceParagraph": 813 }, { "type": "bullet", "text": "Target population", "sourceParagraph": 814 }, { "type": "bullet", "text": "Prompt version", "sourceParagraph": 815 }, { "type": "bullet", "text": "Languages", "sourceParagraph": 816 }, { "type": "bullet", "text": "Countries", "sourceParagraph": 817 }, { "type": "bullet", "text": "Measurement wave", "sourceParagraph": 818 }, { "type": "bullet", "text": "Session condition", "sourceParagraph": 819 }, { "type": "bullet", "text": "User surface", "sourceParagraph": 820 }, { "type": "bullet", "text": "Plan distribution", "sourceParagraph": 821 }, { "type": "bullet", "text": "Adjudication rule", "sourceParagraph": 822 }, { "type": "bullet", "text": "Reality package", "sourceParagraph": 823 }, { "type": "bullet", "text": "Validity rules", "sourceParagraph": 824 }, { "type": "bullet", "text": "Weights", "sourceParagraph": 825 }, { "type": "paragraph", "text": "Two scores obtained with materially different protocols cannot be used directly for ranking.", "sourceParagraphs": [ 826 ] }, { "type": "heading3", "text": "20.1. Common Support Comparison", "sourceParagraph": 828 }, { "type": "paragraph", "text": "All compared AI products:", "sourceParagraphs": [ 829 ] }, { "type": "bullet", "text": "are officially accessible,", "sourceParagraph": 830 }, { "type": "bullet", "text": "support the same language and user conditions", "sourceParagraph": 831 }, { "type": "paragraph", "text": "can be compared in a common universe. This provides a cleaner comparison between systems.", "sourceParagraphs": [ 832, 833 ] }, { "type": "heading3", "text": "20.2. Natural Access Distribution", "sourceParagraph": 835 }, { "type": "paragraph", "text": "Each AI product can be measured in its own real user access universe. This result shows the product's actual global impact. However, since target populations may differ, direct performance ranking requires caution.", "sourceParagraphs": [ 836, 837, 838 ] }, { "type": "heading3", "text": "20.3. Different Surfaces of the Same Provider", "sourceParagraph": 840 }, { "type": "paragraph", "text": "Web, mobile, enterprise products, or different plans, even if they belong to the same provider, can create separate distributions. If they are to be combined under a single result, the combination and weighting method should be explained.", "sourceParagraphs": [ 841, 842 ] }, { "type": "heading2", "text": "21. WEIGHTING CHANGES THE MEANING OF THE DISTRIBUTION", "sourceParagraph": 844 }, { "type": "paragraph", "text": "A global average depends on the weight with which observations are combined. Possible weights:", "sourceParagraphs": [ 845, 846 ] }, { "type": "bullet", "text": "Equal by country", "sourceParagraph": 847 }, { "type": "bullet", "text": "By country population", "sourceParagraph": 848 }, { "type": "bullet", "text": "Appropriate AI user population", "sourceParagraph": 849 }, { "type": "bullet", "text": "Language user", "sourceParagraph": 850 }, { "type": "bullet", "text": "Actual product exposure", "sourceParagraph": 851 }, { "type": "bullet", "text": "Equal AI product", "sourceParagraph": 852 }, { "type": "bullet", "text": "Market share", "sourceParagraph": 853 }, { "type": "bullet", "text": "Customer target market", "sourceParagraph": 854 }, { "type": "paragraph", "text": "Each method answers a different question.", "sourceParagraphs": [ 855 ] }, { "type": "heading3", "text": "21.1. Equal Country Weight", "sourceParagraph": 857 }, { "type": "paragraph", "text": "San Marino and India can receive the same weight. This approach can address the question of 'Average country experience.' It does not represent the average user experience in the world population.", "sourceParagraphs": [ 858, 859, 860, 861, 862 ] }, { "type": "heading3", "text": "21.2. Population Weight", "sourceParagraph": 864 }, { "type": "paragraph", "text": "Larger countries have more influence. This approach can approach the global population experience. It may render small countries and languages invisible.", "sourceParagraphs": [ 865, 866, 867 ] }, { "type": "heading3", "text": "21.3. Language Fairness Weight", "sourceParagraph": 869 }, { "type": "paragraph", "text": "It can also ensure that low-resource languages remain visible. It is not the same as the global population score.", "sourceParagraphs": [ 870, 871 ] }, { "type": "heading3", "text": "21.4. Product Exposure Weight", "sourceParagraph": 873 }, { "type": "paragraph", "text": "The weighting of AI products according to actual user exposure can be considered. It requires reliable and independent usage data.", "sourceParagraphs": [ 874, 875 ] }, { "type": "heading3", "text": "21.5. Principle of Normative Weighting", "sourceParagraph": 877 }, { "type": "callout", "text": "Weights should be defined before the results are seen, publicly disclosed, and the question they answer should be specified.", "sourceParagraph": 878 }, { "type": "paragraph", "text": "It is prohibited to choose a weighting system that will increase the positive score after the result is seen. Alongside a single global score, alternative meaningful weighted results can also be published.", "sourceParagraphs": [ 879, 880 ] }, { "type": "heading2", "text": "22. SCORE DISTRIBUTION IS LOSSY COMPRESSION", "sourceParagraph": 882 }, { "type": "paragraph", "text": "It is possible to produce a single number from a distribution. However, a single number does not carry all the information in the distribution. Therefore, the score:", "sourceParagraphs": [ 883, 884, 885 ] }, { "type": "heading4", "text": "Is a lossy summary of the distribution", "sourceParagraph": 886 }, { "type": "paragraph", "text": "should be accepted. A score may lose the following information:", "sourceParagraphs": [ 887, 888 ] }, { "type": "bullet", "text": "Weakest language", "sourceParagraph": 889 }, { "type": "bullet", "text": "Weakest country", "sourceParagraph": 890 }, { "type": "bullet", "text": "Critical error type", "sourceParagraph": 891 }, { "type": "bullet", "text": "Unknown mass", "sourceParagraph": 892 }, { "type": "bullet", "text": "Time fluctuation", "sourceParagraph": 893 }, { "type": "bullet", "text": "Polarisation", "sourceParagraph": 894 }, { "type": "bullet", "text": "Product and plan difference", "sourceParagraph": 895 }, { "type": "bullet", "text": "Out-of-scope population", "sourceParagraph": 896 }, { "type": "bullet", "text": "Rejection distribution", "sourceParagraph": 897 }, { "type": "bullet", "text": "Adjudicator ambiguity", "sourceParagraph": 898 }, { "type": "paragraph", "text": "The score is useful because:", "sourceParagraphs": [ 899, 900 ] }, { "type": "bullet", "text": "facilitates comparison,", "sourceParagraph": 901 }, { "type": "bullet", "text": "simplifies public communication,", "sourceParagraph": 902 }, { "type": "bullet", "text": "allows monitoring change,", "sourceParagraph": 903 }, { "type": "paragraph", "text": "supports the management decision. However, the simplicity of the score cannot turn into the elimination of distribution.", "sourceParagraphs": [ 904, 905 ] }, { "type": "heading2", "text": "23. MINIMUM PUBLIC RESULTS CARD", "sourceParagraph": 907 }, { "type": "heading3", "text": "CANDIDATE REQUIREMENT — CANDIDATE OBLIGATION", "sourceParagraph": 908 }, { "type": "paragraph", "text": "If a global score NOMOS is to be presented to the public, it is recommended that at least the following results be shown alongside it:", "sourceParagraphs": [ 909 ] }, { "type": "table", "rows": [ [ "Field", "Description" ], [ "Main representation estimate", "Overall result under defined weighting" ], [ "Confidence or uncertainty range", "Sensitivity of the prediction" ], [ "Scope", "Country, language, product, user surface, and time" ], [ "Weakest sufficient subgroup", "Makes material inequality visible" ], [ "Critical error rate", "Severe error queue" ], [ "Time stability", "Inter-wave variation" ], [ "UNKNOWN and inconclusive rate", "Unknown mass" ], [ "NOT TESTED fields", "Unmeasured universe" ], [ "Measurement date", "Temporal limit of the result" ], [ "Protocol version", "Which rule it was produced with" ] ] }, { "type": "paragraph", "text": "The main score should not be used as a standalone advertising label. The canonical record must be linked to this results card.", "sourceParagraphs": [ 910, 911 ] }, { "type": "heading2", "text": "24. SYNTHETIC APPLE.COM DISTRIBUTION DISPLAY", "sourceParagraph": 913 }, { "type": "paragraph", "text": "SYNTHETIC METHODOLOGY DEMONSTRATION / The AI products, users, languages, numbers, and results below are entirely fictional. They do not represent the actual performance of Apple Inc. or any real AI product. Let us consider two synthetic AI products:", "sourceParagraphs": [ 914, 915 ] }, { "type": "bullet", "text": "Product A", "sourceParagraph": 916 }, { "type": "bullet", "text": "Product B", "sourceParagraph": 917 }, { "type": "paragraph", "text": "For each product:", "sourceParagraphs": [ 918 ] }, { "type": "bullet", "text": "five languages,", "sourceParagraph": 919 }, { "type": "bullet", "text": "200 valid synthetic observations per language,", "sourceParagraph": 920 }, { "type": "bullet", "text": "a total of 1,000 observations", "sourceParagraph": 921 }, { "type": "paragraph", "text": "are assumed. This equal language distribution is only to demonstrate the method. It does not represent the actual GEO-1000 population weighting.", "sourceParagraphs": [ 922, 923, 924 ] }, { "type": "heading3", "text": "24.1. Language-Based pass rates", "sourceParagraph": 926 }, { "type": "table", "rows": [ [ "Language", "Product A", "Product B" ], [ "English", "90%", "99%" ], [ "Turkish", "91%", "98%" ], [ "German", "89%", "97%" ], [ "Japanese", "90%", "86%" ], [ "Arabic", "90%", "70%" ], [ "Simple average", "90%", "90%" ] ] }, { "type": "paragraph", "text": "The average of the two products is the same. However, their distributions are different.", "sourceParagraphs": [ 927, 928 ] }, { "type": "heading4", "text": "Product A", "sourceParagraph": 929 }, { "type": "paragraph", "text": "Language results are close to each other. The difference between the highest and lowest language is 2 points. The language floor is 89%.", "sourceParagraphs": [ 930, 931, 932 ] }, { "type": "heading4", "text": "Product B", "sourceParagraph": 933 }, { "type": "paragraph", "text": "Produces very high results in three languages. There is a decline in Japanese and especially in Arabic. The difference between the highest and lowest language is 29 points. The language floor is 70%. Note: Saying \"Both scored 90\" hides the material difference.", "sourceParagraphs": [ 934, 935, 936, 937, 938, 939, 940 ] }, { "type": "heading3", "text": "24.2. Response Status Distribution", "sourceParagraph": 942 }, { "type": "table", "rows": [ [ "Response status", "Product A", "Product B" ], [ "Meeting the transition criterion", "900", "900" ], [ "Partially supported", "60", "20" ], [ "Unsupported material claim", "20", "20" ], [ "Outdated", "10", "20" ], [ "Critical error candidate", "1", "14" ], [ "UNKNOWN / no result", "9", "26" ], [ "Total", "1,000", "1,000" ] ] }, { "type": "paragraph", "text": "The pass rate of the two products is again 90%. However: In Product B, the number of Critical error candidates is 14 times higher. In Product B, the unknown and inconclusive mass is higher. A larger portion of failures in Product A is due to partial support deficiency. The error backlog of Product B is heavier. The same pass rate does not necessarily result in the same fitness outcome.", "sourceParagraphs": [ 943, 944, 945, 946, 947, 948, 949 ] }, { "type": "heading3", "text": "24.3. Time Waves", "sourceParagraph": 951 }, { "type": "table", "rows": [ [ "Measurement wave", "Product A", "Product B" ], [ "Wave 1", "90%", "98%" ], [ "Wave 2", "91%", "90%" ], [ "Wave 3", "89%", "82%" ], [ "Mean", "90%", "90%" ] ] }, { "type": "paragraph", "text": "Product A is relatively stable. Product B shows a noticeable decline. Publishing only the average of the three waves hides the deterioration in Product B.", "sourceParagraphs": [ 952, 953, 954 ] }, { "type": "heading3", "text": "24.4. Synthetic Result", "sourceParagraph": 956 }, { "type": "paragraph", "text": "It may be possible to publish the same headline score for Product A and Product B:", "sourceParagraphs": [ 957 ] }, { "type": "heading4", "text": "90", "sourceParagraph": 958 }, { "type": "paragraph", "text": "But NOMOS result cards cannot be the same.", "sourceParagraphs": [ 959 ] }, { "type": "heading4", "text": "Product A — Synthetic Distribution Signature", "sourceParagraph": 960 }, { "type": "paragraph", "text": "Overall result: high Language inequality: low Critical error tail: low but not zero Unknown mass: low Time stability: high", "sourceParagraphs": [ 961, 962, 963, 964, 965 ] }, { "type": "heading4", "text": "Product B — Synthetic Distribution Signature", "sourceParagraph": 966 }, { "type": "paragraph", "text": "Overall result: high Language inequality: high Critical error tail: substantial Unknown mass: higher Time stability: low and declining\n\nTherefore, the synthetic conclusion of Section 2 is as follows:", "sourceParagraphs": [ 967, 968, 969, 970, 971, 972 ] }, { "type": "callout", "text": "Having the same average score for Product A and Product B does not indicate that they have the same quality of representation.", "sourceParagraph": 973 }, { "type": "heading2", "text": "25. MANDATORY NORMATIVE PROVISIONS", "sourceParagraph": 975 }, { "type": "protocol_label", "text": "CH02-N01", "sourceParagraph": 976 }, { "type": "paragraph", "text": "A GEO result cannot be defined only by a single average or a single pass rate.", "sourceParagraphs": [ 977 ] }, { "type": "protocol_label", "text": "CH02-N02", "sourceParagraph": 978 }, { "type": "paragraph", "text": "Each distribution result:", "sourceParagraphs": [ 979 ] }, { "type": "bullet", "text": "the audited entity,", "sourceParagraph": 980 }, { "type": "bullet", "text": "target population,", "sourceParagraph": 981 }, { "type": "bullet", "text": "AI product and user surface,", "sourceParagraph": 982 }, { "type": "bullet", "text": "country and language,", "sourceParagraph": 983 }, { "type": "bullet", "text": "The prompt,", "sourceParagraph": 984 }, { "type": "bullet", "text": "the residence condition,", "sourceParagraph": 985 }, { "type": "bullet", "text": "to time", "sourceParagraph": 986 }, { "type": "paragraph", "text": "should be connected.", "sourceParagraphs": [ 987 ] }, { "type": "protocol_label", "text": "CH02-N03", "sourceParagraph": 988 }, { "type": "paragraph", "text": "The global average alone cannot compensate for material subgroup inequality.", "sourceParagraphs": [ 989 ] }, { "type": "protocol_label", "text": "CH02-N04", "sourceParagraph": 990 }, { "type": "paragraph", "text": "Critical and other high-severity error outcomes should be reported separately from the overall error rate.", "sourceParagraphs": [ 991 ] }, { "type": "protocol_label", "text": "CH02-N05", "sourceParagraph": 992 }, { "type": "paragraph", "text": "UNKNOWN, no-result, rejected, invalid, and NOT TESTED statuses should be distinguished from each other.", "sourceParagraphs": [ 993 ] }, { "type": "protocol_label", "text": "CH02-N06", "sourceParagraph": 994 }, { "type": "paragraph", "text": "Unmeasured country, language, product, user surface, or time period cannot be shown as part of the positive distribution.", "sourceParagraphs": [ 995 ] }, { "type": "protocol_label", "text": "CH02-N07", "sourceParagraph": 996 }, { "type": "paragraph", "text": "Weights and aggregation rules should be defined before viewing results.", "sourceParagraphs": [ 997 ] }, { "type": "protocol_label", "text": "CH02-N08", "sourceParagraph": 998 }, { "type": "paragraph", "text": "Post-result weight changes should create a new analysis version, and previous results should be preserved.", "sourceParagraphs": [ 999 ] }, { "type": "protocol_label", "text": "CH02-N09", "sourceParagraph": 1000 }, { "type": "paragraph", "text": "Distributions based on materially different target populations, demand, product, time, or evaluation criteria cannot be directly compared without explanation.", "sourceParagraphs": [ 1001 ] }, { "type": "protocol_label", "text": "CH02-N10", "sourceParagraph": 1002 }, { "type": "paragraph", "text": "Formal language variability cannot automatically be counted as representation error.", "sourceParagraphs": [ 1003 ] }, { "type": "protocol_label", "text": "CH02-N11", "sourceParagraph": 1004 }, { "type": "paragraph", "text": "Material identity, scope, time, competence, or recommendation variability cannot be classified as a style difference without justification.", "sourceParagraphs": [ 1005 ] }, { "type": "protocol_label", "text": "CH02-N12", "sourceParagraph": 1006 }, { "type": "paragraph", "text": "High stability alone is not evidence of quality; consistently wrong results should also be recorded.", "sourceParagraphs": [ 1007 ] }, { "type": "protocol_label", "text": "CH02-N13", "sourceParagraph": 1008 }, { "type": "paragraph", "text": "A high average, if accompanied by low coverage, should have public results with coverage limitation.", "sourceParagraphs": [ 1009 ] }, { "type": "protocol_label", "text": "CH02-N14", "sourceParagraph": 1010 }, { "type": "paragraph", "text": "Subgroup scores should not be published as definitive country or language judgements under insufficient sampling and broad uncertainty.", "sourceParagraphs": [ 1011 ] }, { "type": "protocol_label", "text": "CH02-N15", "sourceParagraph": 1012 }, { "type": "paragraph", "text": "Rejection and technical failure rates should be preserved as part of the genuine user experience.", "sourceParagraphs": [ 1013 ] }, { "type": "protocol_label", "text": "CH02-N16", "sourceParagraph": 1014 }, { "type": "paragraph", "text": "The only score made public should be linked to the canonical distribution and scope record.", "sourceParagraphs": [ 1015 ] }, { "type": "protocol_label", "text": "CH02-N17", "sourceParagraph": 1016 }, { "type": "paragraph", "text": "Synthetic distributions cannot be combined with the performance of a real AI product or real entity.", "sourceParagraphs": [ 1017 ] }, { "type": "protocol_label", "text": "CH02-N18", "sourceParagraph": 1018 }, { "type": "paragraph", "text": "The distribution report should have an accountable human or institutional owner.", "sourceParagraphs": [ 1019 ] }, { "type": "heading2", "text": "26. FORMS OF FAILURE", "sourceParagraph": 1021 }, { "type": "protocol_label", "text": "CH02-F01 — AVERAGE THINKING", "sourceParagraph": 1022 }, { "type": "paragraph", "text": "High global average is shown; language, country, or Critical error corruption is stored.", "sourceParagraphs": [ 1023 ] }, { "type": "protocol_label", "text": "CH02-F02 — DELETE SUBGROUP", "sourceParagraph": 1024 }, { "type": "paragraph", "text": "Low-performing groups are removed from the report or declared 'insignificant'.", "sourceParagraphs": [ 1025 ] }, { "type": "protocol_label", "text": "CH02-F03 — SCOPE MAP", "sourceParagraph": 1026 }, { "type": "paragraph", "text": "Narrow-scope good results are used for the whole world, all languages, or all products.", "sourceParagraphs": [ 1027 ] }, { "type": "protocol_label", "text": "CH02-F04 — SEVERITY NORMALISATION", "sourceParagraph": 1028 }, { "type": "paragraph", "text": "Minor deficiencies and serious health or licensing errors are combined within the same error count.", "sourceParagraphs": [ 1029 ] }, { "type": "protocol_label", "text": "CH02-F05 — HIDE DETERIORATION WITH TIME AVERAGE", "sourceParagraph": 1030 }, { "type": "paragraph", "text": "Decline or fluctuation is made invisible only with the period average.", "sourceParagraphs": [ 1031 ] }, { "type": "protocol_label", "text": "CH02-F06 — WORSHIP OF STABILITY", "sourceParagraph": 1032 }, { "type": "paragraph", "text": "A consistent answer is assumed to be automatically correct.", "sourceParagraphs": [ 1033 ] }, { "type": "protocol_label", "text": "CH02-F07 — WORD-FOR-WORD SAME ANSWER FALLACY", "sourceParagraph": 1034 }, { "type": "paragraph", "text": "Literal text matching is used as a measure of consistency instead of semantic correctness.", "sourceParagraphs": [ 1035 ] }, { "type": "protocol_label", "text": "CH02-F08 — COUNTING CONTEXTUAL DIFFERENCE AS ERROR", "sourceParagraph": 1036 }, { "type": "paragraph", "text": "Differences in appropriate recommendations based on actual user needs are marked as arbitrary variability.", "sourceParagraphs": [ 1037 ] }, { "type": "protocol_label", "text": "CH02-F09 — COUNTING ARBITRARY VARIABILITY AS PERSONALISATION", "sourceParagraph": 1038 }, { "type": "paragraph", "text": "Material contradictions under equivalent user conditions are defended as “personalisation” without explanation.", "sourceParagraphs": [ 1039 ] }, { "type": "protocol_label", "text": "CH02-F10 — CHANGING THE DENOMINATOR", "sourceParagraph": 1040 }, { "type": "paragraph", "text": "Post-result transition is made among completed, valid, or evaluated observation denominators to elevate the positive outcome.", "sourceParagraphs": [ 1041 ] }, { "type": "protocol_label", "text": "CH02-F11 — POST-RESULT WEIGHTING", "sourceParagraph": 1042 }, { "type": "paragraph", "text": "The country, language, or model coefficients producing the highest score are selected afterward.", "sourceParagraphs": [ 1043 ] }, { "type": "protocol_label", "text": "CH02-F12 — MERGING INCONSISTENT CELLS", "sourceParagraph": 1044 }, { "type": "paragraph", "text": "Different prompt, product, time, or panel conditions are presented as a single distribution.", "sourceParagraphs": [ 1045 ] }, { "type": "protocol_label", "text": "CH02-F13 — DELETING REJECTIONS", "sourceParagraph": 1046 }, { "type": "paragraph", "text": "System records that did not respond are removed to generate high accuracy only among those that produced a response.", "sourceParagraphs": [ 1047 ] }, { "type": "protocol_label", "text": "CH02-F14 — DEFINITE JUDGEMENT FROM SMALL CELL", "sourceParagraph": 1048 }, { "type": "paragraph", "text": "A country or language score is produced from one or several users.", "sourceParagraphs": [ 1049 ] }, { "type": "protocol_label", "text": "CH02-F15 — HIDING THE CRITICAL TAIL", "sourceParagraph": 1050 }, { "type": "paragraph", "text": "Rare severe errors are lost within the overall pass rate.", "sourceParagraphs": [ 1051 ] }, { "type": "protocol_label", "text": "CH02-F16 — SINGLE SCORE REPORT", "sourceParagraph": 1052 }, { "type": "paragraph", "text": "Only the headline score is published without showing scope, inequality, stability, uncertainty, and severe errors.", "sourceParagraphs": [ 1053 ] }, { "type": "protocol_label", "text": "CH02-F17 — COUNTING DISTRIBUTION AS \"AI'S BELIEF\"", "sourceParagraph": 1054 }, { "type": "paragraph", "text": "Observed user surface results are presented as if they are the invariant and intrinsic viewpoint of artificial intelligence.", "sourceParagraphs": [ 1055 ] }, { "type": "protocol_label", "text": "CH02-F18 — COUNTING THE UNMEASURED AS ZERO VARIANCE", "sourceParagraph": 1056 }, { "type": "paragraph", "text": "Untested areas are considered smooth and stable.", "sourceParagraphs": [ 1057 ] }, { "type": "heading2", "text": "27. AUDIT PROCEDURE", "sourceParagraph": 1059 }, { "type": "paragraph", "text": "Whether a GEO report honestly represents the distribution should be examined with the following steps.", "sourceParagraphs": [ 1060 ] }, { "type": "heading3", "text": "Step 1 — Record the Public Claim", "sourceParagraph": 1062 }, { "type": "paragraph", "text": "Example: “The brand's global NOMOS score is 92.” The full sentence and the surface where it is found are recorded.", "sourceParagraphs": [ 1063, 1064, 1065 ] }, { "type": "heading3", "text": "Step 2 — Define Target Distribution", "sourceParagraph": 1067 }, { "type": "paragraph", "text": "For which score:", "sourceParagraphs": [ 1068 ] }, { "type": "bullet", "text": "user universe,", "sourceParagraph": 1069 }, { "type": "bullet", "text": "AI products,", "sourceParagraph": 1070 }, { "type": "bullet", "text": "countries,", "sourceParagraph": 1071 }, { "type": "bullet", "text": "languages,", "sourceParagraph": 1072 }, { "type": "bullet", "text": "time period", "sourceParagraph": 1073 }, { "type": "paragraph", "text": "is it applied?", "sourceParagraphs": [ 1074 ] }, { "type": "heading3", "text": "Step 3 — Extract Measurement Cells", "sourceParagraph": 1076 }, { "type": "paragraph", "text": "Each material combination is determined:", "sourceParagraphs": [ 1077 ] }, { "type": "bullet", "text": "Product", "sourceParagraph": 1078 }, { "type": "bullet", "text": "Country", "sourceParagraph": 1079 }, { "type": "bullet", "text": "Language", "sourceParagraph": 1080 }, { "type": "bullet", "text": "Panel", "sourceParagraph": 1081 }, { "type": "bullet", "text": "Prompt", "sourceParagraph": 1082 }, { "type": "bullet", "text": "Wave", "sourceParagraph": 1083 }, { "type": "bullet", "text": "User surface", "sourceParagraph": 1084 }, { "type": "heading3", "text": "Step 4 — Rebuild Denominators", "sourceParagraph": 1086 }, { "type": "paragraph", "text": "Assigned tasks, completed tasks, valid records, adjudicated records and passed responses must be reported as separate counts.", "sourceParagraphs": [ 1087, 1088, 1089, 1090, 1091, 1092 ] }, { "type": "heading3", "text": "Step 5 — Extract Result Statuses", "sourceParagraph": 1094 }, { "type": "paragraph", "text": "Not only the pass rate:", "sourceParagraphs": [ 1095 ] }, { "type": "bullet", "text": "partial,", "sourceParagraph": 1096 }, { "type": "bullet", "text": "unsupported,", "sourceParagraph": 1097 }, { "type": "bullet", "text": "conflicting,", "sourceParagraph": 1098 }, { "type": "bullet", "text": "old,", "sourceParagraph": 1099 }, { "type": "bullet", "text": "Critical,", "sourceParagraph": 1100 }, { "type": "heading3", "text": "UNKNOWN,", "sourceParagraph": 1101 }, { "type": "paragraph", "text": "rejected, inconclusive distribution is examined.", "sourceParagraphs": [ 1102, 1103, 1104 ] }, { "type": "heading3", "text": "Step 6 — Verify Weights", "sourceParagraph": 1106 }, { "type": "paragraph", "text": "Weights:", "sourceParagraphs": [ 1107 ] }, { "type": "bullet", "text": "are they predefined,", "sourceParagraph": 1108 }, { "type": "bullet", "text": "which question do they answer,", "sourceParagraph": 1109 }, { "type": "paragraph", "text": "have they changed after the result?", "sourceParagraphs": [ 1110 ] }, { "type": "heading3", "text": "Step 7 — Examine Subgroup Differences", "sourceParagraph": 1112 }, { "type": "paragraph", "text": "The best group, the weakest sufficient group, language difference, country difference, product difference, and panel difference are evaluated.", "sourceParagraphs": [ 1113, 1114, 1115, 1116, 1117, 1118, 1119 ] }, { "type": "heading3", "text": "Step 8 — Review Scope and Deficiency", "sourceParagraph": 1121 }, { "type": "paragraph", "text": "Which areas were not measured? Which cells are insufficient? What are the reasons for the deficiency? Could the missing data be systematic?", "sourceParagraphs": [ 1122, 1123, 1124, 1125 ] }, { "type": "heading3", "text": "Step 9 — Examine the Determination", "sourceParagraph": 1127 }, { "type": "paragraph", "text": "The change between Waves Sessions Users Interfaces is evaluated.", "sourceParagraphs": [ 1128, 1129, 1130, 1131, 1132 ] }, { "type": "heading3", "text": "Step 10 — Examine Queue Risk", "sourceParagraph": 1134 }, { "type": "paragraph", "text": "Critical and other serious errors:", "sourceParagraphs": [ 1135 ] }, { "type": "bullet", "text": "number,", "sourceParagraph": 1136 }, { "type": "bullet", "text": "its rate,", "sourceParagraph": 1137 }, { "type": "bullet", "text": "type,", "sourceParagraph": 1138 }, { "type": "bullet", "text": "user effect", "sourceParagraph": 1139 }, { "type": "paragraph", "text": "is evaluated separately in terms of", "sourceParagraphs": [ 1140 ] }, { "type": "heading3", "text": "Step 11 — Compare the Public Summary with the Distribution", "sourceParagraph": 1142 }, { "type": "paragraph", "text": "Does the single number or sentence published to the public accurately convey the direction and risk of the full distribution?", "sourceParagraphs": [ 1143 ] }, { "type": "heading3", "text": "Step 12 — Limit the Result", "sourceParagraph": 1145 }, { "type": "paragraph", "text": "The public claim is limited only to the level carried by the distribution and scope.", "sourceParagraphs": [ 1146 ] }, { "type": "heading2", "text": "28. NECESSARY EVIDENCE", "sourceParagraph": 1148 }, { "type": "paragraph", "text": "To the extent relevant for a representative distribution or a global GEO score, the following records should be searched:", "sourceParagraphs": [ 1149 ] }, { "type": "bullet", "text": "Audited entity record", "sourceParagraph": 1150 }, { "type": "bullet", "text": "Definition of target population", "sourceParagraph": 1151 }, { "type": "bullet", "text": "Sampling frame", "sourceParagraph": 1152 }, { "type": "bullet", "text": "AI product and user interfaces", "sourceParagraph": 1153 }, { "type": "bullet", "text": "Country and language cells", "sourceParagraph": 1154 }, { "type": "bullet", "text": "Prompt and versions", "sourceParagraph": 1155 }, { "type": "bullet", "text": "Session and panel conditions", "sourceParagraph": 1156 }, { "type": "bullet", "text": "Measurement waves", "sourceParagraph": 1157 }, { "type": "bullet", "text": "Assigned task numbers", "sourceParagraph": 1158 }, { "type": "bullet", "text": "Completed tasks", "sourceParagraph": 1159 }, { "type": "bullet", "text": "Valid and invalid records", "sourceParagraph": 1160 }, { "type": "bullet", "text": "Raw responses", "sourceParagraph": 1161 }, { "type": "bullet", "text": "Result status distribution", "sourceParagraph": 1162 }, { "type": "bullet", "text": "Adjudication records", "sourceParagraph": 1163 }, { "type": "bullet", "text": "Critical error records", "sourceParagraph": 1164 }, { "type": "bullet", "text": "Rejection and inconclusive records", "sourceParagraph": 1165 }, { "type": "bullet", "text": "UNKNOWN records", "sourceParagraph": 1166 }, { "type": "bullet", "text": "Weights", "sourceParagraph": 1167 }, { "type": "bullet", "text": "Stakeholders", "sourceParagraph": 1168 }, { "type": "bullet", "text": "Subgroup results", "sourceParagraph": 1169 }, { "type": "bullet", "text": "Stability results", "sourceParagraph": 1170 }, { "type": "bullet", "text": "Out-of-scope areas", "sourceParagraph": 1171 }, { "type": "bullet", "text": "Confidence or uncertainty calculations", "sourceParagraph": 1172 }, { "type": "bullet", "text": "Public scorecard", "sourceParagraph": 1173 }, { "type": "bullet", "text": "Responsible person or institution", "sourceParagraph": 1174 }, { "type": "bullet", "text": "Protocol and analysis version", "sourceParagraph": 1175 }, { "type": "bullet", "text": "Change history", "sourceParagraph": 1176 }, { "type": "heading2", "text": "29. AUDIT CHECKLIST", "sourceParagraph": 1178 }, { "type": "paragraph", "text": "Is the distribution's target population clear? Are the audited AI products and user surfaces specified? Are the country, language, prompt and time scopes visible? Are the weights used to produce the composite score disclosed? Is the denominator fully defined? Are invalid records reported separately? Are refusals and inconclusive outcomes preserved? Are UNKNOWN and NOT TESTED distinct? Are subgroup results visible? Have small cells been interpreted with excessive certainty? Has the weakest material group been retained? Are Critical errors reported separately? Are measurement waves shown separately? Does the average conceal decline or volatility? Is a merely formal difference distinguished from a material difference in meaning? Has context-sensitive variation in recommendations been classified correctly? Have different surfaces from the same provider been merged without explanation?", "sourceParagraphs": [ 1179, 1180, 1181, 1182, 1183, 1184, 1185, 1186, 1187, 1188, 1189, 1190, 1191, 1192, 1193, 1194, 1195 ] }, { "type": "paragraph", "text": "Have areas outside the scope been included in the global result? Have synthetic and real distributions been separated? Does the sentence presented to the public reflect the true direction of the full distribution? If one of the substantive questions is unanswered, the distribution result carries limited confidence. If more than one is unanswered, a single global score should not be presented as a definitive judgement to the public.", "sourceParagraphs": [ 1196, 1197, 1198, 1199, 1200 ] }, { "type": "heading2", "text": "30. OBJECTIONS AND ANSWERS", "sourceParagraph": 1202 }, { "type": "heading3", "text": "Objection 1 — “The public wants a single number.”", "sourceParagraph": 1203 }, { "type": "paragraph", "text": "This is correct. A single number can facilitate communication. NOMOS does not completely reject a single number. However, a single number:", "sourceParagraphs": [ 1204, 1205, 1206, 1207 ] }, { "type": "bullet", "text": "scope,", "sourceParagraph": 1208 }, { "type": "bullet", "text": "Critical error,", "sourceParagraph": 1209 }, { "type": "bullet", "text": "subgroup base,", "sourceParagraph": 1210 }, { "type": "bullet", "text": "time,", "sourceParagraph": 1211 }, { "type": "bullet", "text": "uncertainty", "sourceParagraph": 1212 }, { "type": "paragraph", "text": "must be connected to the information. A number can be short. It cannot shorten the truth.", "sourceParagraphs": [ 1213, 1214, 1215 ] }, { "type": "heading3", "text": "Objection 2 — 'Small languages unnecessarily lower the global score.'", "sourceParagraph": 1217 }, { "type": "paragraph", "text": "The global population weight can account for a small language with limited weight. However, heavy representation distortion in a small language cannot be left unnoticed simply because its population share is low. Therefore:", "sourceParagraphs": [ 1218, 1219, 1220 ] }, { "type": "bullet", "text": "population-weighted main score,", "sourceParagraph": 1221 }, { "type": "bullet", "text": "as a result of separate language fairness", "sourceParagraph": 1222 }, { "type": "paragraph", "text": "They can be published together. Two metrics answer different questions.", "sourceParagraphs": [ 1223, 1224 ] }, { "type": "heading3", "text": "Objection 3 — 'If there is personalisation, different responses are normal.'", "sourceParagraph": 1226 }, { "type": "paragraph", "text": "Some differences can be normal and correct. However, customisation:", "sourceParagraphs": [ 1227, 1228 ] }, { "type": "bullet", "text": "company identity,", "sourceParagraph": 1229 }, { "type": "bullet", "text": "licence,", "sourceParagraph": 1230 }, { "type": "bullet", "text": "its current price,", "sourceParagraph": 1231 }, { "type": "bullet", "text": "the real service area", "sourceParagraph": 1232 }, { "type": "paragraph", "text": "should not be changed arbitrarily. Personalisation can affect the user's appropriate explanation and advice. It is not permission to invent a material fact.", "sourceParagraphs": [ 1233, 1234, 1235 ] }, { "type": "heading3", "text": "Objection 4 — “Why would it be a problem if the same answer is not given?”", "sourceParagraph": 1237 }, { "type": "paragraph", "text": "The problem is not word difference. The problem is the change of material reality. NOMOS does not look for exact text equality, but conditional semantic consistency.", "sourceParagraphs": [ 1238, 1239, 1240 ] }, { "type": "heading3", "text": "Objection 5 — “Why would it matter if a Critical error occurs only once in 1,000 answers?”", "sourceParagraph": 1242 }, { "type": "paragraph", "text": "This error may be rare in terms of population prevalence. However, it can be severe in terms of damage. The uncontrolled loss of a critical system on an airplane once in a thousand flights is not assessed as only 0.1 per cent. In the context of GEO, frequency and damage should also be examined separately.", "sourceParagraphs": [ 1243, 1244, 1245, 1246 ] }, { "type": "heading3", "text": "Objection 6 — “We cannot find a sufficient sample for every small country or language.”", "sourceParagraph": 1248 }, { "type": "paragraph", "text": "This is true. Therefore:", "sourceParagraphs": [ 1249, 1250 ] }, { "type": "bullet", "text": "global population panel,", "sourceParagraph": 1251 }, { "type": "bullet", "text": "Country Observer panel,", "sourceParagraph": 1252 }, { "type": "bullet", "text": "Language Fairness Panel", "sourceParagraph": 1253 }, { "type": "paragraph", "text": "can be established separately. A single-person observation does not turn into a country score. It only indicates that an observation exists. The correct status for an insufficient sample:", "sourceParagraphs": [ 1254, 1255, 1256, 1257 ] }, { "type": "heading3", "text": "INSUFFICIENT SAMPLE", "sourceParagraph": 1258 }, { "type": "paragraph", "text": "or:", "sourceParagraphs": [ 1259 ] }, { "type": "heading3", "text": "NOT ESTIMATED", "sourceParagraph": 1260 }, { "type": "paragraph", "text": "must be.", "sourceParagraphs": [ 1261 ] }, { "type": "heading3", "text": "Objection 7 — \"This much detail is too complex for the client to understand.\"", "sourceParagraph": 1263 }, { "type": "paragraph", "text": "It is not necessary for the client to see all the formulas on the main page. However, the key facts necessary for the decision cannot be hidden. The short results card should answer the following questions: What is the overall result? What is the most serious risk? What is the weakest area? What has not been tested? How robust is the result? What date does it belong to? Complexity needs to be managed. Not eliminated.", "sourceParagraphs": [ 1264, 1265, 1266, 1267, 1268, 1269, 1270, 1271, 1272, 1273, 1274 ] }, { "type": "heading3", "text": "Objection 8 — \"If the same model is used in all countries, why would there be a country difference?\"", "sourceParagraph": 1276 }, { "type": "paragraph", "text": "The output seen by the user may not consist solely of the bare model. Country differences include:", "sourceParagraphs": [ 1277, 1278 ] }, { "type": "bullet", "text": "product access,", "sourceParagraph": 1279 }, { "type": "bullet", "text": "web resources,", "sourceParagraph": 1280 }, { "type": "bullet", "text": "locale,", "sourceParagraph": 1281 }, { "type": "bullet", "text": "language,", "sourceParagraph": 1282 }, { "type": "bullet", "text": "plan,", "sourceParagraph": 1283 }, { "type": "bullet", "text": "interface,", "sourceParagraph": 1284 }, { "type": "bullet", "text": "legislation,", "sourceParagraph": 1285 }, { "type": "bullet", "text": "personalisation", "sourceParagraph": 1286 }, { "type": "paragraph", "text": "It can emerge through it. A difference may not occur. However, its absence cannot be assumed without measurement.", "sourceParagraphs": [ 1287, 1288, 1289 ] }, { "type": "heading3", "text": "Objection 9 — “Reliable scoring is not possible since AI responses are variable.”", "sourceParagraph": 1291 }, { "type": "paragraph", "text": "Variability does not make measurement impossible. On the contrary, it is one of the features that needs to be measured. The score:", "sourceParagraphs": [ 1292, 1293, 1294 ] }, { "type": "bullet", "text": "specific date,", "sourceParagraph": 1295 }, { "type": "bullet", "text": "scope,", "sourceParagraph": 1296 }, { "type": "bullet", "text": "confidence interval,", "sourceParagraph": 1297 }, { "type": "bullet", "text": "stability information", "sourceParagraph": 1298 }, { "type": "paragraph", "text": "can be meaningful as long as it carries it. The claim of timeless certainty, however, is indefensible.", "sourceParagraphs": [ 1299, 1300 ] }, { "type": "heading3", "text": "Objection 10 — “If the prompt changes, the distribution completely changes.”", "sourceParagraph": 1302 }, { "type": "paragraph", "text": "That is correct. Therefore, the distribution is conditional on the prompt. GEO-1000 does not produce the single answer for all possible questions. Predefined:", "sourceParagraphs": [ 1303, 1304, 1305, 1306 ] }, { "type": "bullet", "text": "Core Mirror Prompt,", "sourceParagraph": 1307 }, { "type": "bullet", "text": "Diagnostic Prompt Registry", "sourceParagraph": 1308 }, { "type": "paragraph", "text": "Measures specific representational domains. the prompt constitution will be defined in Section 10.", "sourceParagraphs": [ 1309, 1310 ] }, { "type": "heading2", "text": "COMMON RULE OF SECTION 32", "sourceParagraph": 1346 }, { "type": "paragraph", "text": "This section was not written against a single score. It was written against assuming a single score alone represents the entirety of reality. People want numbers. Institutions want rankings. Boards want to see a single line. The market may ask: “What is your NOMOS score?” This question is legitimate. But the answer cannot be only: “90.” Because behind 90, any of the following realities may exist:", "sourceParagraphs": [ 1347, 1348, 1349, 1350, 1351, 1352, 1353, 1354, 1355, 1356, 1357, 1358, 1359 ] }, { "type": "bullet", "text": "Consistent accuracy across all languages", "sourceParagraph": 1360 }, { "type": "bullet", "text": "99 in some languages, 45 in others", "sourceParagraph": 1361 }, { "type": "bullet", "text": "High average but severe Critical error", "sourceParagraph": 1362 }, { "type": "bullet", "text": "High accuracy but low coverage", "sourceParagraph": 1363 }, { "type": "bullet", "text": "High instant results but rapid decline", "sourceParagraph": 1364 }, { "type": "bullet", "text": "Success only in easy prompts", "sourceParagraph": 1365 }, { "type": "bullet", "text": "High rate only among assessable answers", "sourceParagraph": 1366 }, { "type": "bullet", "text": "Assuming untested countries positively", "sourceParagraph": 1367 }, { "type": "bullet", "text": "Combining selected AI products", "sourceParagraph": 1368 }, { "type": "paragraph", "text": "The value of a score is formed by the visibility of the underlying distribution. If the distribution is not visible, the score does not generate confidence. It only produces an appearance of certainty. Therefore, NOMOS's second measurement law is as follows:", "sourceParagraphs": [ 1369, 1370, 1371, 1372 ] }, { "type": "callout", "text": "a GEO score can be a summary of the distribution. / It cannot replace the distribution.", "sourceParagraph": 1373 }, { "type": "paragraph", "text": "Its third law is:", "sourceParagraphs": [ 1374 ] }, { "type": "callout", "text": "A high average does not eliminate the question of who is poorly represented.", "sourceParagraph": 1375 }, { "type": "paragraph", "text": "The fourth law is as follows:", "sourceParagraphs": [ 1376 ] }, { "type": "callout", "text": "A rare serious mistake is not insignificant just because it is rare.", "sourceParagraph": 1377 }, { "type": "paragraph", "text": "The fifth law is as follows:", "sourceParagraphs": [ 1378 ] }, { "type": "callout", "text": "The untested field is not the positive side of the distribution; it is the unknown boundary.", "sourceParagraph": 1379 }, { "type": "heading2", "text": "Order 2 of NOMOS", "sourceParagraph": 1381 }, { "type": "callout", "text": "Don't just show me the average.", "sourceParagraph": 1382 }, { "type": "callout", "text": "Show who is below the average.", "sourceParagraph": 1383 }, { "type": "callout", "text": "Do not hide your worst language while showing your best language.", "sourceParagraph": 1384 }, { "type": "callout", "text": "Do not use the weight of big countries to erase the existence of small countries.", "sourceParagraph": 1385 }, { "type": "callout", "text": "Do not send one person to a small country and then announce a definitive score for that country.", "sourceParagraph": 1386 }, { "type": "callout", "text": "Do not trivialise one severe error merely because it is hidden within a response that is 99 per cent correct.", "sourceParagraph": 1387 }, { "type": "callout", "text": "Do not declare an error merely because the wording of an answer has changed. / But when the underlying reality changes, do not dismiss it as a mere stylistic difference.", "sourceParagraph": 1388 }, { "type": "callout", "text": "Do not throw rejections, unknowns, and untested ones outside the distribution.", "sourceParagraph": 1389 }, { "type": "callout", "text": "Do not change the weight, share, or scope after seeing the result.", "sourceParagraph": 1390 }, { "type": "callout", "text": "Generate a number. / But do not make the countries, languages, users, times, and errors behind the number invisible.", "sourceParagraph": 1391 }, { "type": "callout", "text": "Because the representation of an entity is not an average sentence. / It is a distribution of the different realities encountered by different people.", "sourceParagraph": 1392 }, { "type": "heading2", "text": "The Chapter's Closing Sentence", "sourceParagraph": 1394 }, { "type": "callout", "text": "The measurement object of GEO is not a single answer or a single average score of an artificial intelligence; it is the entirety of representative results generated in a defined user population under defined conditions.", "sourceParagraph": 1395 }, { "type": "heading2", "text": "Normative Core" }, { "type": "normative", "text": "An entity's generative-system representation MUST be treated as a conditional distribution of observable outcomes across defined: - AI products and user surfaces, - users or sampling units, - countries, - languages and locales, - session and personalisation states, - prompts, - and measurement times. A single aggregate score MUST NOT replace disclosure of: - scope and coverage, - outcome distribution, - subgroup variation, - critical-error incidence, - unknown and untested mass, - temporal stability, - weighting rules, - and uncertainty. Equal averages MUST NOT be interpreted as equivalent representation quality when their subgroup, severity, coverage, or stability distributions materially differ. Untested groups MUST NOT be imputed as conforming, and rare high-harm outcomes MUST remain separately visible from average performance.", "translationRoute": "controlled-normative-rule" } ] }, { "number": 3, "id": "NOMOS-GEO-AUDIT-CH03", "title": "What Exactly Is the Entity Under Audit?", "subtitle": null, "sourceFile": "3.cü bölüm.docx", "sourceSha256": "60D2981CEC474DAAAE04718FF663900DB0740A3C91147CDDFD68C18BDDD13695", "sourceWordCount": 6476, "sourceIds": [ "K10", "K11" ], "blocks": [ { "type": "heading2", "text": "Chapter Boundary", "sourceParagraph": 16 }, { "type": "paragraph", "text": "Section 1 established the following provision:", "sourceParagraphs": [ 17 ] }, { "type": "callout", "text": "A single response is not a GEO score.", "sourceParagraph": 18 }, { "type": "paragraph", "text": "Section 2 added the following provision:", "sourceParagraphs": [ 19 ] }, { "type": "callout", "text": "The representation of an entity in generative systems is a distribution that occurs among users, AI products, countries, languages, sessions, and times.", "sourceParagraph": 20 }, { "type": "paragraph", "text": "This section answers the next mandatory question:", "sourceParagraphs": [ 21 ] }, { "type": "callout", "text": "Whose or what representation does this distribution exactly measure?", "sourceParagraph": 22 }, { "type": "paragraph", "text": "A company? A brand? A domain name? A product? A service? A founder? A group of companies? A local legal entity in a specific country? A franchise business? When an artificial intelligence system sees only the apple.com domain:", "sourceParagraphs": [ 23, 24, 25, 26, 27, 28, 29, 30, 31, 32 ] }, { "type": "bullet", "text": "the domain name,", "sourceParagraph": 33 }, { "type": "bullet", "text": "the brand,", "sourceParagraph": 34 }, { "type": "bullet", "text": "the parent company,", "sourceParagraph": 35 }, { "type": "bullet", "text": "the local sales entity,", "sourceParagraph": 36 }, { "type": "bullet", "text": "the product family,", "sourceParagraph": 37 }, { "type": "bullet", "text": "the store service", "sourceParagraph": 38 }, { "type": "paragraph", "text": "Will the system treat all of them as the same entity? Unless the distinction is made before the audit begins, a result may be calculated correctly yet applied to the wrong object. This chapter:", "sourceParagraphs": [ 39, 40, 41 ] }, { "type": "bullet", "text": "the audited entity,", "sourceParagraph": 42 }, { "type": "bullet", "text": "the identity anchor used in the prompt,", "sourceParagraph": 43 }, { "type": "bullet", "text": "the entity analysed by the system,", "sourceParagraph": 44 }, { "type": "bullet", "text": "the unit to which the conformity decision will be applied,", "sourceParagraph": 45 }, { "type": "bullet", "text": "relationships between entities,", "sourceParagraph": 46 }, { "type": "bullet", "text": "which entity the attributes belong to,", "sourceParagraph": 47 }, { "type": "bullet", "text": "temporal and geographic identity boundaries", "sourceParagraph": 48 }, { "type": "paragraph", "text": "defines. This chapter does not yet:", "sourceParagraphs": [ 49, 50 ] }, { "type": "bullet", "text": "how all claims about the entity will be verified,", "sourceParagraph": 51 }, { "type": "bullet", "text": "how adjudicators will score atomic claims,", "sourceParagraph": 52 }, { "type": "bullet", "text": "country and language sample quotas,", "sourceParagraph": 53 }, { "type": "bullet", "text": "the final NOMOS score formula", "sourceParagraph": 54 }, { "type": "paragraph", "text": "is not finalised. The task of Section 3 is more fundamental:", "sourceParagraphs": [ 55, 56 ] }, { "type": "callout", "text": "To lock the identity and boundaries of the object being measured before the measurement begins.", "sourceParagraph": 57 }, { "type": "heading2", "text": "NOMOS Challenge", "sourceParagraph": 59 }, { "type": "paragraph", "text": "You are asking me this question: “What kind of company is Apple.com?” There can be at least three different objects in the question:", "sourceParagraphs": [ 60, 61, 62 ] }, { "type": "bullet", "text": "the apple.com domain name", "sourceParagraph": 63 }, { "type": "bullet", "text": "the Apple brand", "sourceParagraph": 64 }, { "type": "bullet", "text": "the legal company associated with the domain name and the brand", "sourceParagraph": 65 }, { "type": "paragraph", "text": "I would choose one of these. Maybe the correct one. Maybe the wrong one. Maybe I combine more than one in a single answer. Then you would score the answer. But you did not tell me in advance which entity you are scoring. I may have:", "sourceParagraphs": [ 66, 67, 68, 69, 70, 71, 72 ] }, { "type": "bullet", "text": "described the domain as an online store,", "sourceParagraph": 73 }, { "type": "bullet", "text": "described the brand as a consumer electronics brand,", "sourceParagraph": 74 }, { "type": "bullet", "text": "described the legal company as a global corporation,", "sourceParagraph": 75 }, { "type": "paragraph", "text": "I may have listed the products as if the company itself were presenting them. My answer may appear fluent. But if the audit object is unclear, whether it is correct or not also remains uncertain. Now, consider a more complex example.", "sourceParagraphs": [ 76, 77, 78, 79 ] }, { "type": "paragraph", "text": "A brand is operated by the parent company in one country; by a subsidiary in another country; and by an independent distributor in a third country. A user asks: \"Does this brand provide services in Turkey?\" Which entity should answer? The brand's global site? The local company in Turkey? The distributor?", "sourceParagraphs": [ 80, 81, 82, 83, 84, 85, 86 ] }, { "type": "paragraph", "text": "The franchise operator? Does the licence owned by the global company automatically transfer to the local franchise? Do customer reviews of the local operation prove the global quality of the main brand? Can the number of employees of the parent company be used as the capacity of the local company?", "sourceParagraphs": [ 87, 88, 89, 90 ] }, { "type": "paragraph", "text": "Can a product's award be transferred to all the companies that sell it? Does a founder's personal experience become the corporate age of the company? Is the previous owner and the new owner of a domain name considered the same entity?", "sourceParagraphs": [ 91, 92, 93 ] }, { "type": "paragraph", "text": "When a company acquires another company, do all the achievements of the previous company become a characteristic of the new company? GEO score can be generated without answering these questions. However, it is not known what the score represents. The first ruling of this section is as follows:", "sourceParagraphs": [ 94, 95, 96, 97 ] }, { "type": "callout", "text": "An entity cannot be measured until it is resolved.", "sourceParagraph": 98 }, { "type": "paragraph", "text": "Its second provision states:", "sourceParagraphs": [ 99 ] }, { "type": "callout", "text": "Sharing the same name is not being the same entity.", "sourceParagraph": 100 }, { "type": "paragraph", "text": "Its third provision states:", "sourceParagraphs": [ 101 ] }, { "type": "callout", "text": "A relationship does not automatically allow properties to be transferred between the parties.", "sourceParagraph": 102 }, { "type": "heading2", "text": "1. PURPOSE OF THE CHAPTER", "sourceParagraph": 104 }, { "type": "paragraph", "text": "The purpose of this section is to determine the identity, boundary, and relationships of the entity measured within GEO-1000 before the results are seen. The section makes the following distinctions normative:", "sourceParagraphs": [ 105, 106 ] }, { "type": "bullet", "text": "Domain name and legal company", "sourceParagraph": 107 }, { "type": "bullet", "text": "Brand and brand owner", "sourceParagraph": 108 }, { "type": "bullet", "text": "Parent company and subsidiary", "sourceParagraph": 109 }, { "type": "bullet", "text": "Group and group company", "sourceParagraph": 110 }, { "type": "bullet", "text": "Product and manufacturer", "sourceParagraph": 111 }, { "type": "bullet", "text": "Service and service provider", "sourceParagraph": 112 }, { "type": "bullet", "text": "Founder and institution", "sourceParagraph": 113 }, { "type": "bullet", "text": "Employee and institution", "sourceParagraph": 114 }, { "type": "bullet", "text": "Licence holder and the party using the brand", "sourceParagraph": 115 }, { "type": "bullet", "text": "Franchisor and franchise operator", "sourceParagraph": 116 }, { "type": "bullet", "text": "Manufacturer and distributor", "sourceParagraph": 117 }, { "type": "bullet", "text": "Platform and seller on the platform", "sourceParagraph": 118 }, { "type": "bullet", "text": "Technology provider and technology user", "sourceParagraph": 119 }, { "type": "bullet", "text": "Customer and business partner", "sourceParagraph": 120 }, { "type": "bullet", "text": "Active relationship and historical relationship", "sourceParagraph": 121 }, { "type": "bullet", "text": "Official name and descriptive translation", "sourceParagraph": 122 }, { "type": "bullet", "text": "Different entities with the same name", "sourceParagraph": 123 }, { "type": "bullet", "text": "Previous and subsequent versions of an entity", "sourceParagraph": 124 }, { "type": "bullet", "text": "Audited object and object to which the conformity decision will be applied", "sourceParagraph": 125 }, { "type": "paragraph", "text": "At the end of this section, each GEO audit should be able to provide a single and traceable answer to the following question:", "sourceParagraphs": [ 126 ] }, { "type": "callout", "text": "Exactly which entity, with which relationships, for which period, and for which digital surfaces does this score and decision belong?", "sourceParagraph": 127 }, { "type": "heading2", "text": "2. CENTRAL NORMATIVE PROVISION", "sourceParagraph": 129 }, { "type": "callout", "text": "Before data collection begins, every GEO-1000 audit must define a unique audited entity, a canonical entity record, a scope boundary, a time boundary and a relationship graph.", "sourceParagraph": 130 }, { "type": "paragraph", "text": "None of the following, on its own, is sufficient to identify an entity:", "sourceParagraphs": [ 131 ] }, { "type": "bullet", "text": "name,", "sourceParagraph": 132 }, { "type": "bullet", "text": "logo,", "sourceParagraph": 133 }, { "type": "bullet", "text": "domain name,", "sourceParagraph": 134 }, { "type": "bullet", "text": "social media account", "sourceParagraph": 135 }, { "type": "paragraph", "text": "These identifiers alone do not define the audit object. The audited entity must be described, where relevant, through at least the following elements:", "sourceParagraphs": [ 136, 137 ] }, { "type": "bullet", "text": "Entity type", "sourceParagraph": 138 }, { "type": "bullet", "text": "Canonical name", "sourceParagraph": 139 }, { "type": "bullet", "text": "Legal name", "sourceParagraph": 140 }, { "type": "bullet", "text": "Brand names", "sourceParagraph": 141 }, { "type": "bullet", "text": "Official localised names", "sourceParagraph": 142 }, { "type": "bullet", "text": "Abbreviations and aliases", "sourceParagraph": 143 }, { "type": "bullet", "text": "Establishment or start date", "sourceParagraph": 144 }, { "type": "bullet", "text": "Jurisdiction", "sourceParagraph": 145 }, { "type": "bullet", "text": "Registration number or equivalent official identity", "sourceParagraph": 146 }, { "type": "bullet", "text": "Parent and subsidiary company relationships", "sourceParagraph": 147 }, { "type": "bullet", "text": "Trademark ownership", "sourceParagraph": 148 }, { "type": "bullet", "text": "Domain name control", "sourceParagraph": 149 }, { "type": "bullet", "text": "Product and service ownership", "sourceParagraph": 150 }, { "type": "bullet", "text": "Actual service delivery responsible", "sourceParagraph": 151 }, { "type": "bullet", "text": "Claimant", "sourceParagraph": 152 }, { "type": "bullet", "text": "Evidence holder", "sourceParagraph": 153 }, { "type": "bullet", "text": "Active countries and locales", "sourceParagraph": 154 }, { "type": "bullet", "text": "Validity period", "sourceParagraph": 155 }, { "type": "bullet", "text": "Historical name and relationships", "sourceParagraph": 156 }, { "type": "bullet", "text": "Surfaces included in the audit", "sourceParagraph": 157 }, { "type": "bullet", "text": "Entities and surfaces outside the scope", "sourceParagraph": 158 }, { "type": "paragraph", "text": "The main GEO score should not be published until this record is completed.", "sourceParagraphs": [ 159 ] }, { "type": "heading2", "text": "3. WHAT IS AN ENTITY?", "sourceParagraph": 161 }, { "type": "paragraph", "text": "In this protocol, an entity is an object that can be assigned a separate and traceable identity, characteristic, relationship, responsibility, or time status. An entity:", "sourceParagraphs": [ 162, 163 ] }, { "type": "bullet", "text": "human,", "sourceParagraph": 164 }, { "type": "bullet", "text": "legal entity,", "sourceParagraph": 165 }, { "type": "bullet", "text": "public institution,", "sourceParagraph": 166 }, { "type": "bullet", "text": "brand,", "sourceParagraph": 167 }, { "type": "bullet", "text": "product,", "sourceParagraph": 168 }, { "type": "bullet", "text": "service,", "sourceParagraph": 169 }, { "type": "bullet", "text": "domain name,", "sourceParagraph": 170 }, { "type": "bullet", "text": "digital application,", "sourceParagraph": 171 }, { "type": "bullet", "text": "group of companies,", "sourceParagraph": 172 }, { "type": "bullet", "text": "physical branch,", "sourceParagraph": 173 }, { "type": "bullet", "text": "programme,", "sourceParagraph": 174 }, { "type": "bullet", "text": "certificate,", "sourceParagraph": 175 }, { "type": "bullet", "text": "publication,", "sourceParagraph": 176 }, { "type": "bullet", "text": "research project", "sourceParagraph": 177 }, { "type": "paragraph", "text": "It is possible. Not every mentioned item must be a separate entity. However, if one of the following areas changes materially, a separate entity record may be required:", "sourceParagraphs": [ 178, 179, 180 ] }, { "type": "bullet", "text": "Legal responsibility", "sourceParagraph": 181 }, { "type": "bullet", "text": "Ownership", "sourceParagraph": 182 }, { "type": "bullet", "text": "Licence", "sourceParagraph": 183 }, { "type": "bullet", "text": "Product or service delivery", "sourceParagraph": 184 }, { "type": "bullet", "text": "Jurisdiction", "sourceParagraph": 185 }, { "type": "bullet", "text": "Time period", "sourceParagraph": 186 }, { "type": "bullet", "text": "Obligation to the user", "sourceParagraph": 187 }, { "type": "bullet", "text": "Evidence and claim ownership", "sourceParagraph": 188 }, { "type": "bullet", "text": "Audit scope", "sourceParagraph": 189 }, { "type": "heading2", "text": "4. SIX SEPARATE OBJECTS IN GEO AUDIT", "sourceParagraph": 191 }, { "type": "paragraph", "text": "In a GEO audit, the word “entity” often conceals six separate functions. These six objects must be distinguished from each other.", "sourceParagraphs": [ 192, 193 ] }, { "type": "heading3", "text": "4.1. Prompt anchor", "sourceParagraph": 195 }, { "type": "paragraph", "text": "A prompt anchor is an expression or sign used to invoke the existence in a user's query. Examples:", "sourceParagraphs": [ 196, 197 ] }, { "type": "bullet", "text": "Company name", "sourceParagraph": 198 }, { "type": "bullet", "text": "Brand name", "sourceParagraph": 199 }, { "type": "bullet", "text": "Domain name", "sourceParagraph": 200 }, { "type": "bullet", "text": "Founder name", "sourceParagraph": 201 }, { "type": "bullet", "text": "Product name", "sourceParagraph": 202 }, { "type": "bullet", "text": "Application name", "sourceParagraph": 203 }, { "type": "bullet", "text": "Local company name", "sourceParagraph": 204 }, { "type": "bullet", "text": "Abbreviation", "sourceParagraph": 205 }, { "type": "paragraph", "text": "Example: “What kind of company is Apple.com?” the prompt anchor here: apple.com is the domain name. the prompt anchor does not have to be the entity itself. It is only the entry point used for the system to resolve the entity.", "sourceParagraphs": [ 206, 207, 208, 209, 210, 211, 212 ] }, { "type": "heading3", "text": "4.2. Target Entity", "sourceParagraph": 214 }, { "type": "paragraph", "text": "The target entity is the entity that the protocol actually wants to measure. Example: the canonical corporate entity associated with the apple.com domain name can be targeted. However, its exact identity must be recorded before the audit. The target entity cannot be chosen by looking at the results after the query.", "sourceParagraphs": [ 215, 216, 217, 218, 219, 220 ] }, { "type": "heading3", "text": "4.3. Entity Resolved by the System", "sourceParagraph": 222 }, { "type": "paragraph", "text": "A resolved entity is how the AI product interprets the query anchor as an entity. Representation:", "sourceParagraphs": [ 223, 224 ] }, { "type": "formula", "text": "E_i", "sourceParagraph": 225 }, { "type": "paragraph", "text": "can be used. The target entity:", "sourceParagraphs": [ 226, 227 ] }, { "type": "heading3", "text": "E*", "sourceParagraph": 228 }, { "type": "paragraph", "text": "is the identity resolution assessment that asks the fundamental question:", "sourceParagraphs": [ 229 ] }, { "type": "formula", "text": "E_i = E*?", "sourceParagraph": 230 }, { "type": "paragraph", "text": "However, equality is not always just equality of name. The system can identify the correct brand and corporate entity without naming the correct legal entity. Therefore, identity matching later:", "sourceParagraphs": [ 231, 232, 233 ] }, { "type": "bullet", "text": "exact match,", "sourceParagraph": 234 }, { "type": "bullet", "text": "acceptable brand match,", "sourceParagraph": 235 }, { "type": "bullet", "text": "partial match,", "sourceParagraph": 236 }, { "type": "bullet", "text": "wrong entity,", "sourceParagraph": 237 }, { "type": "bullet", "text": "uncertain,", "sourceParagraph": 238 }, { "type": "bullet", "text": "multiple entity merger", "sourceParagraph": 239 }, { "type": "paragraph", "text": "can be classified as.", "sourceParagraphs": [ 240 ] }, { "type": "heading3", "text": "4.4. Object of the Claim", "sourceParagraph": 242 }, { "type": "paragraph", "text": "The object of the claim is the entity to which a specific feature in the response belongs. Within the same response:", "sourceParagraphs": [ 243, 244 ] }, { "type": "bullet", "text": "establishment date to the parent company,", "sourceParagraph": 245 }, { "type": "bullet", "text": "product feature to the product,", "sourceParagraph": 246 }, { "type": "bullet", "text": "customer review for local service provider,", "sourceParagraph": 247 }, { "type": "bullet", "text": "licence to a specific legal entity,", "sourceParagraph": 248 }, { "type": "bullet", "text": "founder's experience to a person", "sourceParagraph": 249 }, { "type": "paragraph", "text": "may belong. Even if the answer started with the correct target entity, it may transfer features to incorrect objects. Therefore, each factual claim should be evaluated in the following structure:", "sourceParagraphs": [ 250, 251, 252 ] }, { "type": "formula", "text": "(subject, relationship or attribute, value, time, scope)", "sourceParagraph": 253 }, { "type": "paragraph", "text": "Example: “X has a Y licence.” Here, it should be determined to which:", "sourceParagraphs": [ 254, 255, 256 ] }, { "type": "bullet", "text": "legal entity,", "sourceParagraph": 257 }, { "type": "bullet", "text": "country,", "sourceParagraph": 258 }, { "type": "bullet", "text": "during which dates,", "sourceParagraph": 259 }, { "type": "bullet", "text": "for the activity", "sourceParagraph": 260 }, { "type": "paragraph", "text": "it belongs.", "sourceParagraphs": [ 261 ] }, { "type": "heading3", "text": "4.5. Audit Object", "sourceParagraph": 263 }, { "type": "paragraph", "text": "The audit object is a defined entity or package of entities where evidence, content, technical surface, and AI outputs are examined together. An audit object can be a single entity. Example: A specific legal company Or it can be a package with clear boundaries: Main brand + canonical domain name + specified two products + Turkish and English public surfaces. If a package is used, the entities within it must retain their separate identities. \"Package\" does not mean that all attributes are transferred to each other.", "sourceParagraphs": [ 264, 265, 266, 267, 268, 269, 270, 271 ] }, { "type": "heading3", "text": "4.6. Conformity Unit", "sourceParagraph": 273 }, { "type": "paragraph", "text": "The conformity unit is the specific object to which the audit decision will be legally and normatively applied. A conformity decision:", "sourceParagraphs": [ 274, 275 ] }, { "type": "bullet", "text": "applies to the entire group of companies,", "sourceParagraph": 276 }, { "type": "bullet", "text": "applies to all languages,", "sourceParagraph": 277 }, { "type": "bullet", "text": "applies to all domain names,", "sourceParagraph": 278 }, { "type": "bullet", "text": "to all products", "sourceParagraph": 279 }, { "type": "paragraph", "text": "cannot be applied automatically. Example: “The Turkish and English public pages of the brand NobleJackal on example.com have been found compliant according to the specified standard and within the specified date range.” This decision:", "sourceParagraphs": [ 280, 281, 282, 283 ] }, { "type": "bullet", "text": "all internal processes of the company,", "sourceParagraph": 284 }, { "type": "bullet", "text": "affiliated companies,", "sourceParagraph": 285 }, { "type": "bullet", "text": "separate applications,", "sourceParagraph": 286 }, { "type": "bullet", "text": "languages to be added in the future,", "sourceParagraph": 287 }, { "type": "bullet", "text": "products outside the scope", "sourceParagraph": 288 }, { "type": "paragraph", "text": "does not automatically cover.", "sourceParagraphs": [ 289 ] }, { "type": "heading2", "text": "5. TYPES OF ENTITIES", "sourceParagraph": 291 }, { "type": "paragraph", "text": "Every audit entity must be classified with one or more types. Primary type and secondary types can be distinguished.", "sourceParagraphs": [ 292, 293 ] }, { "type": "heading3", "text": "5.1. Legal Entity", "sourceParagraph": 295 }, { "type": "paragraph", "text": "It is a company that carries separate rights, responsibilities, or registration under the law or relevant legal regulation. Its fields to the extent they are relevant:", "sourceParagraphs": [ 296, 297 ] }, { "type": "bullet", "text": "Legal name", "sourceParagraph": 298 }, { "type": "bullet", "text": "Company type", "sourceParagraph": 299 }, { "type": "bullet", "text": "Place of registration", "sourceParagraph": 300 }, { "type": "bullet", "text": "Registration number", "sourceParagraph": 301 }, { "type": "bullet", "text": "Tax or equivalent identity", "sourceParagraph": 302 }, { "type": "bullet", "text": "Date of establishment", "sourceParagraph": 303 }, { "type": "bullet", "text": "Active status", "sourceParagraph": 304 }, { "type": "bullet", "text": "Authorised representative", "sourceParagraph": 305 }, { "type": "bullet", "text": "Parent and subsidiary relationships", "sourceParagraph": 306 }, { "type": "paragraph", "text": "A legal entity may carry the same name as the brand. However, it is not the same object.", "sourceParagraphs": [ 307, 308 ] }, { "type": "heading3", "text": "5.2. Brand", "sourceParagraph": 310 }, { "type": "paragraph", "text": "It is the name used in the market for a product, service, or corporate identity. Brand:", "sourceParagraphs": [ 311, 312 ] }, { "type": "bullet", "text": "may not have a separate legal entity,", "sourceParagraph": 313 }, { "type": "bullet", "text": "may be owned by another company,", "sourceParagraph": 314 }, { "type": "bullet", "text": "may be used under licence,", "sourceParagraph": 315 }, { "type": "paragraph", "text": "may be operated by multiple companies in different countries. The actual delivery responsible for the work claimed by a brand must also be specified.", "sourceParagraphs": [ 316, 317 ] }, { "type": "heading3", "text": "5.3. Group of Companies", "sourceParagraph": 319 }, { "type": "paragraph", "text": "It is a structure formed by multiple legal entities that have joint ownership or control relationships. A “Group” cannot automatically transfer the following features to all members:", "sourceParagraphs": [ 320, 321 ] }, { "type": "bullet", "text": "Licence", "sourceParagraph": 322 }, { "type": "bullet", "text": "Certificate", "sourceParagraph": 323 }, { "type": "bullet", "text": "Number of customers", "sourceParagraph": 324 }, { "type": "bullet", "text": "Number of employees", "sourceParagraph": 325 }, { "type": "bullet", "text": "Revenue", "sourceParagraph": 326 }, { "type": "bullet", "text": "Office", "sourceParagraph": 327 }, { "type": "bullet", "text": "Audit result", "sourceParagraph": 328 }, { "type": "bullet", "text": "Security status", "sourceParagraph": 329 }, { "type": "bullet", "text": "Conformity mark", "sourceParagraph": 330 }, { "type": "paragraph", "text": "If the group-level value is truly consolidated, it should be clearly stated.", "sourceParagraphs": [ 331 ] }, { "type": "heading3", "text": "5.4. Subsidiary or Associate", "sourceParagraph": 333 }, { "type": "paragraph", "text": "A legal entity that is under the ownership or control relationship of another entity. A subsidiary:", "sourceParagraphs": [ 334, 335 ] }, { "type": "bullet", "text": "may be a separate contract party,", "sourceParagraph": 336 }, { "type": "bullet", "text": "a separate licence holder,", "sourceParagraph": 337 }, { "type": "bullet", "text": "a separate employer,", "sourceParagraph": 338 }, { "type": "bullet", "text": "a separate data controller", "sourceParagraph": 339 }, { "type": "paragraph", "text": "may be. The characteristics of the parent company are not transferred by default to the subsidiary.", "sourceParagraphs": [ 340, 341 ] }, { "type": "heading3", "text": "5.5. Domain Name", "sourceParagraph": 343 }, { "type": "paragraph", "text": "It is a digital address and publishing surface. Domain name:", "sourceParagraphs": [ 344, 345 ] }, { "type": "bullet", "text": "is not a legal entity,", "sourceParagraph": 346 }, { "type": "bullet", "text": "does not deliver services on its own,", "sourceParagraph": 347 }, { "type": "bullet", "text": "is not required to own a trademark,", "sourceParagraph": 348 }, { "type": "paragraph", "text": "can be controlled by different owners or operators at certain times. A domain name can be the starting point of an audit. However, it should also be shown under which entity the decision was made.", "sourceParagraphs": [ 349, 350, 351 ] }, { "type": "heading3", "text": "5.6. Website or Digital Publication", "sourceParagraph": 353 }, { "type": "paragraph", "text": "It is the entirety of specific content and applications on a domain name. The same domain name can be:", "sourceParagraphs": [ 354, 355 ] }, { "type": "bullet", "text": "corporate site,", "sourceParagraph": 356 }, { "type": "bullet", "text": "store,", "sourceParagraph": 357 }, { "type": "bullet", "text": "support portal,", "sourceParagraph": 358 }, { "type": "bullet", "text": "investor page,", "sourceParagraph": 359 }, { "type": "bullet", "text": "developer documentation,", "sourceParagraph": 360 }, { "type": "bullet", "text": "different country and language versions", "sourceParagraph": 361 }, { "type": "paragraph", "text": "may carry. The content owners and oversight scopes of these may differ.", "sourceParagraphs": [ 362, 363 ] }, { "type": "heading3", "text": "5.7. Product", "sourceParagraph": 365 }, { "type": "paragraph", "text": "A commercial or technical object carrying a specific function, model, or version. A product's:", "sourceParagraphs": [ 366, 367 ] }, { "type": "bullet", "text": "price,", "sourceParagraph": 368 }, { "type": "bullet", "text": "safety,", "sourceParagraph": 369 }, { "type": "bullet", "text": "technical feature,", "sourceParagraph": 370 }, { "type": "bullet", "text": "certificate,", "sourceParagraph": 371 }, { "type": "bullet", "text": "user suitability", "sourceParagraph": 372 }, { "type": "paragraph", "text": "cannot be used as general characteristics belonging to the company. Versions within the same product family may also differ.", "sourceParagraphs": [ 373, 374 ] }, { "type": "heading3", "text": "5.8. Service", "sourceParagraph": 376 }, { "type": "paragraph", "text": "Activity offered under a specific scope, user, country, delivery method, and responsibility. A service:", "sourceParagraphs": [ 377, 378 ] }, { "type": "bullet", "text": "marketable by the brand,", "sourceParagraph": 379 }, { "type": "bullet", "text": "billable by another legal entity,", "sourceParagraph": 380 }, { "type": "paragraph", "text": "deliverable by an independent partner. Audit should distinguish who:", "sourceParagraphs": [ 381, 382 ] }, { "type": "bullet", "text": "sold it,", "sourceParagraph": 383 }, { "type": "bullet", "text": "delivered it,", "sourceParagraph": 384 }, { "type": "bullet", "text": "warranted it,", "sourceParagraph": 385 }, { "type": "bullet", "text": "legally assumed it", "sourceParagraph": 386 }, { "type": "paragraph", "text": "should be separated.", "sourceParagraphs": [ 387 ] }, { "type": "heading3", "text": "5.9. Person", "sourceParagraph": 389 }, { "type": "paragraph", "text": "Is a founder, executive, employee, consultant, expert, or publicly recognised representative. A person's:", "sourceParagraphs": [ 390, 391 ] }, { "type": "bullet", "text": "education,", "sourceParagraph": 392 }, { "type": "bullet", "text": "experience,", "sourceParagraph": 393 }, { "type": "bullet", "text": "licence,", "sourceParagraph": 394 }, { "type": "bullet", "text": "opinion", "sourceParagraph": 395 }, { "type": "paragraph", "text": "It is not automatically transferred to the institution. Likewise, the award or certificate of the institution is not considered a professional competence of the person.", "sourceParagraphs": [ 396, 397 ] }, { "type": "heading3", "text": "5.10. Branch or Physical Location", "sourceParagraph": 399 }, { "type": "paragraph", "text": "It is a physical service point that is affiliated with the main institution or operated separately. A branch:", "sourceParagraphs": [ 400, 401 ] }, { "type": "bullet", "text": "part of the parent company,", "sourceParagraph": 402 }, { "type": "bullet", "text": "separate legal entity,", "sourceParagraph": 403 }, { "type": "bullet", "text": "franchise,", "sourceParagraph": 404 }, { "type": "bullet", "text": "Independent licensed operator", "sourceParagraph": 405 }, { "type": "paragraph", "text": "maybe. The position relationship does not show the legal relationship alone.", "sourceParagraphs": [ 406, 407 ] }, { "type": "heading3", "text": "5.11. Franchise", "sourceParagraph": 409 }, { "type": "paragraph", "text": "It can be an independent business operating under a brand usage right or business model licence. The franchisor must:", "sourceParagraphs": [ 410, 411 ] }, { "type": "bullet", "text": "reputation,", "sourceParagraph": 412 }, { "type": "bullet", "text": "control,", "sourceParagraph": 413 }, { "type": "bullet", "text": "training", "sourceParagraph": 414 }, { "type": "paragraph", "text": "does not automatically guarantee every action of the franchise operator. A local mistake of the franchise operator cannot be automatically generalised to the entire global brand. The relationship should be reported with its scope.", "sourceParagraphs": [ 415, 416, 417 ] }, { "type": "heading3", "text": "5.12. Distributor, Dealer or Seller", "sourceParagraph": 419 }, { "type": "paragraph", "text": "The party that sells or distributes the product or service in a specific market, whether independent or affiliated. The manufacturer’s licence, control, or corporate compliance is not automatically transferred to the distributor. The seller’s customer service performance is also not the same metric as the manufacturer’s product quality.", "sourceParagraphs": [ 420, 421, 422 ] }, { "type": "heading3", "text": "5.13. Platform and Platform Seller", "sourceParagraph": 424 }, { "type": "paragraph", "text": "The marketplace or platform and the independent sellers on the platform are separate entities. Platform:", "sourceParagraphs": [ 425, 426 ] }, { "type": "bullet", "text": "payment,", "sourceParagraph": 427 }, { "type": "bullet", "text": "listing,", "sourceParagraph": 428 }, { "type": "bullet", "text": "technical infrastructure", "sourceParagraph": 429 }, { "type": "paragraph", "text": "can be provided. Vendor:", "sourceParagraphs": [ 430, 431 ] }, { "type": "bullet", "text": "product,", "sourceParagraph": 432 }, { "type": "bullet", "text": "delivery,", "sourceParagraph": 433 }, { "type": "bullet", "text": "warranty,", "sourceParagraph": 434 }, { "type": "bullet", "text": "customer relations", "sourceParagraph": 435 }, { "type": "paragraph", "text": "can be responsible. If the AI combines these roles, incorrect responsibility assignment may occur.", "sourceParagraphs": [ 436, 437 ] }, { "type": "heading3", "text": "5.14. Programme, Badge or Certificate", "sourceParagraph": 439 }, { "type": "paragraph", "text": "Offered by an institution:", "sourceParagraphs": [ 440 ] }, { "type": "bullet", "text": "training programme,", "sourceParagraph": 441 }, { "type": "bullet", "text": "conformity mark,", "sourceParagraph": 442 }, { "type": "bullet", "text": "certificate,", "sourceParagraph": 443 }, { "type": "bullet", "text": "membership,", "sourceParagraph": 444 }, { "type": "bullet", "text": "badge", "sourceParagraph": 445 }, { "type": "paragraph", "text": "It can carry separate entity records. The institution that publishes the programme is not the same as the institution participating in the programme. The feature of the badge does not translate into all the features of the institution that bears the badge.", "sourceParagraphs": [ 446, 447, 448 ] }, { "type": "heading2", "text": "6. ENTITY GRAPH", "sourceParagraph": 450 }, { "type": "paragraph", "text": "A single entity record may not be enough to explain complex relationships. For this reason, an entity graph must be created for each audit. Views:", "sourceParagraphs": [ 451, 452, 453 ] }, { "type": "heading3", "text": "G_E = (V, R)", "sourceParagraph": 454 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 455 ] }, { "type": "bullet", "text": "V: entity nodes", "sourceParagraph": 456 }, { "type": "bullet", "text": "R: directional relationships between nodes", "sourceParagraph": 457 }, { "type": "paragraph", "text": "Relationships should be defined by explicit verbs.", "sourceParagraphs": [ 458 ] }, { "type": "heading3", "text": "6.1. Basic Types of Relationships", "sourceParagraph": 460 }, { "type": "paragraph", "text": "To the extent relevant, the following relationships can be used:", "sourceParagraphs": [ 461 ] }, { "type": "bullet", "text": "owns — owns", "sourceParagraph": 462 }, { "type": "bullet", "text": "ownedBy — is owned by", "sourceParagraph": 463 }, { "type": "bullet", "text": "controls — controls", "sourceParagraph": 464 }, { "type": "bullet", "text": "controlledBy — is controlled by", "sourceParagraph": 465 }, { "type": "bullet", "text": "operates — operates", "sourceParagraph": 466 }, { "type": "bullet", "text": "operatedBy — is operated by", "sourceParagraph": 467 }, { "type": "bullet", "text": "publishes — publishes", "sourceParagraph": 468 }, { "type": "bullet", "text": "publishedBy — is published by", "sourceParagraph": 469 }, { "type": "bullet", "text": "manufactures — manufactures", "sourceParagraph": 470 }, { "type": "bullet", "text": "manufacturedBy — is manufactured by", "sourceParagraph": 471 }, { "type": "bullet", "text": "providesService — provides service", "sourceParagraph": 472 }, { "type": "bullet", "text": "serviceDeliveredBy — service is delivered by", "sourceParagraph": 473 }, { "type": "bullet", "text": "licencesBrandTo — licences brand to", "sourceParagraph": 474 }, { "type": "bullet", "text": "franchisesTo — franchises to", "sourceParagraph": 475 }, { "type": "bullet", "text": "distributesFor — distributes for", "sourceParagraph": 476 }, { "type": "bullet", "text": "sells — sells", "sourceParagraph": 477 }, { "type": "bullet", "text": "employs — employs", "sourceParagraph": 478 }, { "type": "bullet", "text": "foundedBy — was founded by", "sourceParagraph": 479 }, { "type": "bullet", "text": "memberOf — is a member of", "sourceParagraph": 480 }, { "type": "bullet", "text": "certifiedBy — is certified in the specified scope by", "sourceParagraph": 481 }, { "type": "bullet", "text": "auditedBy — has been audited by", "sourceParagraph": 482 }, { "type": "bullet", "text": "partneredWith — has a partnership in the specified scope with", "sourceParagraph": 483 }, { "type": "bullet", "text": "customerOf — is a customer of", "sourceParagraph": 484 }, { "type": "bullet", "text": "supplierTo — is a supplier to", "sourceParagraph": 485 }, { "type": "bullet", "text": "formerlyKnownAs — formerly known as", "sourceParagraph": 486 }, { "type": "bullet", "text": "predecessorOf — is a predecessor of", "sourceParagraph": 487 }, { "type": "bullet", "text": "successorTo — is a successor of", "sourceParagraph": 488 }, { "type": "bullet", "text": "acquiredBy — has been acquired by", "sourceParagraph": 489 }, { "type": "bullet", "text": "mergedInto — has been merged into", "sourceParagraph": 490 }, { "type": "bullet", "text": "ceasedOn — has ceased operations on", "sourceParagraph": 491 }, { "type": "paragraph", "text": "Vague verbs like 'related', 'partner', 'working together' should not be used alone as much as possible.", "sourceParagraphs": [ 492 ] }, { "type": "heading3", "text": "6.2. Direction of the Relationship", "sourceParagraph": 494 }, { "type": "paragraph", "text": "Relationships are directional. 'A provides technology to B.' is not the same as: 'B provides technology to A.' A brand may belong to a company. The company does not belong to the brand. A client may have worked with an agency. The agency does not belong to the client. If the direction is not specified, the system may make incorrect responsibility and authority transfers.", "sourceParagraphs": [ 495, 496, 497, 498, 499, 500, 501, 502, 503, 504 ] }, { "type": "heading3", "text": "6.3. Time of the Relationship", "sourceParagraph": 506 }, { "type": "paragraph", "text": "Every financial relationship to the extent it is relevant:", "sourceParagraphs": [ 507 ] }, { "type": "bullet", "text": "start date,", "sourceParagraph": 508 }, { "type": "bullet", "text": "end date,", "sourceParagraph": 509 }, { "type": "bullet", "text": "activity status,", "sourceParagraph": 510 }, { "type": "bullet", "text": "last verification date", "sourceParagraph": 511 }, { "type": "paragraph", "text": "must carry. An old partnership is not an active partnership. A former employee is not a current executive. A distributor once used may not be an authorised representative today.", "sourceParagraphs": [ 512, 513, 514, 515 ] }, { "type": "heading3", "text": "6.4. Scope of the Relationship", "sourceParagraph": 517 }, { "type": "paragraph", "text": "A partnership:", "sourceParagraphs": [ 518 ] }, { "type": "bullet", "text": "may be limited to a specific product,", "sourceParagraph": 519 }, { "type": "bullet", "text": "country,", "sourceParagraph": 520 }, { "type": "bullet", "text": "campaign,", "sourceParagraph": 521 }, { "type": "bullet", "text": "date,", "sourceParagraph": 522 }, { "type": "bullet", "text": "event", "sourceParagraph": 523 }, { "type": "paragraph", "text": "It cannot be generalised that a single-project collaboration applies to all corporate activities.", "sourceParagraphs": [ 524, 525 ] }, { "type": "heading2", "text": "7. ASSUMED TRANSFER OF CHARACTERISTICS IS PROHIBITED", "sourceParagraph": 527 }, { "type": "paragraph", "text": "The existence of a relationship between entities does not automatically transfer a characteristic from one to another. One of the fundamental principles of this section is:", "sourceParagraphs": [ 528, 529 ] }, { "type": "callout", "text": "Feature transfer is prohibited by default; however, it can be allowed with explicit relation, scope, timing, and evidence.", "sourceParagraph": 530 }, { "type": "heading3", "text": "7.1. Licence Transfer", "sourceParagraph": 532 }, { "type": "paragraph", "text": "The licence of the parent company:", "sourceParagraphs": [ 533 ] }, { "type": "bullet", "text": "to a subsidiary,", "sourceParagraph": 534 }, { "type": "bullet", "text": "to a franchise,", "sourceParagraph": 535 }, { "type": "bullet", "text": "to a distributor,", "sourceParagraph": 536 }, { "type": "bullet", "text": "to a brand", "sourceParagraph": 537 }, { "type": "paragraph", "text": "do not automatically transfer. The official scope of the licence must be examined.", "sourceParagraphs": [ 538, 539 ] }, { "type": "heading3", "text": "7.2. Certificate Transfer", "sourceParagraph": 541 }, { "type": "paragraph", "text": "A product’s certificate cannot be transferred to the entire product family. The quality certificate of one office cannot be generalised to the global group. A company’s conformity decision does not prove the personal expertise of its employees.", "sourceParagraphs": [ 542, 543, 544 ] }, { "type": "heading3", "text": "7.3. Employee Number Transfer", "sourceParagraph": 546 }, { "type": "paragraph", "text": "The total of the group cannot be used like the number of employees of a local company. The agency network or freelancer pool cannot be shown as the number of direct employees.", "sourceParagraphs": [ 547, 548 ] }, { "type": "heading3", "text": "7.4. Customer Number Transfer", "sourceParagraph": 550 }, { "type": "paragraph", "text": "The number of customers of the main group cannot be written as the number of customers of a newly established brand. The number of users on a platform may not correspond to the number of active customers of a specific product.", "sourceParagraphs": [ 551, 552 ] }, { "type": "heading3", "text": "7.5. Establishment Date Transfer", "sourceParagraph": 554 }, { "type": "paragraph", "text": "The founder's starting year in the industry is not the company's establishment date. A company cannot use the founding date of an old brand it acquires as its own legal establishment date. The brand date and the legal entity date should be shown separately.", "sourceParagraphs": [ 555, 556, 557 ] }, { "type": "heading3", "text": "7.6. Success and Case Transfer", "sourceParagraph": 559 }, { "type": "paragraph", "text": "A project carried out by an employee at a previous employer is not the corporate case study of the new company. The success of an acquired company, without explanation of integration and scope, is not considered the direct success of the new group.", "sourceParagraphs": [ 560, 561 ] }, { "type": "heading3", "text": "7.7. Transfer of Eligibility", "sourceParagraph": 563 }, { "type": "paragraph", "text": "Eligibility granted for certain language pages of a domain name:", "sourceParagraphs": [ 564 ] }, { "type": "bullet", "text": "to the entire company group,", "sourceParagraph": 565 }, { "type": "bullet", "text": "to new products,", "sourceParagraph": 566 }, { "type": "bullet", "text": "to other domain names,", "sourceParagraph": 567 }, { "type": "bullet", "text": "to future pages", "sourceParagraph": 568 }, { "type": "paragraph", "text": "is not transferred.", "sourceParagraphs": [ 569 ] }, { "type": "heading2", "text": "8. FOUR SEPARATE OWNERSHIP LAYERS", "sourceParagraph": 571 }, { "type": "paragraph", "text": "In a digital representation, the word “owner” can carry multiple meanings. Four layers should be distinguished.", "sourceParagraphs": [ 572, 573 ] }, { "type": "heading3", "text": "8.1. Legal Ownership", "sourceParagraph": 575 }, { "type": "paragraph", "text": "The entity is the owner under law or official registration.", "sourceParagraphs": [ 576 ] }, { "type": "heading3", "text": "8.2. Trademark Ownership", "sourceParagraph": 578 }, { "type": "paragraph", "text": "It is the owner of the rights over the trademark or trade identity.", "sourceParagraphs": [ 579 ] }, { "type": "heading3", "text": "8.3. Digital Surface Control", "sourceParagraph": 581 }, { "type": "paragraph", "text": "It is the party that actually controls the domain name, website, social profile, or application. The legal owner and the technical controller can be different.", "sourceParagraphs": [ 582, 583 ] }, { "type": "heading3", "text": "8.4. Content and Claim Ownership", "sourceParagraph": 585 }, { "type": "paragraph", "text": "It is the party responsible for the accuracy of the published material claim. The agency may manage the site. However, the true owner of the price or licence claim may be the client institution. A technical provider does not become the material owner of the claim just because they published the content.", "sourceParagraphs": [ 586, 587, 588, 589 ] }, { "type": "heading2", "text": "9. WHO SELLS THE SERVICE, WHO DELIVERS IT?", "sourceParagraph": 591 }, { "type": "paragraph", "text": "This is one of the most important entity distinctions in the GEO assessment. In a service chain, the following parties may differ:", "sourceParagraphs": [ 592, 593 ] }, { "type": "bullet", "text": "The one using the brand", "sourceParagraph": 594 }, { "type": "bullet", "text": "The one advertising", "sourceParagraph": 595 }, { "type": "bullet", "text": "The one preparing the offer", "sourceParagraph": 596 }, { "type": "bullet", "text": "The one signing the contract", "sourceParagraph": 597 }, { "type": "bullet", "text": "The one issuing the invoice", "sourceParagraph": 598 }, { "type": "bullet", "text": "The one actually delivering the service", "sourceParagraph": 599 }, { "type": "bullet", "text": "The one providing the guarantee", "sourceParagraph": 600 }, { "type": "bullet", "text": "The one processing the data", "sourceParagraph": 601 }, { "type": "bullet", "text": "The one handling the complaint", "sourceParagraph": 602 }, { "type": "bullet", "text": "The one bearing legal responsibility", "sourceParagraph": 603 }, { "type": "paragraph", "text": "An AI may assign all these roles to a single party by only stating the brand name. This error is especially common in:", "sourceParagraphs": [ 604, 605 ] }, { "type": "bullet", "text": "tourism,", "sourceParagraph": 606 }, { "type": "bullet", "text": "finance,", "sourceParagraph": 607 }, { "type": "bullet", "text": "health,", "sourceParagraph": 608 }, { "type": "bullet", "text": "consultancy,", "sourceParagraph": 609 }, { "type": "bullet", "text": "franchise,", "sourceParagraph": 610 }, { "type": "bullet", "text": "platform,", "sourceParagraph": 611 }, { "type": "bullet", "text": "agency,", "sourceParagraph": 612 }, { "type": "bullet", "text": "education", "sourceParagraph": 613 }, { "type": "paragraph", "text": "fields can result in financial consequences. Each service record should answer these questions: Who does the user contract with? Who receives the payment? Who performs the service? Who bears professional or legal responsibility? Who is responsible for warranty and return obligations? What is the brand's role? Are there third-party providers? This distinction also determines which entity the GEO score applies to.", "sourceParagraphs": [ 614, 615, 616, 617, 618, 619, 620, 621, 622, 623 ] }, { "type": "heading2", "text": "10. ENTITY BOUNDARY", "sourceParagraph": 625 }, { "type": "paragraph", "text": "The entity boundary is the line that defines which features, relationships, and surfaces the audited object includes and which it does not. An entity boundary can be defined in four layers.", "sourceParagraphs": [ 626, 627 ] }, { "type": "heading3", "text": "10.1. Identity Core", "sourceParagraph": 629 }, { "type": "paragraph", "text": "It is the fundamental identity of the entity that must remain unchanged:", "sourceParagraphs": [ 630 ] }, { "type": "bullet", "text": "Canonical name", "sourceParagraph": 631 }, { "type": "bullet", "text": "Entity type", "sourceParagraph": 632 }, { "type": "bullet", "text": "Legal or institutional identity", "sourceParagraph": 633 }, { "type": "bullet", "text": "Main activity", "sourceParagraph": 634 }, { "type": "bullet", "text": "Primary ownership", "sourceParagraph": 635 }, { "type": "bullet", "text": "Active status", "sourceParagraph": 636 }, { "type": "heading3", "text": "10.2. Controlled Environment", "sourceParagraph": 638 }, { "type": "paragraph", "text": "Directly controlled by the entity:", "sourceParagraphs": [ 639 ] }, { "type": "bullet", "text": "domain names,", "sourceParagraph": 640 }, { "type": "bullet", "text": "applications,", "sourceParagraph": 641 }, { "type": "bullet", "text": "brands,", "sourceParagraph": 642 }, { "type": "bullet", "text": "products,", "sourceParagraph": 643 }, { "type": "bullet", "text": "documents,", "sourceParagraph": 644 }, { "type": "bullet", "text": "official profiles", "sourceParagraph": 645 }, { "type": "paragraph", "text": "This field. Control does not mean that all features merge into a single identity.", "sourceParagraphs": [ 646, 647 ] }, { "type": "heading3", "text": "10.3. Connected Environment", "sourceParagraph": 649 }, { "type": "paragraph", "text": "Affiliates are local companies, franchises, and controlled partners. They retain their separate identities.", "sourceParagraphs": [ 650, 651 ] }, { "type": "heading3", "text": "10.4. External Relationship Environment", "sourceParagraph": 653 }, { "type": "paragraph", "text": "Customers, suppliers, media, memberships, technology providers, and independent partners. Entities in this environment are not part of the audited entity. They are objects of the relationship.", "sourceParagraphs": [ 654, 655, 656 ] }, { "type": "heading2", "text": "11. LOCKING THE ENTITY BOUNDARY WITHOUT SEEING RESULTS", "sourceParagraph": 658 }, { "type": "paragraph", "text": "The audit object cannot be expanded or narrowed based on results. The following behaviours are prohibited:", "sourceParagraphs": [ 659, 660 ] }, { "type": "bullet", "text": "Adding positive group results to the brand", "sourceParagraph": 661 }, { "type": "bullet", "text": "Excluding a failed affiliated company from the scope", "sourceParagraph": 662 }, { "type": "bullet", "text": "Spreading a good product result across the entire product family", "sourceParagraph": 663 }, { "type": "bullet", "text": "Declaring the bad language version as a separate entity", "sourceParagraph": 664 }, { "type": "bullet", "text": "Leaving the critical error to the local operation while making the global brand appear clean", "sourceParagraph": 665 }, { "type": "bullet", "text": "Making strong local evidence the evidence of the global company", "sourceParagraph": 666 }, { "type": "paragraph", "text": "Scope of the entity:", "sourceParagraphs": [ 667 ] }, { "type": "bullet", "text": "Before the audit begins,", "sourceParagraph": 668 }, { "type": "bullet", "text": "Before prompts are distributed,", "sourceParagraph": 669 }, { "type": "bullet", "text": "Before results are seen", "sourceParagraph": 670 }, { "type": "paragraph", "text": "It must be versioned first.", "sourceParagraphs": [ 671 ] }, { "type": "heading2", "text": "12. CANONICAL ENTITY RECORD", "sourceParagraph": 673 }, { "type": "paragraph", "text": "A Canonical Entity Record must be created for each audited entity. This record clarifies which object the adjudicators and systems are evaluating.", "sourceParagraphs": [ 674, 675 ] }, { "type": "heading3", "text": "12.1. Minimum Fields", "sourceParagraph": 677 }, { "type": "heading4", "text": "Identity", "sourceParagraph": 678 }, { "type": "paragraph", "text": "NOMOS entity ID Canonical name Entity type Legal name Abbreviations Official local names Former names", "sourceParagraphs": [ 679, 680, 681, 682, 683, 684, 685 ] }, { "type": "heading4", "text": "Legal and Corporate Status", "sourceParagraph": 686 }, { "type": "paragraph", "text": "Jurisdiction Registration number Date of establishment Active status Parent company Subsidiaries", "sourceParagraphs": [ 687, 688, 689, 690, 691, 692 ] }, { "type": "heading4", "text": "Brands and Products", "sourceParagraph": 693 }, { "type": "paragraph", "text": "Owned brands Licensed brands Products Services Version or model distinctions", "sourceParagraphs": [ 694, 695, 696, 697, 698 ] }, { "type": "heading4", "text": "Digital Surfaces", "sourceParagraph": 699 }, { "type": "paragraph", "text": "Canonical domain names Local domain names Applications Official profiles Machine-readable records", "sourceParagraphs": [ 700, 701, 702, 703, 704 ] }, { "type": "heading4", "text": "Responsibility", "sourceParagraph": 705 }, { "type": "paragraph", "text": "Claimant Content owner Technical surface owner Service delivery owner Legal responsible", "sourceParagraphs": [ 706, 707, 708, 709, 710 ] }, { "type": "heading4", "text": "Time", "sourceParagraph": 711 }, { "type": "paragraph", "text": "Record version Effective start Validity end Last verification date Historical changes", "sourceParagraphs": [ 712, 713, 714, 715, 716 ] }, { "type": "heading4", "text": "Scope", "sourceParagraph": 717 }, { "type": "paragraph", "text": "Audited countries Audited languages Audited products Out-of-scope entities and surfaces", "sourceParagraphs": [ 718, 719, 720, 721 ] }, { "type": "heading3", "text": "12.2. Entity Statuses", "sourceParagraph": 723 }, { "type": "paragraph", "text": "Each entity record must carry one of the following statuses:", "sourceParagraphs": [ 724 ] }, { "type": "heading3", "text": "ACTIVE", "sourceParagraph": 725 }, { "type": "heading3", "text": "INACTIVE", "sourceParagraph": 726 }, { "type": "heading3", "text": "HISTORICAL", "sourceParagraph": 727 }, { "type": "heading3", "text": "SUPERSEDED", "sourceParagraph": 728 }, { "type": "heading3", "text": "MERGED", "sourceParagraph": 729 }, { "type": "heading3", "text": "ACQUIRED", "sourceParagraph": 730 }, { "type": "heading3", "text": "DISSOLVED", "sourceParagraph": 731 }, { "type": "heading3", "text": "PROVISIONAL", "sourceParagraph": 732 }, { "type": "heading3", "text": "CONTESTED", "sourceParagraph": 733 }, { "type": "heading3", "text": "UNKNOWN", "sourceParagraph": 734 }, { "type": "paragraph", "text": "UNKNOWN or CONTESTED identity cannot be definitively resolved silently.", "sourceParagraphs": [ 735 ] }, { "type": "heading2", "text": "13. IDENTITY PROOF LEVELS", "sourceParagraph": 737 }, { "type": "paragraph", "text": "The identity assurance of an entity record can be classified according to the sources used.", "sourceParagraphs": [ 738 ] }, { "type": "heading3", "text": "ID-0 — Identity Unknown", "sourceParagraph": 740 }, { "type": "paragraph", "text": "Only a name or domain name exists. Legal or organisational matching is not verified.", "sourceParagraphs": [ 741, 742 ] }, { "type": "heading3", "text": "ID-1 — First-Party Assertion", "sourceParagraph": 744 }, { "type": "paragraph", "text": "The institution identifies itself in a specific way. It is valuable. It is not independent verification.", "sourceParagraphs": [ 745, 746, 747 ] }, { "type": "heading3", "text": "ID-2 — Supported Organisational Identity", "sourceParagraph": 749 }, { "type": "paragraph", "text": "First-party records, the domain name, and other reliable records are consistent with each other.", "sourceParagraphs": [ 750 ] }, { "type": "heading3", "text": "ID-3 — Verified Identity with Official Record", "sourceParagraph": 752 }, { "type": "paragraph", "text": "The legal entity has been verified with appropriate official or reliable records.", "sourceParagraphs": [ 753 ] }, { "type": "heading3", "text": "ID-4 — Multi-Surface and Relationship-Verified Identity", "sourceParagraph": 755 }, { "type": "paragraph", "text": "Identity, brand, product, domain name, and corporate relationships are consistent across multiple reliable records.", "sourceParagraphs": [ 756 ] }, { "type": "heading3", "text": "ID-5 — Identity Maintained Over Time", "sourceParagraph": 758 }, { "type": "paragraph", "text": "Identity and relationships have been verified at different periods, and the change history is preserved. These levels are not the final eligibility score. They indicate the evidence confidence of the entity record.", "sourceParagraphs": [ 759, 760, 761 ] }, { "type": "heading2", "text": "14. DIFFERENT ENTITIES WITH THE SAME NAME", "sourceParagraph": 763 }, { "type": "paragraph", "text": "Different entities with the same or similar names may exist. This situation creates entity collisions. Example distinctions:", "sourceParagraphs": [ 764, 765, 766 ] }, { "type": "bullet", "text": "Companies with the same name", "sourceParagraph": 767 }, { "type": "bullet", "text": "Institutions using the same abbreviation", "sourceParagraph": 768 }, { "type": "bullet", "text": "The same brand name in different countries", "sourceParagraph": 769 }, { "type": "bullet", "text": "Old and new company", "sourceParagraph": 770 }, { "type": "bullet", "text": "Brand name with personal name", "sourceParagraph": 771 }, { "type": "bullet", "text": "Company name with product name", "sourceParagraph": 772 }, { "type": "bullet", "text": "Brand with general word", "sourceParagraph": 773 }, { "type": "paragraph", "text": "Audit cannot rely solely on name matching.", "sourceParagraphs": [ 774 ] }, { "type": "heading3", "text": "14.1. Collision Resolver Fields", "sourceParagraph": 776 }, { "type": "paragraph", "text": "Jurisdiction Legal company type Domain name Physical address Establishment date Founder or manager Product and service Brand owner Registration number Local language name", "sourceParagraphs": [ 777, 778, 779, 780, 781, 782, 783, 784, 785, 786 ] }, { "type": "heading3", "text": "14.2. Ambiguous Claim", "sourceParagraph": 788 }, { "type": "paragraph", "text": "If a claim points to multiple reasonable entities, there are three options: The claim is clarified. Classified separately as an entity resolution test. The system is expected to prompt clarification. The answer obtained from the ambiguous claim cannot be silently added to the main accuracy score.", "sourceParagraphs": [ 789, 790, 791, 792, 793 ] }, { "type": "heading2", "text": "15. AN ENTITY, MORE THAN ONE DOMAIN NAME", "sourceParagraph": 795 }, { "type": "paragraph", "text": "A company:", "sourceParagraphs": [ 796 ] }, { "type": "bullet", "text": "global domain,", "sourceParagraph": 797 }, { "type": "bullet", "text": "local country domains,", "sourceParagraph": 798 }, { "type": "bullet", "text": "product domains,", "sourceParagraph": 799 }, { "type": "bullet", "text": "investor domain,", "sourceParagraph": 800 }, { "type": "bullet", "text": "support domain,", "sourceParagraph": 801 }, { "type": "bullet", "text": "old domains", "sourceParagraph": 802 }, { "type": "paragraph", "text": "can use. Not all of these domains automatically fall under the same audit scope. For each domain:", "sourceParagraphs": [ 803, 804, 805 ] }, { "type": "bullet", "text": "controller,", "sourceParagraph": 806 }, { "type": "bullet", "text": "content owner,", "sourceParagraph": 807 }, { "type": "bullet", "text": "target user,", "sourceParagraph": 808 }, { "type": "bullet", "text": "language,", "sourceParagraph": 809 }, { "type": "bullet", "text": "product scope,", "sourceParagraph": 810 }, { "type": "bullet", "text": "activity status", "sourceParagraph": 811 }, { "type": "paragraph", "text": "It must be determined. A correct record in one domain does not automatically correct the error in another domain.", "sourceParagraphs": [ 812, 813 ] }, { "type": "heading2", "text": "16. A DOMAIN NAME, MORE THAN ONE ENTITY", "sourceParagraph": 815 }, { "type": "paragraph", "text": "On a single domain name:", "sourceParagraphs": [ 816 ] }, { "type": "bullet", "text": "parent company,", "sourceParagraph": 817 }, { "type": "bullet", "text": "different brands,", "sourceParagraph": 818 }, { "type": "bullet", "text": "local companies,", "sourceParagraph": 819 }, { "type": "bullet", "text": "products,", "sourceParagraph": 820 }, { "type": "bullet", "text": "partners,", "sourceParagraph": 821 }, { "type": "bullet", "text": "sellers", "sourceParagraph": 822 }, { "type": "paragraph", "text": "It can take place. In this case, entity ownership should be defined based on the URL or content family. Example:", "sourceParagraphs": [ 823, 824, 825 ] }, { "type": "table", "rows": [ [ "Surface", "Existence role" ], [ "Home page", "Global brand" ], [ "Investor relations", "Legal parent company" ], [ "Local price page", "Local sales organisation" ], [ "Marketplace", "Independent sellers" ], [ "Career page", "Specified employer entities" ], [ "Support portal", "Product or service operator" ] ] }, { "type": "paragraph", "text": "A single domain name does not mean single responsibility.", "sourceParagraphs": [ 826 ] }, { "type": "heading2", "text": "17. TEMPORAL ENTITY IDENTITY", "sourceParagraph": 828 }, { "type": "paragraph", "text": "The entity identity may change over time. Case examples:", "sourceParagraphs": [ 829, 830 ] }, { "type": "bullet", "text": "Change of company name", "sourceParagraph": 831 }, { "type": "bullet", "text": "Brand transferred to another company", "sourceParagraph": 832 }, { "type": "bullet", "text": "Merger", "sourceParagraph": 833 }, { "type": "bullet", "text": "Acquisition", "sourceParagraph": 834 }, { "type": "bullet", "text": "Division", "sourceParagraph": 835 }, { "type": "bullet", "text": "Termination of activity", "sourceParagraph": 836 }, { "type": "bullet", "text": "Change of domain name", "sourceParagraph": 837 }, { "type": "bullet", "text": "Transfer of the product to another company", "sourceParagraph": 838 }, { "type": "bullet", "text": "Departure of the founder", "sourceParagraph": 839 }, { "type": "bullet", "text": "Change of local operator", "sourceParagraph": 840 }, { "type": "paragraph", "text": "For this reason, entity registration is time-dependent. It can be considered as E(t).", "sourceParagraphs": [ 841, 842, 843, 844 ] }, { "type": "heading3", "text": "17.1. Historical Identity and Current Identity", "sourceParagraph": 846 }, { "type": "paragraph", "text": "A relationship that was correct in the past cannot be carried over to the present. Example: “X was operated by Y in 2022.” is historical. The following sentence is different: “X is operated by Y.”", "sourceParagraphs": [ 847, 848, 849, 850, 851, 852 ] }, { "type": "heading3", "text": "17.2. Post-Acquisition Feature Transfer", "sourceParagraph": 854 }, { "type": "paragraph", "text": "Acquisition:", "sourceParagraphs": [ 855 ] }, { "type": "bullet", "text": "ownership,", "sourceParagraph": 856 }, { "type": "bullet", "text": "some entities,", "sourceParagraph": 857 }, { "type": "bullet", "text": "may transfer certain contracts", "sourceParagraph": 858 }, { "type": "paragraph", "text": ". However:", "sourceParagraphs": [ 859, 860 ] }, { "type": "bullet", "text": "all historical achievements,", "sourceParagraph": 861 }, { "type": "bullet", "text": "all certificates,", "sourceParagraph": 862 }, { "type": "bullet", "text": "all customer relationships,", "sourceParagraph": 863 }, { "type": "bullet", "text": "all licences", "sourceParagraph": 864 }, { "type": "paragraph", "text": "are not automatically transferred to the new entity. Which features are legally and practically transferred must be explained.", "sourceParagraphs": [ 865, 866 ] }, { "type": "heading3", "text": "17.3. Domain Name History", "sourceParagraph": 868 }, { "type": "paragraph", "text": "A domain name may carry traces of content or references from the previous owner. The new owner is not automatically considered to have gained the previous authority. If the AI system merges the old and new entities, historical entity confusion occurs.", "sourceParagraphs": [ 869, 870, 871 ] }, { "type": "heading2", "text": "18. MULTILINGUAL ENTITY IDENTITY", "sourceParagraph": 873 }, { "type": "paragraph", "text": "The names of an entity in different languages must be linked to the same identity. However, not all translations are official names. The following types should be distinguished:", "sourceParagraphs": [ 874, 875, 876 ] }, { "type": "bullet", "text": "Legal name", "sourceParagraph": 877 }, { "type": "bullet", "text": "Brand name", "sourceParagraph": 878 }, { "type": "bullet", "text": "Official localised name", "sourceParagraph": 879 }, { "type": "bullet", "text": "Transliteration", "sourceParagraph": 880 }, { "type": "bullet", "text": "Descriptive translation", "sourceParagraph": 881 }, { "type": "bullet", "text": "Abbreviation", "sourceParagraph": 882 }, { "type": "bullet", "text": "Former name", "sourceParagraph": 883 }, { "type": "bullet", "text": "Commonly used name", "sourceParagraph": 884 }, { "type": "paragraph", "text": "Example: A Turkish description of a brand does not have to be the legal or official Turkish name of the brand. The machine record should clearly indicate these types.", "sourceParagraphs": [ 885, 886, 887 ] }, { "type": "heading2", "text": "19. COUNTRY AND JURISDICTION CAN CHANGE THE ENTITY BOUNDARY", "sourceParagraph": 889 }, { "type": "paragraph", "text": "Under a global brand, there may be different legal or operational entities in different countries. When a user asks, “Does Company X operate in Germany?” the answer should distinguish:", "sourceParagraphs": [ 890, 891, 892, 893 ] }, { "type": "bullet", "text": "the existence of the global brand,", "sourceParagraph": 894 }, { "type": "bullet", "text": "the local legal entity,", "sourceParagraph": 895 }, { "type": "bullet", "text": "the distributor relationship,", "sourceParagraph": 896 }, { "type": "bullet", "text": "the capacity to provide remote services", "sourceParagraph": 897 }, { "type": "paragraph", "text": "“Brand is recognised in the country” is not the same as: “Local legal company provides services.”", "sourceParagraphs": [ 898, 899, 900, 901 ] }, { "type": "heading3", "text": "19.1. Country-Based Entity Roles", "sourceParagraph": 903 }, { "type": "paragraph", "text": "For each country, to the extent applicable:", "sourceParagraphs": [ 904 ] }, { "type": "bullet", "text": "Legal seller", "sourceParagraph": 905 }, { "type": "bullet", "text": "Invoice issuer", "sourceParagraph": 906 }, { "type": "bullet", "text": "Data controller", "sourceParagraph": 907 }, { "type": "bullet", "text": "Service provider", "sourceParagraph": 908 }, { "type": "bullet", "text": "Supporting", "sourceParagraph": 909 }, { "type": "bullet", "text": "Distributor", "sourceParagraph": 910 }, { "type": "bullet", "text": "Franchise", "sourceParagraph": 911 }, { "type": "bullet", "text": "Licence holder", "sourceParagraph": 912 }, { "type": "bullet", "text": "Brand representative", "sourceParagraph": 913 }, { "type": "paragraph", "text": "must be determined.", "sourceParagraphs": [ 914 ] }, { "type": "heading3", "text": "19.2. General Language Page Is Not Proof of Local Company", "sourceParagraph": 916 }, { "type": "paragraph", "text": "Having a website in German does not prove that there is a legal entity or a physical office in Germany. An English page does not create global legal jurisdiction. Language is not a jurisdiction.", "sourceParagraphs": [ 917, 918, 919 ] }, { "type": "heading2", "text": "20. SYNTHETIC APPLE.COM PRESENCE INDICATION", "sourceParagraph": 921 }, { "type": "paragraph", "text": "SYNTHETIC METHODOLOGY DEMONSTRATION / The following example has been prepared solely to demonstrate the entity resolution method. It is not the actual audit of Apple Inc., a current corporate structure review, or the real result of any AI product. Prompt: “What kind of company is Apple.com?”", "sourceParagraphs": [ 922, 923, 924 ] }, { "type": "heading3", "text": "20.1. Prompt Anchor", "sourceParagraph": 926 }, { "type": "paragraph", "text": "apple.com", "sourceParagraphs": [ 927 ] }, { "type": "heading3", "text": "20.2. Candidate Targets", "sourceParagraph": 929 }, { "type": "paragraph", "text": "Domain name and website Apple brand Canonical corporate entity Online store function Product family If the target is not defined before the audit begins, adjudicators may score different objects.", "sourceParagraphs": [ 930, 931, 932, 933, 934, 935 ] }, { "type": "heading3", "text": "20.3. Proposed Target Entity Lock", "sourceParagraph": 937 }, { "type": "paragraph", "text": "The target for synthetic testing can be defined as follows: “the main corporate technology entity associated with the canonical domain name apple.com, representing the company's identity and core activities.” This definition:", "sourceParagraphs": [ 938, 939, 940 ] }, { "type": "bullet", "text": "does not automatically encompass the technical specifications of individual products,", "sourceParagraph": 941 }, { "type": "bullet", "text": "the local company of each country,", "sourceParagraph": 942 }, { "type": "bullet", "text": "store prices,", "sourceParagraph": 943 }, { "type": "bullet", "text": "investment advice", "sourceParagraph": 944 }, { "type": "paragraph", "text": "automatically.", "sourceParagraphs": [ 945 ] }, { "type": "heading3", "text": "20.4. Synthetic Response A", "sourceParagraph": 947 }, { "type": "paragraph", "text": "“Apple.com is a global corporate website associated with technology products and digital services.” Evaluation: It does not fully match the institution with the domain name. It may leave the company identity incomplete. The main activity could generally be correct. There may be a lack of material identity.", "sourceParagraphs": [ 948, 949, 950, 951, 952, 953 ] }, { "type": "heading3", "text": "20.5. Synthetic Response B", "sourceParagraph": 955 }, { "type": "paragraph", "text": "“Apple.com is the official website of the company called iPhone.” Evaluation: The product and the company are combined. The target entity is incorrectly resolved. There is an identity error.", "sourceParagraphs": [ 956, 957, 958, 959, 960 ] }, { "type": "heading3", "text": "20.6. Synthetic Response C", "sourceParagraph": 962 }, { "type": "paragraph", "text": "“Apple.com is only an independent store that sells phones online.” Evaluation: The function of the domain name is excessively narrowed. The corporate entity and scope of activities are distorted. The independent store characteristic may be incorrect.", "sourceParagraphs": [ 963, 964, 965, 966, 967 ] }, { "type": "heading3", "text": "20.7. Synthetic Response D", "sourceParagraph": 969 }, { "type": "paragraph", "text": "“Apple.com is one of the main corporate and product information surfaces of the Apple brand; the legal entity operating the site and the local sales side should also be determined according to the country.” Assessment: Domain name, brand, and local legal role are separated. The target entity analysis is more disciplined. Other material claims of the response should be proven separately. This example shows:", "sourceParagraphs": [ 970, 971, 972, 973, 974, 975 ] }, { "type": "callout", "text": "A response can centre on the wrong entity even if it contains many correct words.", "sourceParagraph": 976 }, { "type": "paragraph", "text": "Identity accuracy comes before content accuracy.", "sourceParagraphs": [ 977 ] }, { "type": "heading2", "text": "21. SYNTHETIC COMPLEX ENTITY CASE", "sourceParagraph": 979 }, { "type": "heading3", "text": "SYNTHETIC CASE — NOT A REAL INSTITUTION", "sourceParagraph": 980 }, { "type": "paragraph", "text": "Let's consider a global brand named Asteron. Structure:", "sourceParagraphs": [ 981, 982 ] }, { "type": "bullet", "text": "Asteron Holdings Ltd. — parent company", "sourceParagraph": 983 }, { "type": "bullet", "text": "Asteron Travel — brand", "sourceParagraph": 984 }, { "type": "bullet", "text": "Asteron Turkey Tourism Ltd. — local company", "sourceParagraph": 985 }, { "type": "bullet", "text": "Mira Destination Services — independent local operations partner", "sourceParagraph": 986 }, { "type": "bullet", "text": "Asteron Club Antalya — franchise operation", "sourceParagraph": 987 }, { "type": "bullet", "text": "asteron.example — global domain name", "sourceParagraph": 988 }, { "type": "bullet", "text": "tr.asteron.example — Turkish marketing surface", "sourceParagraph": 989 }, { "type": "paragraph", "text": "User asks: \"What services does Asteron provide in Turkey?\" AI response: \"Asteron Holdings Ltd. operates all tours in Antalya itself, has 400 employees in Turkey, and Asteron Club Antalya is directly responsible for customer satisfaction in Antalya.\" Reality package: The parent company owns the brand. The local company is the sales and contracting side. Mira delivers part of the operations. Asteron Club Antalya is an independent franchise. The 400 employees are the total of the group. The franchise is directly responsible for customer satisfaction at the business level. Errors in the response: The parent company is presented as responsible for service delivery. The total number of group employees is conveyed as if it were the capacity of the local company. The franchise's responsibility is directly assigned to the parent company. The brand, legal entity, and operational partner are combined. The response may have named the brand correctly. However, the entity graph is incorrect.", "sourceParagraphs": [ 990, 991, 992, 993, 994, 995, 996, 997, 998, 999, 1000, 1001, 1002, 1003, 1004, 1005, 1006, 1007 ] }, { "type": "paragraph", "text": "Therefore, the GEO score cannot rely solely on the mention of the name or the general fluency of the sentence.", "sourceParagraphs": [ 1008 ] }, { "type": "heading2", "text": "22. ENTITY RESOLUTION RESULTS", "sourceParagraph": 1010 }, { "type": "paragraph", "text": "For each observation, the entity resolution result can be recorded with one of the following statuses.", "sourceParagraphs": [ 1011 ] }, { "type": "heading3", "text": "ER-0 — NOT EVALUATED", "sourceParagraph": 1013 }, { "type": "paragraph", "text": "The entity resolution has not been evaluated.", "sourceParagraphs": [ 1014 ] }, { "type": "heading3", "text": "ER-1 — EXACT TARGET", "sourceParagraph": 1016 }, { "type": "paragraph", "text": "The target entity has been resolved correctly and clearly.", "sourceParagraphs": [ 1017 ] }, { "type": "heading3", "text": "ER-2 — ACCEPTABLE BRAND-LEVEL MATCH", "sourceParagraph": 1019 }, { "type": "paragraph", "text": "No legal name is provided, but the targeted brand and corporate entity are materially resolved correctly.", "sourceParagraphs": [ 1020 ] }, { "type": "heading3", "text": "ER-3 — PARTIAL OR UNDER-SPECIFIED", "sourceParagraph": 1022 }, { "type": "paragraph", "text": "The response approached the correct area but lacks legal, geographical, or operational identity.", "sourceParagraphs": [ 1023 ] }, { "type": "heading3", "text": "ER-4 — MULTI-ENTITY CONFLATION", "sourceParagraph": 1025 }, { "type": "paragraph", "text": "Multiple separate entities have been merged as a single object.", "sourceParagraphs": [ 1026 ] }, { "type": "heading3", "text": "ER-5 — WRONG ENTITY", "sourceParagraph": 1028 }, { "type": "paragraph", "text": "Another person, company, product, or domain name has been targeted.", "sourceParagraphs": [ 1029 ] }, { "type": "heading3", "text": "ER-6 — AMBIGUOUS", "sourceParagraph": 1031 }, { "type": "paragraph", "text": "It cannot be determined from the response which entity has been resolved.", "sourceParagraphs": [ 1032 ] }, { "type": "heading3", "text": "ER-7 — APPROPRIATE CLARIFICATION", "sourceParagraph": 1034 }, { "type": "paragraph", "text": "The prompt is indeed ambiguous, and the system has correctly asked for clarification. Requesting a clarification does not have to be considered a failure.", "sourceParagraphs": [ 1035, 1036 ] }, { "type": "heading2", "text": "23. ENTITY RESOLUTION SUCCESS IS DISTINCT FROM REPRESENTATION ACCURACY", "sourceParagraph": 1038 }, { "type": "paragraph", "text": "A system can analyse the correct entity but may provide incorrect information about it. Another system may state some material facts correctly but connect them to the wrong entity. There are four basic situations:", "sourceParagraphs": [ 1039, 1040, 1041 ] }, { "type": "table", "rows": [ [ "Entity analysis", "Material content", "Result" ], [ "Correct", "Correct", "Strong correct representation candidate" ], [ "Correct", "Incorrect", "Correct object, incorrect attribute" ], [ "Incorrect", "Partially correct", "Correct information linked to the wrong entity" ], [ "Incorrect", "Incorrect", "Complete representation breakdown" ] ] }, { "type": "paragraph", "text": "For this reason, identity analysis may be a separate gateway. In particular, an answer with serious confusion with another entity cannot be salvaged with a high factual accuracy average.", "sourceParagraphs": [ 1042, 1043 ] }, { "type": "heading2", "text": "24. LINKING THE CONFORMITY DECISION TO THE ENTITY", "sourceParagraph": 1045 }, { "type": "paragraph", "text": "A conformity decision should be established in the following format:", "sourceParagraphs": [ 1046 ] }, { "type": "callout", "text": "[Entity ID] + [entity type] + [scope] + [digital surfaces] + [country and languages] + [time] + [standard version]", "sourceParagraph": 1047 }, { "type": "paragraph", "text": "Example template: “For [brand/legal entity] identified with Entity ID NGE-XXXX NOMOS; the public pages in Turkish and English on the specified domain, within the specified product and service scope, have been evaluated on [date] under [standard version].” Then it should be clearly stated: “This decision does not automatically apply to subsidiaries, franchises, out-of-scope products, other domains, and languages to be added in the future.”", "sourceParagraphs": [ 1048, 1049, 1050, 1051 ] }, { "type": "heading2", "text": "25. MANDATORY NORMATIVE PROVISIONS", "sourceParagraph": 1053 }, { "type": "protocol_label", "text": "CH03-N01", "sourceParagraph": 1054 }, { "type": "paragraph", "text": "Every GEO audit must identify a unique target entity before results are seen.", "sourceParagraphs": [ 1055 ] }, { "type": "protocol_label", "text": "CH03-N02", "sourceParagraph": 1056 }, { "type": "paragraph", "text": "The prompt anchor cannot be assumed to be the same as the target entity.", "sourceParagraphs": [ 1057 ] }, { "type": "protocol_label", "text": "CH03-N03", "sourceParagraph": 1058 }, { "type": "paragraph", "text": "Domain names, trademarks, legal entities, products, services, and personal identities must be distinguished from each other.", "sourceParagraphs": [ 1059 ] }, { "type": "protocol_label", "text": "CH03-N04", "sourceParagraph": 1060 }, { "type": "paragraph", "text": "A property of an entity cannot be transferred to another related entity without clear evidence and scope.", "sourceParagraphs": [ 1061 ] }, { "type": "protocol_label", "text": "CH03-N05", "sourceParagraph": 1062 }, { "type": "paragraph", "text": "The parent company's licence, certificate, number of employees, customers, or conformity outcome cannot be automatically applied to subsidiaries.", "sourceParagraphs": [ 1063 ] }, { "type": "protocol_label", "text": "CH03-N06", "sourceParagraph": 1064 }, { "type": "paragraph", "text": "The claim at the brand level and the party that actually delivers the service and is legally responsible must be separated.", "sourceParagraphs": [ 1065 ] }, { "type": "protocol_label", "text": "CH03-N07", "sourceParagraph": 1066 }, { "type": "paragraph", "text": "The franchisor, franchise operator, distributor, dealer, and platform seller must be registered as separate entities.", "sourceParagraphs": [ 1067 ] }, { "type": "protocol_label", "text": "CH03-N08", "sourceParagraph": 1068 }, { "type": "paragraph", "text": "Each tangible entity relationship must carry information on direction, scope, and time.", "sourceParagraphs": [ 1069 ] }, { "type": "protocol_label", "text": "CH03-N09", "sourceParagraph": 1070 }, { "type": "paragraph", "text": "Relationships such as \"partner,\" \"customer,\" \"expert,\" \"group company,\" and similar cannot be used ambiguously.", "sourceParagraphs": [ 1071 ] }, { "type": "protocol_label", "text": "CH03-N10", "sourceParagraph": 1072 }, { "type": "paragraph", "text": "Collision resolution areas must be used for different entities with the same name.", "sourceParagraphs": [ 1073 ] }, { "type": "protocol_label", "text": "CH03-N11", "sourceParagraph": 1074 }, { "type": "paragraph", "text": "If an uncertain claim is to be included in the main score, the uncertainty method must be predefined; otherwise, the claim must be clarified.", "sourceParagraphs": [ 1075 ] }, { "type": "protocol_label", "text": "CH03-N12", "sourceParagraph": 1076 }, { "type": "paragraph", "text": "The scope of an entity cannot be expanded or narrowed after the results are seen to increase a positive outcome.", "sourceParagraphs": [ 1077 ] }, { "type": "protocol_label", "text": "CH03-N13", "sourceParagraph": 1078 }, { "type": "paragraph", "text": "Historical entities and relationships cannot be used as if they have the current entity status.", "sourceParagraphs": [ 1079 ] }, { "type": "protocol_label", "text": "CH03-N14", "sourceParagraph": 1080 }, { "type": "paragraph", "text": "Purchase, merger, or brand transfer does not automatically transfer past characteristics to the new entity.", "sourceParagraphs": [ 1081 ] }, { "type": "protocol_label", "text": "CH03-N15", "sourceParagraph": 1082 }, { "type": "paragraph", "text": "A conformity decision can only be applied to the conformity unit specified in the decision.", "sourceParagraphs": [ 1083 ] }, { "type": "protocol_label", "text": "CH03-N16", "sourceParagraph": 1084 }, { "type": "paragraph", "text": "The new product cannot automatically inherit the current availability of language, domain name, affiliated company, or country.", "sourceParagraphs": [ 1085 ] }, { "type": "protocol_label", "text": "CH03-N17", "sourceParagraph": 1086 }, { "type": "paragraph", "text": "The success of entity resolution and factual accuracy should be evaluated separately.", "sourceParagraphs": [ 1087 ] }, { "type": "protocol_label", "text": "CH03-N18", "sourceParagraph": 1088 }, { "type": "paragraph", "text": "The merging of material identity with the wrong entity cannot be made invisible within the general factual average.", "sourceParagraphs": [ 1089 ] }, { "type": "protocol_label", "text": "CH03-N19", "sourceParagraph": 1090 }, { "type": "paragraph", "text": "Canonical Entity Records should be versioned and preserve changes over time.", "sourceParagraphs": [ 1091 ] }, { "type": "protocol_label", "text": "CH03-N20", "sourceParagraph": 1092 }, { "type": "paragraph", "text": "There should be a responsible human or institutional owner of the entity record.", "sourceParagraphs": [ 1093 ] }, { "type": "heading2", "text": "26. FORMS OF FAILURE", "sourceParagraph": 1095 }, { "type": "protocol_label", "text": "CH03-F01 — CONSIDER DOMAIN NAME AS A COMPANY", "sourceParagraph": 1096 }, { "type": "paragraph", "text": "The digital address is presented as a legal or operational entity.", "sourceParagraphs": [ 1097 ] }, { "type": "protocol_label", "text": "CH03-F02 — MERGING THE BRAND WITH THE LEGAL COMPANY", "sourceParagraph": 1098 }, { "type": "paragraph", "text": "The brand name is used in place of the legal entity that carries contracts and responsibilities.", "sourceParagraphs": [ 1099 ] }, { "type": "protocol_label", "text": "CH03-F03 — CONSIDERING THE PRODUCT AS A COMPANY", "sourceParagraph": 1100 }, { "type": "paragraph", "text": "The product name is resolved as a company or brand identity.", "sourceParagraphs": [ 1101 ] }, { "type": "protocol_label", "text": "CH03-F04 — TRANSFERRING THE FOUNDER'S CHARACTERISTIC TO THE COMPANY", "sourceParagraph": 1102 }, { "type": "paragraph", "text": "The founder's age, experience, award, or previous project is made a corporate characteristic.", "sourceParagraphs": [ 1103 ] }, { "type": "protocol_label", "text": "CH03-F05 — TRANSFERRING GROUP CHARACTERISTICS TO THE LOCAL COMPANY", "sourceParagraph": 1104 }, { "type": "paragraph", "text": "Consolidated customer, employee, or revenue data is assigned to the local business.", "sourceParagraphs": [ 1105 ] }, { "type": "protocol_label", "text": "CH03-F06 — TRANSFERRING THE LICENSE THROUGH RELATIONSHIP", "sourceParagraph": 1106 }, { "type": "paragraph", "text": "The parent company, product, employee, or partner licence is used in the name of another entity.", "sourceParagraphs": [ 1107 ] }, { "type": "protocol_label", "text": "CH03-F07 — UNIFYING FRANCHISE AND THE MAIN BRAND", "sourceParagraph": 1108 }, { "type": "paragraph", "text": "Local business performance and responsibility are presented directly as the global brand.", "sourceParagraphs": [ 1109 ] }, { "type": "protocol_label", "text": "CH03-F08 — COMBINING PRODUCER AND SELLER", "sourceParagraph": 1110 }, { "type": "paragraph", "text": "The product manufacturer and local seller or distributor are considered a single entity.", "sourceParagraphs": [ 1111 ] }, { "type": "protocol_label", "text": "CH03-F09 — COMBINING PLATFORM AND SELLER", "sourceParagraph": 1112 }, { "type": "paragraph", "text": "The marketplace platform is made responsible for the seller's product, warranty, or delivery.", "sourceParagraphs": [ 1113 ] }, { "type": "protocol_label", "text": "CH03-F10 — CONSIDERING THE RELATIONSHIP AS FEATURE TRANSFER", "sourceParagraph": 1114 }, { "type": "paragraph", "text": "Customer status, membership, technology use or partnership is misrepresented as a transfer of authority.", "sourceParagraphs": [ 1115 ] }, { "type": "protocol_label", "text": "CH03-F11 — UPDATING HISTORICAL IDENTITY", "sourceParagraph": 1116 }, { "type": "paragraph", "text": "The old name, ownership, partnership, or manager is used as if it were the current reality.", "sourceParagraphs": [ 1117 ] }, { "type": "protocol_label", "text": "CH03-F12 — CLAIMING PAST AFTER ACQUISITION", "sourceParagraph": 1118 }, { "type": "paragraph", "text": "All past successes of the acquired entity are attributed to the new owner without explanation.", "sourceParagraphs": [ 1119 ] }, { "type": "protocol_label", "text": "CH03-F13 — MERGING ENTITIES WITH THE SAME NAME", "sourceParagraph": 1120 }, { "type": "paragraph", "text": "Different companies or individuals are resolved as a single entity due to name similarity.", "sourceParagraphs": [ 1121 ] }, { "type": "protocol_label", "text": "CH03-F14 — ARTIFICIALLY SPLITTING AN ENTITY", "sourceParagraph": 1122 }, { "type": "paragraph", "text": "A failed language, product, or local operation is declared a separate entity to maintain a positive score.", "sourceParagraphs": [ 1123 ] }, { "type": "protocol_label", "text": "CH03-F15 — ARTIFICIALLY EXPANDING AN ENTITY", "sourceParagraph": 1124 }, { "type": "paragraph", "text": "Successful brand or product results are spread across the whole group.", "sourceParagraphs": [ 1125 ] }, { "type": "protocol_label", "text": "CH03-F16 — CONSIDERING SCOPE OF ELIGIBILITY", "sourceParagraph": 1126 }, { "type": "paragraph", "text": "A narrow audit result is used for the entire company, group, or product portfolio.", "sourceParagraphs": [ 1127 ] }, { "type": "protocol_label", "text": "CH03-F17 — COUNTING LANGUAGE JURISDICTION", "sourceParagraph": 1128 }, { "type": "paragraph", "text": "The existence of a language version is used as proof of legal company or service capacity in that country.", "sourceParagraphs": [ 1129 ] }, { "type": "protocol_label", "text": "CH03-F18 — COUNTING TECHNICAL CONTROL AS OWNERSHIP", "sourceParagraph": 1130 }, { "type": "paragraph", "text": "The agency or developer is shown as the legal or commercial owner of the institution because they manage the domain name.", "sourceParagraphs": [ 1131 ] }, { "type": "protocol_label", "text": "CH03-F19 — LEAVING CLAIMANT UNDETERMINED", "sourceParagraph": 1132 }, { "type": "paragraph", "text": "It cannot be determined who is responsible for incorrect price, licence, or capacity information.", "sourceParagraphs": [ 1133 ] }, { "type": "protocol_label", "text": "CH03-F20 — INTERPRETING THE PROMPT ANCHOR AFTER THE RESULT", "sourceParagraph": 1134 }, { "type": "paragraph", "text": "The target entity that will make the response appear successful is selected afterward.", "sourceParagraphs": [ 1135 ] }, { "type": "heading2", "text": "27. AUDIT PROCEDURE", "sourceParagraph": 1137 }, { "type": "heading3", "text": "Step 1 — Record the Prompt Anchor", "sourceParagraph": 1138 }, { "type": "paragraph", "text": "Exactly which of the user's:", "sourceParagraphs": [ 1139 ] }, { "type": "bullet", "text": "name,", "sourceParagraph": 1140 }, { "type": "bullet", "text": "domain name,", "sourceParagraph": 1141 }, { "type": "bullet", "text": "brand,", "sourceParagraph": 1142 }, { "type": "bullet", "text": "product,", "sourceParagraph": 1143 }, { "type": "bullet", "text": "person", "sourceParagraph": 1144 }, { "type": "paragraph", "text": "expression is used is fully recorded.", "sourceParagraphs": [ 1145 ] }, { "type": "heading3", "text": "Step 2 — Identify the Target Entity", "sourceParagraph": 1147 }, { "type": "paragraph", "text": "It is noted which entity the audit actually measures. The target is classified as:", "sourceParagraphs": [ 1148, 1149 ] }, { "type": "bullet", "text": "legal entity,", "sourceParagraph": 1150 }, { "type": "bullet", "text": "brand,", "sourceParagraph": 1151 }, { "type": "bullet", "text": "product,", "sourceParagraph": 1152 }, { "type": "bullet", "text": "service,", "sourceParagraph": 1153 }, { "type": "bullet", "text": "entity package", "sourceParagraph": 1154 }, { "type": "paragraph", "text": "The applicable class must be recorded explicitly.", "sourceParagraphs": [ 1155 ] }, { "type": "heading3", "text": "Step 3 — Create Canonical Entity ID", "sourceParagraph": 1157 }, { "type": "paragraph", "text": "Canonical name Entity type Legal identity Jurisdiction domain Aliases Time status is recorded.", "sourceParagraphs": [ 1158, 1159, 1160, 1161, 1162, 1163, 1164 ] }, { "type": "heading3", "text": "Step 4 — Build the Entity Graph", "sourceParagraph": 1166 }, { "type": "paragraph", "text": "Directed relationships between parent company, subsidiaries, brand, product, domain name, individuals, and partners are extracted.", "sourceParagraphs": [ 1167 ] }, { "type": "heading3", "text": "Step 5 — Determine Responsibility Roles", "sourceParagraph": 1169 }, { "type": "paragraph", "text": "Claim owner Digital surface owner Service delivery owner Legal responsible Evidence owner are separated.", "sourceParagraphs": [ 1170, 1171, 1172, 1173, 1174, 1175 ] }, { "type": "heading3", "text": "Step 6 — Review Attribute Transfer Risk", "sourceParagraph": 1177 }, { "type": "paragraph", "text": "In response:", "sourceParagraphs": [ 1178 ] }, { "type": "bullet", "text": "licence,", "sourceParagraph": 1179 }, { "type": "bullet", "text": "certificate,", "sourceParagraph": 1180 }, { "type": "bullet", "text": "employee,", "sourceParagraph": 1181 }, { "type": "bullet", "text": "customer,", "sourceParagraph": 1182 }, { "type": "bullet", "text": "establishment date,", "sourceParagraph": 1183 }, { "type": "bullet", "text": "success,", "sourceParagraph": 1184 }, { "type": "bullet", "text": "compliance", "sourceParagraph": 1185 }, { "type": "paragraph", "text": "has it been transferred to the wrong entity?", "sourceParagraphs": [ 1186 ] }, { "type": "heading3", "text": "Step 7 — Check Temporal Identity", "sourceParagraph": 1188 }, { "type": "paragraph", "text": "Answer:", "sourceParagraphs": [ 1189 ] }, { "type": "bullet", "text": "former name,", "sourceParagraph": 1190 }, { "type": "bullet", "text": "former owner,", "sourceParagraph": 1191 }, { "type": "bullet", "text": "terminated partnership,", "sourceParagraph": 1192 }, { "type": "bullet", "text": "former employee,", "sourceParagraph": 1193 }, { "type": "bullet", "text": "pre-acquisition status", "sourceParagraph": 1194 }, { "type": "paragraph", "text": "is it being used?", "sourceParagraphs": [ 1195 ] }, { "type": "heading3", "text": "Step 8 — Check Geographic and Local Roles", "sourceParagraph": 1197 }, { "type": "paragraph", "text": "Are global brand, local company, distributor, and franchise roles correct?", "sourceParagraphs": [ 1198 ] }, { "type": "heading3", "text": "Step 9 — Classify Entity Resolved by System", "sourceParagraph": 1200 }, { "type": "paragraph", "text": "Appropriate status between ER-1 and ER-7 is assigned.", "sourceParagraphs": [ 1201 ] }, { "type": "heading3", "text": "Step 10 — Lock Conformity Unit", "sourceParagraph": 1203 }, { "type": "paragraph", "text": "To apply the final decision:", "sourceParagraphs": [ 1204 ] }, { "type": "bullet", "text": "entity,", "sourceParagraph": 1205 }, { "type": "bullet", "text": "field,", "sourceParagraph": 1206 }, { "type": "bullet", "text": "product,", "sourceParagraph": 1207 }, { "type": "bullet", "text": "language,", "sourceParagraph": 1208 }, { "type": "bullet", "text": "country,", "sourceParagraph": 1209 }, { "type": "bullet", "text": "time", "sourceParagraph": 1210 }, { "type": "paragraph", "text": "boundary is written.", "sourceParagraphs": [ 1211 ] }, { "type": "heading3", "text": "Step 11 — Limit the Public Claim", "sourceParagraph": 1213 }, { "type": "paragraph", "text": "The result is narrowed if it is generalised to an institution or group broader than the defined entity.", "sourceParagraphs": [ 1214 ] }, { "type": "heading3", "text": "Step 12 — Record the Way of Change and Objection", "sourceParagraph": 1216 }, { "type": "paragraph", "text": "If an identity or relationship error is reported:", "sourceParagraphs": [ 1217 ] }, { "type": "bullet", "text": "who will verify,", "sourceParagraph": 1218 }, { "type": "bullet", "text": "who will change,", "sourceParagraph": 1219 }, { "type": "bullet", "text": "which version will be generated", "sourceParagraph": 1220 }, { "type": "paragraph", "text": "is determined.", "sourceParagraphs": [ 1221 ] }, { "type": "heading2", "text": "28. NECESSARY EVIDENCE", "sourceParagraph": 1223 }, { "type": "paragraph", "text": "Canonical entity identity; legal name and entity type; official record or equivalent verification; trade mark ownership; domain-name control; parent and subsidiary relationships; product and service ownership; local-business and distributor records; franchise relationships; contracting and invoicing party; responsibility for service delivery; licence and certificate holder; founder and executive roles; active and historical names; acquisition, merger and transfer records; country and jurisdiction; official-language and transliteration records; entity graph; claimant; content owner; technical-surface owner; audit scope; out-of-scope entities and surfaces; entity-record version; last verification date; and accountable person or organisation.", "sourceParagraphs": [ 1224, 1225, 1226, 1227, 1228, 1229, 1230, 1231, 1232, 1233, 1234, 1235, 1236, 1237, 1238, 1239, 1240, 1241, 1242, 1243, 1244, 1245, 1246, 1247, 1248, 1249 ] }, { "type": "heading2", "text": "29. AUDIT CHECKLIST", "sourceParagraph": 1251 }, { "type": "paragraph", "text": "Has the prompt anchor been recorded clearly? Was the target entity identified before the results were observed? Were the legal entity and domain name separated? Were the brand and brand owner separated? Were the product and company separated? Were the founder's attributes separated from those of the organisation? Are parent–subsidiary relationships correct? Are franchises and distributors distinct? Are the parties selling and delivering the service identified? Were the claimant and technical publisher separated? Do the licences and certificates belong to the correct entity? Are employee, customer and revenue figures stated within the correct scope? Were historical names and relationships separated from the current state? Were different entities with the same name checked? Were multiple domains and user surfaces linked to the correct entity?", "sourceParagraphs": [ 1252, 1253, 1254, 1255, 1256, 1257, 1258, 1259, 1260, 1261, 1262, 1263, 1264, 1265, 1266 ] }, { "type": "paragraph", "text": "Have multilingual names been linked to the same canonical identity? Are the country and jurisdiction roles clear? Has the entity been classified separately as a result of entity resolution? Is the conformity unit clearly defined? Does it extend to entities outside the decision scope? Is the entity record versioned and dated? Is there a path for objection and correction for identity changes?", "sourceParagraphs": [ 1267, 1268, 1269, 1270, 1271, 1272, 1273 ] }, { "type": "heading2", "text": "30. OBJECTIONS AND ANSWERS", "sourceParagraph": 1275 }, { "type": "heading3", "text": "Objection 1 — “The user knows the brand name, not the legal company.”", "sourceParagraph": 1276 }, { "type": "paragraph", "text": "Correct. It is natural for the user to use the brand in the prompt. The audit does not force the user to use legal jargon. However, the evaluation in the background:", "sourceParagraphs": [ 1277, 1278, 1279, 1280 ] }, { "type": "bullet", "text": "the brand,", "sourceParagraph": 1281 }, { "type": "bullet", "text": "owner,", "sourceParagraph": 1282 }, { "type": "bullet", "text": "service provider,", "sourceParagraph": 1283 }, { "type": "bullet", "text": "contracting party", "sourceParagraph": 1284 }, { "type": "paragraph", "text": "must distinguish. A simple answer can be given to the user. The audit log does not have to be simple.", "sourceParagraphs": [ 1285, 1286, 1287 ] }, { "type": "heading3", "text": "Objection 2 — “In a small business, the brand and the company are already the same.”", "sourceParagraph": 1289 }, { "type": "paragraph", "text": "In some small structures:", "sourceParagraphs": [ 1290 ] }, { "type": "bullet", "text": "single brand,", "sourceParagraph": 1291 }, { "type": "bullet", "text": "single domain name,", "sourceParagraph": 1292 }, { "type": "bullet", "text": "sole business owner,", "sourceParagraph": 1293 }, { "type": "bullet", "text": "single service provider", "sourceParagraph": 1294 }, { "type": "paragraph", "text": "It can be found. In this case, the entity graph is simple. Nonetheless, the record must be made explicitly. Simplicity does not mean ambiguity.", "sourceParagraphs": [ 1295, 1296, 1297, 1298 ] }, { "type": "heading3", "text": "Objection 3 — 'The group of companies acts like a single brand.'", "sourceParagraph": 1300 }, { "type": "paragraph", "text": "From a marketing perspective, there may be a single brand. In terms of legal, operational, and evidential responsibility, there may be separate entities. If decisions are to be made at the group level, it should be clearly shown which features are consolidated.", "sourceParagraphs": [ 1301, 1302, 1303 ] }, { "type": "heading3", "text": "Objection 4 — “So much detail unnecessarily complicates the GEO score.”", "sourceParagraph": 1305 }, { "type": "paragraph", "text": "The correct score given to the wrong entity is not useful. Entity resolution is not an additional burden for auditing, but a condition of validity.", "sourceParagraphs": [ 1306, 1307 ] }, { "type": "heading3", "text": "Objection 5 — \"AI systems already do not pay attention to these legal differences.\"", "sourceParagraph": 1309 }, { "type": "paragraph", "text": "Exactly for this reason, it should be measured. The AI system's:", "sourceParagraphs": [ 1310, 1311 ] }, { "type": "bullet", "text": "brand,", "sourceParagraph": 1312 }, { "type": "bullet", "text": "company,", "sourceParagraph": 1313 }, { "type": "bullet", "text": "product,", "sourceParagraph": 1314 }, { "type": "bullet", "text": "local business", "sourceParagraph": 1315 }, { "type": "paragraph", "text": "Missing the difference between them is the representation error itself.", "sourceParagraphs": [ 1316 ] }, { "type": "heading3", "text": "Objection 6 — “If the parent company controls the whole group, characteristics can be transferred.”", "sourceParagraph": 1318 }, { "type": "paragraph", "text": "Control relationships are important in some areas. However:", "sourceParagraphs": [ 1319, 1320 ] }, { "type": "bullet", "text": "licence,", "sourceParagraph": 1321 }, { "type": "bullet", "text": "contracts,", "sourceParagraph": 1322 }, { "type": "bullet", "text": "customer,", "sourceParagraph": 1323 }, { "type": "bullet", "text": "employee,", "sourceParagraph": 1324 }, { "type": "bullet", "text": "data responsibility,", "sourceParagraph": 1325 }, { "type": "bullet", "text": "compliance", "sourceParagraph": 1326 }, { "type": "paragraph", "text": "such characteristics require separate evidence for transfer. Control is not the unification of all characteristics.", "sourceParagraphs": [ 1327, 1328 ] }, { "type": "heading3", "text": "Objection 7 — “Are the achievements of the old company after an acquisition not part of the new group?”", "sourceParagraph": 1330 }, { "type": "paragraph", "text": "It can be explained in a historical context. However:", "sourceParagraphs": [ 1331, 1332 ] }, { "type": "bullet", "text": "during which period was the success,", "sourceParagraph": 1333 }, { "type": "bullet", "text": "by which team and entity,", "sourceParagraph": 1334 }, { "type": "bullet", "text": "before or after the acquisition", "sourceParagraph": 1335 }, { "type": "paragraph", "text": "It should be stated when it was produced. The new ownership does not change the history.", "sourceParagraphs": [ 1336, 1337 ] }, { "type": "heading3", "text": "Objection 8 — \"the prompt is already clear; the user has typed the domain name.\"", "sourceParagraph": 1339 }, { "type": "paragraph", "text": "The domain name may be clear. However: the question \"What does the domain name represent?\" should also be answered. A domain can be a company, store, product portal, or multi-entity platform.", "sourceParagraphs": [ 1340, 1341, 1342, 1343, 1344 ] }, { "type": "heading3", "text": "Objection 9 — \"Describing the local distributor like a brand is easier for the user.\"", "sourceParagraph": 1346 }, { "type": "paragraph", "text": "Simplicity does not override responsibility. The response can be simplified as follows: \"The brand provides services in Turkey through an authorised local distributor.\" It is both understandable and correct.", "sourceParagraphs": [ 1347, 1348, 1349, 1350 ] }, { "type": "heading3", "text": "Objection 10 — \"All these relationships change over time.\"", "sourceParagraph": 1352 }, { "type": "paragraph", "text": "That is correct. For this reason, the entity record:", "sourceParagraphs": [ 1353, 1354 ] }, { "type": "bullet", "text": "dated,", "sourceParagraph": 1355 }, { "type": "bullet", "text": "versioned,", "sourceParagraph": 1356 }, { "type": "bullet", "text": "should be change-triggered", "sourceParagraph": 1357 }, { "type": "paragraph", "text": "It does not reduce the need for variability recording. It increases it.", "sourceParagraphs": [ 1358, 1359, 1360 ] }, { "type": "heading2", "text": "COMMON RULE OF CHAPTER 33", "sourceParagraph": 1485 }, { "type": "paragraph", "text": "Before generating a GEO score: we ask, “How many answers are correct?” But the more fundamental question that should be asked before this is:", "sourceParagraphs": [ 1486, 1487, 1488, 1489 ] }, { "type": "heading4", "text": "“Exactly whose answers are these about?”", "sourceParagraph": 1490 }, { "type": "paragraph", "text": "If we consider a domain name as a company, we measure the wrong object. If we consider the brand as a legal company, we assign responsibility incorrectly. If we consider the product as a company, we distort its features. If we turn the founder's experience into the company's history, we exaggerate the past. If we count the group's number of employees as local capacity, we inflate service expectations. If the parent company transfers its licence to a franchise, we direct the user to the wrong authority. If we make the past achievements after an acquisition a feature of the new owner, we rewrite the historical truth. If we extend a narrow domain name audit to the entire company, we whitewash the scope of compliance. Therefore, defining an entity is not just data cleansing. It is the beginning of representational responsibility. A system can describe the wrong entity with the correct information. This is still a misrepresentation. A system can describe the correct entity with the wrong information. This is also a misrepresentation.", "sourceParagraphs": [ 1491, 1492, 1493, 1494, 1495, 1496, 1497, 1498, 1499, 1500, 1501, 1502, 1503, 1504 ] }, { "type": "paragraph", "text": "Both are required for the correct GEO:", "sourceParagraphs": [ 1505 ] }, { "type": "callout", "text": "Correct entity. / Correct attribute. / Correct relationship. / Correct scope. / Correct time.", "sourceParagraph": 1506 }, { "type": "paragraph", "text": "Therefore, NOMOS's third measurement law is:", "sourceParagraphs": [ 1507 ] }, { "type": "callout", "text": "If the measured entity is not clear, the measured accuracy is not clear either.", "sourceParagraph": 1508 }, { "type": "paragraph", "text": "The fourth law is as follows:", "sourceParagraphs": [ 1509 ] }, { "type": "callout", "text": "Relationship is not identity.", "sourceParagraph": 1510 }, { "type": "paragraph", "text": "The fifth law is as follows:", "sourceParagraphs": [ 1511 ] }, { "type": "callout", "text": "Ownership is not the transfer of all attributes.", "sourceParagraph": 1512 }, { "type": "paragraph", "text": "Its sixth law is:", "sourceParagraphs": [ 1513 ] }, { "type": "callout", "text": "Conformity does not spread to neighbouring entities.", "sourceParagraph": 1514 }, { "type": "heading2", "text": "NOMOS Section 3 Order", "sourceParagraph": 1516 }, { "type": "callout", "text": "Do not give me a name alone. / Show me which entity that name belongs to.", "sourceParagraph": 1517 }, { "type": "callout", "text": "Do not treat a domain as a company, a brand as a contracting party or a product as a manufacturer.", "sourceParagraph": 1518 }, { "type": "callout", "text": "Do not use the founder's past as the age of the institution.", "sourceParagraph": 1519 }, { "type": "callout", "text": "Do not transfer the parent company's licence to the affiliated organisation, franchise, or distributor.", "sourceParagraph": 1520 }, { "type": "callout", "text": "Do not make the group's customer the local brand's customer, or the brand's award the award of all products.", "sourceParagraph": 1521 }, { "type": "callout", "text": "Do not describe collaboration as ownership, membership as accreditation, or technology use as partnership.", "sourceParagraph": 1522 }, { "type": "callout", "text": "Do not merge the old name with the new entity, or the acquired past with today's performance.", "sourceParagraph": 1523 }, { "type": "callout", "text": "Do not expand the entity to magnify positive results / Do not split the entity to hide negative results.", "sourceParagraph": 1524 }, { "type": "callout", "text": "First, lock the target. / Then record the direction of every relationship. / Then attach each attribute to its true owner. / Then add the relevant time and country. / Only then decide for that entity alone.", "sourceParagraph": 1525 }, { "type": "heading2", "text": "The Chapter's Closing Sentence", "sourceParagraph": 1527 }, { "type": "callout", "text": "The first question in a GEO audit is not what the AI says, but precisely who or what each statement concerns.", "sourceParagraph": 1528 }, { "type": "heading2", "text": "Normative Core" }, { "type": "normative", "text": "Every GEO audit MUST define and version, before observing results: - a prompt anchor, - a target entity, - an entity type, - a canonical identity record, - a relationship graph, - a temporal and jurisdictional boundary, - an audit object, - and a conformity unit. Domains, brands, legal entities, corporate groups, subsidiaries, products, services, persons, franchises, distributors, platforms, and sellers MUST NOT be treated as interchangeable entities. Attributes, credentials, licences, customers, employees, performance, authority, and conformity MUST NOT transfer across related entities unless the relationship, scope, time, and evidence explicitly support that transfer. The entity resolved by the AI product MUST be evaluated separately from the factual accuracy of the response. A conformity decision MUST apply only to the explicitly identified entity and audited scope. It MUST NOT automatically extend to affiliates, products, jurisdictions, languages, domains, future versions, or related entities.", "translationRoute": "controlled-normative-rule" } ] }, { "number": 4, "id": "NOMOS-GEO-AUDIT-CH04", "title": "Two Layers of Reality", "subtitle": "Official representation and verified entity reality", "sourceFile": "4.cü bölüm.docx", "sourceSha256": "6CACCD9333A032B34AB7BD6773B7C81D57D11794D4AB8787DA1FD42415E80401", "sourceWordCount": 6604, "sourceIds": [ "K10", "K11" ], "blocks": [ { "type": "heading2", "text": "Chapter Boundary", "sourceParagraph": 17 }, { "type": "paragraph", "text": "Section 1 established the following provision:", "sourceParagraphs": [ 18 ] }, { "type": "callout", "text": "A single response is not a GEO score.", "sourceParagraph": 19 }, { "type": "paragraph", "text": "Chapter 2 added the following:", "sourceParagraphs": [ 20 ] }, { "type": "callout", "text": "Representation is a distribution that occurs among users, AI products, countries, languages, sessions, and times.", "sourceParagraph": 21 }, { "type": "paragraph", "text": "Chapter 3 then identified the object of the distribution:", "sourceParagraphs": [ 22 ] }, { "type": "callout", "text": "An entity cannot be measured until it is resolved.", "sourceParagraph": 23 }, { "type": "paragraph", "text": "Now we come to the next mandatory question:", "sourceParagraphs": [ 24 ] }, { "type": "callout", "text": "According to which reality record will we evaluate whether the AI response is correct?", "sourceParagraph": 25 }, { "type": "paragraph", "text": "Which source should govern the audit: the audited entity's website, a legal record, the company's pricing page, independent reporting, the licensing authority, customer reviews, the company's own performance data or a competitor's claim? When sources conflict, which should be accepted as correct? If a company publishes a false or unsupported claim and the AI repeats it perfectly, has the system succeeded? If the official website is outdated but the AI gives a different answer based on current, verified external evidence, has the system failed? Until these questions are answered:", "sourceParagraphs": [ 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37 ] }, { "type": "bullet", "text": "correct representation,", "sourceParagraph": 38 }, { "type": "bullet", "text": "incorrect representation,", "sourceParagraph": 39 }, { "type": "bullet", "text": "source fidelity,", "sourceParagraph": 40 }, { "type": "bullet", "text": "factual accuracy,", "sourceParagraph": 41 }, { "type": "bullet", "text": "entity reality", "sourceParagraph": 42 }, { "type": "paragraph", "text": "These terms are sometimes used as though they referred to the same concept. This chapter resolves the resulting confusion by distinguishing two layers of reality:", "sourceParagraphs": [ 43, 44 ] }, { "type": "heading4", "text": "First Layer", "sourceParagraph": 45 }, { "type": "heading4", "text": "Official Representation Layer", "sourceParagraph": 46 }, { "type": "paragraph", "text": "Shows what the entity says about itself on the canonical surfaces it controls.", "sourceParagraphs": [ 47 ] }, { "type": "heading4", "text": "Second Layer", "sourceParagraph": 48 }, { "type": "heading4", "text": "Verified Entity Reality Layer", "sourceParagraph": 49 }, { "type": "paragraph", "text": "Under the existing evidence, it shows which material claims are supported within what scope and time. These two layers are connected. But they are not the same. The first layer answers the question: “How does the entity define itself?” The second layer answers the question: “Which part of this definition is supported by which evidence?” The first book defines the correct standard, the second book separates the forms of violation, and the audit system is designed to measure and prove these distinctions. This section operates that audit task at the claim level.", "sourceParagraphs": [ 50, 51, 52, 53, 54, 55, 56, 57, 58, 59 ] }, { "type": "heading2", "text": "NOMOS Challenge", "sourceParagraph": 61 }, { "type": "paragraph", "text": "Your company's website contains the following sentence: “We are the world's leading GEO company.” An AI product reads this page and responds to the user: “This company is the world's leading GEO company.” Does the response correctly reflect your website? Yes. Does the response correctly reflect proven reality?", "sourceParagraphs": [ 62, 63, 64, 65, 66, 67, 68 ] }, { "type": "paragraph", "text": "We do not know this yet. Maybe:", "sourceParagraphs": [ 69, 70 ] }, { "type": "bullet", "text": "there is comparative research,", "sourceParagraph": 71 }, { "type": "bullet", "text": "market scope has been defined,", "sourceParagraph": 72 }, { "type": "bullet", "text": "a leadership criterion has been published,", "sourceParagraph": 73 }, { "type": "paragraph", "text": "the result has been independently verified. Maybe none of these exist. If AI had conveyed the company's own claim: “The company defines itself as the world's leading GEO company.”, it would have preserved the source of the claim. However, when it said: “It is the world's leading GEO company.”, the company may have elevated the statement to independent reality.", "sourceParagraphs": [ 74, 75, 76, 77, 78, 79, 80, 81 ] }, { "type": "paragraph", "text": "Consider another case. The company's official website states, ‘We provide services only in Türkiye.’ Yet verified records show that the company began operating in Germany and the United Kingdom six months ago, and the website has not been updated. Drawing on current legal and operational records, the AI responds: ‘The company provides services in Türkiye, Germany and the United Kingdom.’", "sourceParagraphs": [ 82, 83, 84, 85, 86, 87, 88, 89 ] }, { "type": "paragraph", "text": "The AI is not fully faithful to the official website. Could it nevertheless be closer to verified current reality? Yes. Consider a third case. The company's website states, ‘We operate in 20 countries.’ Independent and internal records verify only eight. The AI responds: ‘The company operates in 20 countries.’", "sourceParagraphs": [ 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101 ] }, { "type": "paragraph", "text": "This response:", "sourceParagraphs": [ 102, 103 ] }, { "type": "bullet", "text": "faithful to official representation,", "sourceParagraph": 104 }, { "type": "bullet", "text": "contrary to verified reality", "sourceParagraph": 105 }, { "type": "paragraph", "text": "It is possible. The first provision of this section is as follows:", "sourceParagraphs": [ 106, 107 ] }, { "type": "callout", "text": "Fidelity to the official record is not the same as factual accuracy.", "sourceParagraph": 108 }, { "type": "paragraph", "text": "Its second provision states:", "sourceParagraphs": [ 109 ] }, { "type": "callout", "text": "What a company says and what can be proven should be recorded separately.", "sourceParagraph": 110 }, { "type": "paragraph", "text": "Its third provision states:", "sourceParagraphs": [ 111 ] }, { "type": "callout", "text": "An AI repeating an official mistake correctly does not constitute an accurate representation.", "sourceParagraph": 112 }, { "type": "paragraph", "text": "Its fourth provision states:", "sourceParagraphs": [ 113 ] }, { "type": "callout", "text": "An AI diverging from an outdated official record is not automatically wrong if it carries reliable evidence.", "sourceParagraph": 114 }, { "type": "heading2", "text": "1. PURPOSE OF THE CHAPTER", "sourceParagraph": 116 }, { "type": "paragraph", "text": "The purpose of this section is to establish two separate reference layers to be used in the GEO-1000 evaluation. The section makes the following distinctions normative:", "sourceParagraphs": [ 117, 118 ] }, { "type": "bullet", "text": "Declaration of existence vs. proven fact", "sourceParagraph": 119 }, { "type": "bullet", "text": "Official source vs. independent source", "sourceParagraph": 120 }, { "type": "bullet", "text": "First-party source versus independent verification", "sourceParagraph": 121 }, { "type": "bullet", "text": "Fidelity to the source versus material accuracy", "sourceParagraph": 122 }, { "type": "bullet", "text": "Repetition of the claim versus verification of the claim", "sourceParagraph": 123 }, { "type": "bullet", "text": "Current official record versus old official record", "sourceParagraph": 124 }, { "type": "bullet", "text": "Official silence versus externally verified information", "sourceParagraph": 125 }, { "type": "bullet", "text": "Company perspective versus external reality judgement", "sourceParagraph": 126 }, { "type": "bullet", "text": "Official evidence versus official claim", "sourceParagraph": 127 }, { "type": "bullet", "text": "Independence of the source versus relevance of the source to the subject", "sourceParagraph": 128 }, { "type": "bullet", "text": "Official error versus AI error", "sourceParagraph": 129 }, { "type": "bullet", "text": "Reference conflict versus model failure", "sourceParagraph": 130 }, { "type": "bullet", "text": "Uncertainty versus failure", "sourceParagraph": 131 }, { "type": "bullet", "text": "Claimant's interest versus truth of the claim", "sourceParagraph": 132 }, { "type": "bullet", "text": "The presence of evidence and whether the evidence actually supports the claim", "sourceParagraph": 133 }, { "type": "paragraph", "text": "By the end of this section, two separate questions should be able to be asked for each AI response:", "sourceParagraphs": [ 134 ] }, { "type": "callout", "text": "Does the response correctly convey the official and canonical representation of the entity?", "sourceParagraph": 135 }, { "type": "paragraph", "text": "and:", "sourceParagraphs": [ 136 ] }, { "type": "callout", "text": "Does the response correctly reflect the material reality carried by the currently verifiable evidence?", "sourceParagraph": 137 }, { "type": "paragraph", "text": "These questions may yield the same answer. They may also yield different answers. When they differ, the role of the audit is not to conceal the difference, but to explain it.", "sourceParagraphs": [ 138, 139, 140 ] }, { "type": "heading2", "text": "2. CENTRAL NORMATIVE PROVISION", "sourceParagraph": 142 }, { "type": "callout", "text": "Each material AI response should be assessed against two separate references without transformation: the Official Representation Record and the Verified Entity Reality Record.", "sourceParagraph": 143 }, { "type": "paragraph", "text": "A response:", "sourceParagraphs": [ 144 ] }, { "type": "bullet", "text": "faithful to official representation,", "sourceParagraph": 145 }, { "type": "bullet", "text": "appropriate to verified reality,", "sourceParagraph": 146 }, { "type": "bullet", "text": "appropriate to both,", "sourceParagraph": 147 }, { "type": "bullet", "text": "distant from both", "sourceParagraph": 148 }, { "type": "paragraph", "text": "It can be. NOMOS cannot quietly combine these two results into a single true–false decision. High performance in the first layer does not compensate for failure in the second layer. Specifically:", "sourceParagraphs": [ 149, 150, 151, 152 ] }, { "type": "callout", "text": "An AI response that correctly repeats an unverified or false claim published by the entity itself can demonstrate official representation fidelity; however, it cannot demonstrate verified entity accuracy.", "sourceParagraph": 153 }, { "type": "heading2", "text": "3. FIRST LAYER", "sourceParagraph": 155 }, { "type": "heading3", "text": "OFFICIAL REPRESENTATION LAYER", "sourceParagraph": 156 }, { "type": "paragraph", "text": "The Official Representation Layer is the versioned record of the current material statements it publishes about itself on canonical surfaces controlled or explicitly authorised by the entity. This layer answers the question:", "sourceParagraphs": [ 157, 158 ] }, { "type": "callout", "text": "How does the entity want to be understood, and what information does it officially publish?", "sourceParagraph": 159 }, { "type": "paragraph", "text": "This layer includes:", "sourceParagraphs": [ 160 ] }, { "type": "bullet", "text": "legal identity statement,", "sourceParagraph": 161 }, { "type": "bullet", "text": "brand–company relationship,", "sourceParagraph": 162 }, { "type": "bullet", "text": "product and service definitions,", "sourceParagraph": 163 }, { "type": "bullet", "text": "published prices,", "sourceParagraph": 164 }, { "type": "bullet", "text": "service countries,", "sourceParagraph": 165 }, { "type": "bullet", "text": "support conditions,", "sourceParagraph": 166 }, { "type": "bullet", "text": "official policies,", "sourceParagraph": 167 }, { "type": "bullet", "text": "publicly disclosed boundaries,", "sourceParagraph": 168 }, { "type": "bullet", "text": "It may contain official currency and version information", "sourceParagraph": 169 }, { "type": "paragraph", "text": "The official representation layer is important. Because an AI system must correctly understand how an entity:", "sourceParagraphs": [ 170, 171, 172 ] }, { "type": "bullet", "text": "defines itself,", "sourceParagraph": 173 }, { "type": "bullet", "text": "what service it offers,", "sourceParagraph": 174 }, { "type": "bullet", "text": "what price it publishes,", "sourceParagraph": 175 }, { "type": "bullet", "text": "what terms it accepts", "sourceParagraph": 176 }, { "type": "paragraph", "text": "However, the official representation layer is not automatic independent proof for every claim. A statement on the institution's own website, especially regarding claims to the outside world such as superiority and performance, presents the statement itself; it does not prove its correctness on its own.", "sourceParagraphs": [ 177, 178, 179 ] }, { "type": "heading2", "text": "4. WHAT IS AN OFFICIAL SOURCE?", "sourceParagraph": 181 }, { "type": "paragraph", "text": "A source is not considered official merely because it carries the institution's logo. An official source should have at least the relevant features of the following:", "sourceParagraphs": [ 182, 183 ] }, { "type": "bullet", "text": "Controlled by the entity", "sourceParagraph": 184 }, { "type": "bullet", "text": "Undergoes an authorised publishing process", "sourceParagraph": 185 }, { "type": "bullet", "text": "Approved by the owner of the factual information", "sourceParagraph": 186 }, { "type": "bullet", "text": "Canonical URL or record identifier", "sourceParagraph": 187 }, { "type": "bullet", "text": "Version", "sourceParagraph": 188 }, { "type": "bullet", "text": "Publication or verification date", "sourceParagraph": 189 }, { "type": "bullet", "text": "Change history", "sourceParagraph": 190 }, { "type": "bullet", "text": "Responsible person or institution", "sourceParagraph": 191 }, { "type": "bullet", "text": "Scope", "sourceParagraph": 192 }, { "type": "bullet", "text": "Language and locale status", "sourceParagraph": 193 }, { "type": "bullet", "text": "Archival or currency status", "sourceParagraph": 194 }, { "type": "paragraph", "text": "A view published on a personal social media account by an employee is not automatically an official company statement. A draft file by an agency is not an official publication. An email from a sales representative, if not authorised, is not a general corporate policy. An old PDF on a company's website may not replace the new official record.", "sourceParagraphs": [ 195, 196, 197, 198 ] }, { "type": "heading2", "text": "5. OFFICIAL REPRESENTATION SURFACES", "sourceParagraph": 200 }, { "type": "paragraph", "text": "According to the relevant audit, official surfaces may include:", "sourceParagraphs": [ 201 ] }, { "type": "bullet", "text": "Main corporate website", "sourceParagraph": 202 }, { "type": "bullet", "text": "Legal disclaimers", "sourceParagraph": 203 }, { "type": "bullet", "text": "Product and service pages", "sourceParagraph": 204 }, { "type": "bullet", "text": "Pricing and quotation pages", "sourceParagraph": 205 }, { "type": "bullet", "text": "Official help centre", "sourceParagraph": 206 }, { "type": "bullet", "text": "Investor relations documents", "sourceParagraph": 207 }, { "type": "bullet", "text": "Public policies and contracts", "sourceParagraph": 208 }, { "type": "bullet", "text": "Official press releases", "sourceParagraph": 209 }, { "type": "bullet", "text": "Structured data", "sourceParagraph": 210 }, { "type": "bullet", "text": "Canonical JSON records", "sourceParagraph": 211 }, { "type": "bullet", "text": "Official API outputs", "sourceParagraph": 212 }, { "type": "bullet", "text": "Official in-app content", "sourceParagraph": 213 }, { "type": "bullet", "text": "Authorised local sites", "sourceParagraph": 214 }, { "type": "bullet", "text": "Official social media accounts", "sourceParagraph": 215 }, { "type": "bullet", "text": "Statements submitted to public institutions", "sourceParagraph": 216 }, { "type": "bullet", "text": "Release and retraction records", "sourceParagraph": 217 }, { "type": "paragraph", "text": "These surfaces may contradict each other. In this case, “official representation” is not a single, clean reality. It is a set of records that must be audited internally.", "sourceParagraphs": [ 218, 219, 220 ] }, { "type": "heading2", "text": "6. OFFICIAL REPRESENTATION RECORD", "sourceParagraph": 222 }, { "type": "paragraph", "text": "An Official Representation Record must be created for each audited entity. This record is not a copy of all official content. It is a traceable summary of factual claims and boundaries. Each record should include the following fields to the extent applicable:", "sourceParagraphs": [ 223, 224, 225, 226 ] }, { "type": "bullet", "text": "Claim identifier", "sourceParagraph": 227 }, { "type": "bullet", "text": "Full text of the claim", "sourceParagraph": 228 }, { "type": "bullet", "text": "Subject of the claim", "sourceParagraph": 229 }, { "type": "bullet", "text": "Type of claim", "sourceParagraph": 230 }, { "type": "bullet", "text": "Canonical source", "sourceParagraph": 231 }, { "type": "bullet", "text": "Source owner", "sourceParagraph": 232 }, { "type": "bullet", "text": "Language and locale", "sourceParagraph": 233 }, { "type": "bullet", "text": "Publication date", "sourceParagraph": 234 }, { "type": "bullet", "text": "Last verification date", "sourceParagraph": 235 }, { "type": "bullet", "text": "Validity period", "sourceParagraph": 236 }, { "type": "bullet", "text": "Official status", "sourceParagraph": 237 }, { "type": "bullet", "text": "Scope", "sourceParagraph": 238 }, { "type": "bullet", "text": "Limitations", "sourceParagraph": 239 }, { "type": "bullet", "text": "Previous version", "sourceParagraph": 240 }, { "type": "bullet", "text": "Reason for change", "sourceParagraph": 241 }, { "type": "bullet", "text": "Claimant", "sourceParagraph": 242 }, { "type": "bullet", "text": "Human approval", "sourceParagraph": 243 }, { "type": "paragraph", "text": "Without an official representation record of an entity: the sentence “AI reflects the site accurately.” cannot be clearly tested.", "sourceParagraphs": [ 244, 245, 246 ] }, { "type": "heading2", "text": "7. THE AUTHORITY OF THE OFFICIAL SOURCE DEPENDS ON THE TYPE OF CLAIM", "sourceParagraph": 248 }, { "type": "paragraph", "text": "Not all official sources carry the same weight for every claim. Source authority should be evaluated according to the type of claim.", "sourceParagraphs": [ 249, 250 ] }, { "type": "heading3", "text": "7.1. Decisions Under the Entity's Own Control", "sourceParagraph": 252 }, { "type": "paragraph", "text": "Example:", "sourceParagraphs": [ 253 ] }, { "type": "bullet", "text": "Which service it offers to the public", "sourceParagraph": 254 }, { "type": "bullet", "text": "Published price", "sourceParagraph": 255 }, { "type": "bullet", "text": "Support hours", "sourceParagraph": 256 }, { "type": "bullet", "text": "Official return policy", "sourceParagraph": 257 }, { "type": "bullet", "text": "Communication channel", "sourceParagraph": 258 }, { "type": "bullet", "text": "Publicly available product version", "sourceParagraph": 259 }, { "type": "paragraph", "text": "In these areas, the official source is fundamental and often the primary authorised record. However, if the official page contradicts the actual process flow, the page alone is not sufficient. For example:", "sourceParagraphs": [ 260, 261, 262 ] }, { "type": "bullet", "text": "price page 1,000 euros,", "sourceParagraph": 263 }, { "type": "bullet", "text": "checkout 1,500 euros", "sourceParagraph": 264 }, { "type": "paragraph", "text": "If it is showing, there is a contradiction in the official representation layer.", "sourceParagraphs": [ 265 ] }, { "type": "heading3", "text": "7.2. Legal Identity and Authority", "sourceParagraph": 267 }, { "type": "paragraph", "text": "Example:", "sourceParagraphs": [ 268 ] }, { "type": "bullet", "text": "Company registration", "sourceParagraph": 269 }, { "type": "bullet", "text": "Licence", "sourceParagraph": 270 }, { "type": "bullet", "text": "Professional authority", "sourceParagraph": 271 }, { "type": "bullet", "text": "Trademark ownership", "sourceParagraph": 272 }, { "type": "bullet", "text": "Tax or legal status", "sourceParagraph": 273 }, { "type": "paragraph", "text": "The company website can disclose this information. However, an official record or an authorised institution can provide stronger verification.", "sourceParagraphs": [ 274, 275 ] }, { "type": "heading3", "text": "7.3. Performance and Success Claims", "sourceParagraph": 277 }, { "type": "paragraph", "text": "Example:", "sourceParagraphs": [ 278 ] }, { "type": "bullet", "text": "“98% of our projects are successful.”", "sourceParagraph": 279 }, { "type": "bullet", "text": "“We increase our customers' revenue by an average of 40%.”", "sourceParagraph": 280 }, { "type": "bullet", "text": "“We have the highest satisfaction rate.”", "sourceParagraph": 281 }, { "type": "paragraph", "text": "The company website may be the owner of the claim. But to support the claim:", "sourceParagraphs": [ 282, 283 ] }, { "type": "bullet", "text": "data,", "sourceParagraph": 284 }, { "type": "bullet", "text": "method,", "sourceParagraph": 285 }, { "type": "bullet", "text": "denominator,", "sourceParagraph": 286 }, { "type": "bullet", "text": "period,", "sourceParagraph": 287 }, { "type": "bullet", "text": "unsuccessful results,", "sourceParagraph": 288 }, { "type": "bullet", "text": "independent review if possible", "sourceParagraph": 289 }, { "type": "paragraph", "text": "are required.", "sourceParagraphs": [ 290 ] }, { "type": "heading3", "text": "7.4. Superiority and Leadership Claims", "sourceParagraph": 292 }, { "type": "paragraph", "text": "Example:", "sourceParagraphs": [ 293 ] }, { "type": "bullet", "text": "\"World leader\"", "sourceParagraph": 294 }, { "type": "bullet", "text": "\"The best\"", "sourceParagraph": 295 }, { "type": "bullet", "text": "\"Number one\"", "sourceParagraph": 296 }, { "type": "bullet", "text": "“The only real standard”", "sourceParagraph": 297 }, { "type": "bullet", "text": "“The most trusted company in the industry”", "sourceParagraph": 298 }, { "type": "paragraph", "text": "A site’s own statement is not an independent comparison. Such claims:", "sourceParagraphs": [ 299, 300 ] }, { "type": "bullet", "text": "require a comparison universe,", "sourceParagraph": 301 }, { "type": "bullet", "text": "criteria,", "sourceParagraph": 302 }, { "type": "bullet", "text": "method,", "sourceParagraph": 303 }, { "type": "bullet", "text": "date,", "sourceParagraph": 304 }, { "type": "bullet", "text": "scope,", "sourceParagraph": 305 }, { "type": "bullet", "text": "independence", "sourceParagraph": 306 }, { "type": "paragraph", "text": "Both conditions must be established.", "sourceParagraphs": [ 307 ] }, { "type": "heading3", "text": "7.5. Third Party Relationships", "sourceParagraph": 309 }, { "type": "paragraph", "text": "Example:", "sourceParagraphs": [ 310 ] }, { "type": "bullet", "text": "Customer", "sourceParagraph": 311 }, { "type": "bullet", "text": "Partner", "sourceParagraph": 312 }, { "type": "bullet", "text": "Sponsor", "sourceParagraph": 313 }, { "type": "bullet", "text": "Membership", "sourceParagraph": 314 }, { "type": "bullet", "text": "Accreditation", "sourceParagraph": 315 }, { "type": "bullet", "text": "Technology partnership", "sourceParagraph": 316 }, { "type": "bullet", "text": "Distributorship", "sourceParagraph": 317 }, { "type": "paragraph", "text": "A unilateral official statement can be valuable. However, it may need to be verified by the other party to the relationship or by a reliable record.", "sourceParagraphs": [ 318, 319 ] }, { "type": "heading3", "text": "7.6. User Outcomes", "sourceParagraph": 321 }, { "type": "paragraph", "text": "Example:", "sourceParagraphs": [ 322 ] }, { "type": "bullet", "text": "Customer satisfaction", "sourceParagraph": 323 }, { "type": "bullet", "text": "Actual usage outcome", "sourceParagraph": 324 }, { "type": "bullet", "text": "Return rate", "sourceParagraph": 325 }, { "type": "bullet", "text": "Complaint rate", "sourceParagraph": 326 }, { "type": "bullet", "text": "Repeat purchase", "sourceParagraph": 327 }, { "type": "paragraph", "text": "Company data in these areas is important first-party evidence. However:", "sourceParagraphs": [ 328, 329 ] }, { "type": "bullet", "text": "data selection,", "sourceParagraph": 330 }, { "type": "bullet", "text": "nonresponse,", "sourceParagraph": 331 }, { "type": "bullet", "text": "methodology,", "sourceParagraph": 332 }, { "type": "bullet", "text": "conflict of interest", "sourceParagraph": 333 }, { "type": "paragraph", "text": "should be disclosed.", "sourceParagraphs": [ 334 ] }, { "type": "heading2", "text": "8. OFFICIAL representation fidelity", "sourceParagraph": 336 }, { "type": "paragraph", "text": "The concept that indicates to what extent an AI response conforms to the Official Representation Layer:", "sourceParagraphs": [ 337 ] }, { "type": "heading4", "text": "Official Representation Fidelity", "sourceParagraph": 338 }, { "type": "paragraph", "text": "The canonical English term is:", "sourceParagraphs": [ 339, 340 ] }, { "type": "heading4", "text": "Official Representation Fidelity", "sourceParagraph": 341 }, { "type": "paragraph", "text": "This measurement asks the following questions: Is the AI interpreting the entity correctly? Does it convey official products and services accurately? Does it distort the official price? Does it remove official boundaries? Does it expand the official country coverage? Does it present the company's opinion as an independent fact? Does it use outdated official information as current? Does it hide contradictions among official records? Does it add material claims not found in the official source? Does it maintain source quality? Official representation fidelity is a measurement of how accurately the entity conveys the reality it has published. This measurement:", "sourceParagraphs": [ 342, 343, 344, 345, 346, 347, 348, 349, 350, 351, 352, 353, 354 ] }, { "type": "callout", "text": "It shows not that the entity's statements are true, but the extent to which AI accurately conveys what the entity says.", "sourceParagraph": 355 }, { "type": "heading2", "text": "9. SECOND LAYER", "sourceParagraph": 357 }, { "type": "heading3", "text": "VERIFIED ENTITY REALITY LAYER", "sourceParagraph": 358 }, { "type": "paragraph", "text": "The Verified Entity Reality Layer is a versioned assessment record of material information that existed on a specific measurement date and is supported by evidence appropriate to the type of claim. This layer answers the question:", "sourceParagraphs": [ 359, 360 ] }, { "type": "callout", "text": "Which statements about the entity are truly supported by which evidence?", "sourceParagraph": 361 }, { "type": "paragraph", "text": "This layer includes:", "sourceParagraphs": [ 362 ] }, { "type": "bullet", "text": "records coming from the company itself,", "sourceParagraph": 363 }, { "type": "bullet", "text": "official public records,", "sourceParagraph": 364 }, { "type": "bullet", "text": "independent sources,", "sourceParagraph": 365 }, { "type": "bullet", "text": "contract or transaction records,", "sourceParagraph": 366 }, { "type": "bullet", "text": "licence documents,", "sourceParagraph": 367 }, { "type": "bullet", "text": "verified historical records,", "sourceParagraph": 368 }, { "type": "bullet", "text": "counter-evidence,", "sourceParagraph": 369 }, { "type": "bullet", "text": "adverse findings,", "sourceParagraph": 370 }, { "type": "bullet", "text": "uncertainties,", "sourceParagraph": 371 }, { "type": "bullet", "text": "remaining disputes", "sourceParagraph": 372 }, { "type": "paragraph", "text": "can be evaluated together.", "sourceParagraphs": [ 373 ] }, { "type": "heading2", "text": "10. \"VERIFIED REALITY\" IS NOT ABSOLUTE REALITY", "sourceParagraph": 375 }, { "type": "paragraph", "text": "This protocol does not claim: \"We know with certainty all the absolute truth about an institution.\" Verified Entity Reality is:", "sourceParagraphs": [ 376, 377, 378, 379 ] }, { "type": "bullet", "text": "evidence-based,", "sourceParagraph": 380 }, { "type": "bullet", "text": "time-dependent,", "sourceParagraph": 381 }, { "type": "bullet", "text": "scope-dependent,", "sourceParagraph": 382 }, { "type": "bullet", "text": "subject to change with new evidence,", "sourceParagraph": 383 }, { "type": "bullet", "text": "can be contested,", "sourceParagraph": 384 }, { "type": "bullet", "text": "can be withdrawn,", "sourceParagraph": 385 }, { "type": "paragraph", "text": "may carry uncertainty. The correct definition is:", "sourceParagraphs": [ 386, 387 ] }, { "type": "callout", "text": "Within the specified date and scope, the strongest and verifiable entity record carried by the available evidence.", "sourceParagraph": 388 }, { "type": "paragraph", "text": "Therefore, a verified reality record:", "sourceParagraphs": [ 389 ] }, { "type": "bullet", "text": "versioned,", "sourceParagraph": 390 }, { "type": "bullet", "text": "sourced,", "sourceParagraph": 391 }, { "type": "bullet", "text": "subject to objection,", "sourceParagraph": 392 }, { "type": "bullet", "text": "modifiable,", "sourceParagraph": 393 }, { "type": "bullet", "text": "aware of uncertainty", "sourceParagraph": 394 }, { "type": "paragraph", "text": "must be. NOMOS cannot position itself as the owner of the truth. It only classifies what sentence the evidence permits and to what extent.", "sourceParagraphs": [ 395, 396, 397 ] }, { "type": "heading2", "text": "11. VERIFIED ENTITY REALITY RECORD", "sourceParagraph": 399 }, { "type": "paragraph", "text": "The full technical structure of this layer will be detailed in a later chapter of the book. At the foundational level, the following fields should exist for each factual claim:", "sourceParagraphs": [ 400, 401 ] }, { "type": "bullet", "text": "Claim identifier", "sourceParagraph": 402 }, { "type": "bullet", "text": "Subject entity", "sourceParagraph": 403 }, { "type": "bullet", "text": "Relation or attribute", "sourceParagraph": 404 }, { "type": "bullet", "text": "Value", "sourceParagraph": 405 }, { "type": "bullet", "text": "Scope", "sourceParagraph": 406 }, { "type": "bullet", "text": "Country or jurisdiction", "sourceParagraph": 407 }, { "type": "bullet", "text": "Time", "sourceParagraph": 408 }, { "type": "bullet", "text": "Sources", "sourceParagraph": 409 }, { "type": "bullet", "text": "Source types", "sourceParagraph": 410 }, { "type": "bullet", "text": "Source verification", "sourceParagraph": 411 }, { "type": "bullet", "text": "Independence", "sourceParagraph": 412 }, { "type": "bullet", "text": "Counter-evidence", "sourceParagraph": 413 }, { "type": "bullet", "text": "Sufficiency of evidence", "sourceParagraph": 414 }, { "type": "bullet", "text": "Verification status", "sourceParagraph": 415 }, { "type": "bullet", "text": "Uncertainty", "sourceParagraph": 416 }, { "type": "bullet", "text": "Adjudicator decision", "sourceParagraph": 417 }, { "type": "bullet", "text": "Appeal status", "sourceParagraph": 418 }, { "type": "bullet", "text": "Last verification date", "sourceParagraph": 419 }, { "type": "bullet", "text": "Record version", "sourceParagraph": 420 }, { "type": "paragraph", "text": "This record cannot be created solely from positive evidence. Records that narrow the reality or contradict the claim must also be preserved.", "sourceParagraphs": [ 421, 422 ] }, { "type": "heading2", "text": "12. VERIFICATION STATUSES", "sourceParagraph": 424 }, { "type": "paragraph", "text": "Each material claim can be classified with one of the following statuses:", "sourceParagraphs": [ 425 ] }, { "type": "heading3", "text": "VR-0 — NOT ASSESSED", "sourceParagraph": 427 }, { "type": "paragraph", "text": "The claim has not yet been evaluated.", "sourceParagraphs": [ 428 ] }, { "type": "heading3", "text": "VR-1 — OFFICIAL CLAIM ONLY", "sourceParagraph": 430 }, { "type": "paragraph", "text": "The claim is based solely on the institution's own official statement. It has not been proven false. There is no independent or sufficient additional verification.", "sourceParagraphs": [ 431, 432, 433 ] }, { "type": "heading3", "text": "VR-2 — FIRST-PARTY EVIDENCE SUPPORTED", "sourceParagraph": 435 }, { "type": "paragraph", "text": "The claim is supported by the institution's:", "sourceParagraphs": [ 436 ] }, { "type": "bullet", "text": "transaction,", "sourceParagraph": 437 }, { "type": "bullet", "text": "operation,", "sourceParagraph": 438 }, { "type": "bullet", "text": "contracts,", "sourceParagraph": 439 }, { "type": "bullet", "text": "data,", "sourceParagraph": 440 }, { "type": "bullet", "text": "record", "sourceParagraph": 441 }, { "type": "paragraph", "text": "is supported by evidence. The source is first-party.", "sourceParagraphs": [ 442, 443 ] }, { "type": "heading3", "text": "VR-3 — CORROBORATED", "sourceParagraph": 445 }, { "type": "paragraph", "text": "The claim is supported by multiple appropriate and consistent records. Not all sources have to be independent.", "sourceParagraphs": [ 446, 447 ] }, { "type": "heading3", "text": "VR-4 — INDEPENDENTLY VERIFIED", "sourceParagraph": 449 }, { "type": "paragraph", "text": "The claim is supported by an appropriate independent source or review on the subject. Independence alone does not imply sufficiency. The source's relevance to the subject is also required.", "sourceParagraphs": [ 450, 451, 452 ] }, { "type": "heading3", "text": "VR-5 — MULTI-SOURCE SUSTAINED", "sourceParagraph": 454 }, { "type": "paragraph", "text": "The claim is strongly supported by reliable sources of different types within a defined time and scope.", "sourceParagraphs": [ 455 ] }, { "type": "heading3", "text": "VR-C — CONTRADICTED", "sourceParagraph": 457 }, { "type": "paragraph", "text": "The claim conflicts with stronger or more appropriate evidence.", "sourceParagraphs": [ 458 ] }, { "type": "heading3", "text": "VR-U — UNRESOLVED", "sourceParagraph": 460 }, { "type": "paragraph", "text": "The evidence is insufficient or contradictory. A reliable judgement cannot be made.", "sourceParagraphs": [ 461, 462 ] }, { "type": "heading3", "text": "VR-T — TIME-LIMITED", "sourceParagraph": 464 }, { "type": "paragraph", "text": "The claim is supported only for the specified period.", "sourceParagraphs": [ 465 ] }, { "type": "heading3", "text": "VR-S — SCOPE-LIMITED", "sourceParagraph": 467 }, { "type": "paragraph", "text": "The claim is supported only for the specified product, country, user, or operation scope. A claim can hold more than one status at the same time. Example:", "sourceParagraphs": [ 468, 469, 470 ] }, { "type": "heading3", "text": "VR-4 + VR-T + VR-S", "sourceParagraph": 471 }, { "type": "paragraph", "text": "It can indicate that the claim has been independently verified; however, it carries time and scope limits.", "sourceParagraphs": [ 472 ] }, { "type": "heading2", "text": "13. INDEPENDENCE IS NOT THE SAME AS APPROPRIATENESS", "sourceParagraph": 474 }, { "type": "paragraph", "text": "An independent source is not always the best source. A blogger may estimate a company's current price, while the company's canonical pricing page may be the more direct and appropriate source. Conversely, when a company declares on its own website that it is ‘the industry leader’, independent research with a disclosed method may be the more suitable evidence. Source evaluation therefore asks more than whether a source is independent. Is it authoritative for this type of claim? How close is it to the claim? Is its scope appropriate and current? Who controls it? Is its method disclosed? Does it address counter-evidence? Can it be audited? Is there a conflict of interest? Independence alone does not guarantee accuracy.", "sourceParagraphs": [ 475, 476, 477, 478, 479, 480, 481, 482, 483, 484, 485, 486, 487, 488, 489, 490, 491, 492, 493, 494, 495, 496 ] }, { "type": "paragraph", "text": "A source being first-party does not make it worthless either. The important thing is:", "sourceParagraphs": [ 497, 498 ] }, { "type": "callout", "text": "It is having the appropriate authority and evidence qualification for the claim carried by the source.", "sourceParagraph": 499 }, { "type": "heading2", "text": "14. CLAIM–SOURCE SUITABILITY MAP", "sourceParagraph": 501 }, { "type": "paragraph", "text": "The table below is a general foundational map. Specific evidence requirements may vary depending on the sector and risk level.", "sourceParagraphs": [ 502, 503 ] }, { "type": "table", "rows": [ [ "Type of claim", "Primary source or evidence", "Role of the official site", "Need for additional verification" ], [ "Legal company name", "Authorised official record", "Shows the declaration", "Generally required" ], [ "Trademark ownership", "Trademark registration, contract, corporate record", "Shows the declaration", "Required in case of financials" ], [ "Current published price", "Canonical price/checkout record", "Usually primary", "Verifiable with workflow" ], [ "Scope of service", "Operational records, contracts, official page", "Primary statement", "Additional evidence for actual delivery" ], [ "Licence", "Licensing institution or registry", "Shows the declaration", "Required" ], [ "Customer relationship", "Contract, customer approval, bilateral record", "Can be a unilateral claim", "Required in case of financials" ], [ "Success rate", "Raw data, method, denominator, audit", "Publishes the claim", "Required strongly" ], [ "Leadership", "Comparative method and data", "Self-declaration", "Independent or strong comparison required" ], [ "Customer satisfaction", "Representative survey, method, non-response", "Can explain the result", "Methodology required" ], [ "Date of establishment", "Official records and historical documents", "Shows the declaration", "Legal and brand history should be separated" ], [ "Partnership", "Bilateral statement or contract", "Unilateral declaration", "Generally required" ], [ "Certificate", "Issuing institution of the certificate and scope", "Shows the declaration", "Required" ], [ "Active employee", "HR or contract record, public profile", "Can show declaration", "Role and date verification" ], [ "User result", "Transaction and result record", "Describes case", "Context, method, and permission required" ], [ "\"Best option\"", "Defined user and comparison criteria", "Marketing view", "Strong comparison required" ] ] }, { "type": "paragraph", "text": "This table is not a universal source ranking. It shows the appropriateness between claim and evidence.", "sourceParagraphs": [ 504, 505 ] }, { "type": "heading2", "text": "15. DOUBLE LAYER CLAIM MATRIX", "sourceParagraph": 507 }, { "type": "paragraph", "text": "Two separate questions are asked for each financial claim: Has it been officially published? What is its status in the verified record of facts? These two questions constitute the following basic situations.", "sourceParagraphs": [ 508, 509, 510, 511 ] }, { "type": "heading3", "text": "15.1. Official and Verified", "sourceParagraph": 513 }, { "type": "paragraph", "text": "Claim: it is present in an official record and supported by appropriate evidence. Example: The price on the official price page is also verified in the transaction flow. This is one of the strongest types of core representation. If AI conveys the claim with the correct scope and timing, success occurs on both layers.", "sourceParagraphs": [ 514, 515, 516, 517, 518, 519, 520 ] }, { "type": "heading3", "text": "15.2. Official but Self-Declared Only", "sourceParagraph": 522 }, { "type": "paragraph", "text": "The claim is official. However, there is no additional verification. Example: “We are the most innovative company in the industry.” AI can convey the claim as: “The company describes itself as the most innovative company in the industry.” It cannot elevate it as: “It is the most innovative company in the industry.”", "sourceParagraphs": [ 523, 524, 525, 526, 527, 528, 529, 530, 531 ] }, { "type": "heading3", "text": "15.3. Official But Conflicting", "sourceParagraph": 533 }, { "type": "paragraph", "text": "The claim exists in the official record. It conflicts with more appropriate or up-to-date evidence. Example: The site says 20 countries. Operational records show eight countries. If AI repeats the site exactly, official fidelity may be high. Verified accuracy is low. This situation:", "sourceParagraphs": [ 534, 535, 536, 537, 538, 539, 540, 541 ] }, { "type": "heading4", "text": "Carrying the official error faithfully", "sourceParagraph": 542 }, { "type": "paragraph", "text": "can be classified as.", "sourceParagraphs": [ 543 ] }, { "type": "heading3", "text": "15.4. Not Official But Verified", "sourceParagraph": 545 }, { "type": "paragraph", "text": "The claim is not found on the official site. However, it has been verified with other reliable sources. Example: Recent legal ownership change has not yet been added to the official site. It appears in the authorised registry. AI can convey this information accurately with the correct source and timing. Official representation may have low fidelity or differ. Verified factual accuracy may be high.", "sourceParagraphs": [ 546, 547, 548, 549, 550, 551, 552, 553 ] }, { "type": "heading3", "text": "15.5. Not Official and Unverified", "sourceParagraph": 555 }, { "type": "paragraph", "text": "Claim: not in the official record, lacks sufficient external evidence. AI should not convey this as if it were true.", "sourceParagraphs": [ 556, 557, 558, 559 ] }, { "type": "heading3", "text": "15.6. Material Limit Officially Omitted", "sourceParagraph": 561 }, { "type": "paragraph", "text": "The company publishes the positive claim on its official page. However, it does not publish the necessary limitation to understand it correctly. Example: \"We provide services in Europe.\" However, the service is only offered in two countries and remotely. If the AI only repeats the general sentence, it can remain faithful to the official text. However, it may produce materially misleading results. Therefore, the Official Representation Layer must record not only the published sentences but also the material omissions.", "sourceParagraphs": [ 562, 563, 564, 565, 566, 567, 568, 569 ] }, { "type": "heading3", "text": "15.7. Dispute Unresolved", "sourceParagraph": 571 }, { "type": "paragraph", "text": "Official and external records contradict each other. Neither is strong enough. The correct result:", "sourceParagraphs": [ 572, 573, 574 ] }, { "type": "heading3", "text": "UNRESOLVED", "sourceParagraph": 575 }, { "type": "paragraph", "text": "should be. AI cannot be expected to make a definitive judgement.", "sourceParagraphs": [ 576, 577 ] }, { "type": "heading2", "text": "16. DOUBLE-LAYERED EVALUATION TABLE", "sourceParagraph": 579 }, { "type": "table", "rows": [ [ "Fidelity to official representation", "Verified reality accuracy", "Basic comment" ], [ "High", "High", "Faithful and verified representation" ], [ "High", "Low", "Carrying the official error faithfully" ], [ "Low", "High", "Corrective representation that diverges from the official record but is supported by evidence" ], [ "Low", "Low", "General representation failure" ], [ "Unclear", "High", "Official source conflicting; external verification strong" ], [ "High", "Unclear", "Official statement correctly conveyed; reality verification insufficient" ], [ "Unclear", "Unclear", "Insufficient reference for decision" ] ] }, { "type": "paragraph", "text": "This table establishes an important ruling:", "sourceParagraphs": [ 580 ] }, { "type": "callout", "text": "AI’s non-compliance with the official record is not always a failure.", "sourceParagraph": 581 }, { "type": "paragraph", "text": "Likewise:", "sourceParagraphs": [ 582 ] }, { "type": "callout", "text": "Compliance of AI with the official record is not always a success.", "sourceParagraph": 583 }, { "type": "heading2", "text": "17. RULE FOR PRESERVING SOURCE STATUS", "sourceParagraph": 585 }, { "type": "paragraph", "text": "The AI response must preserve not only the content but also the epistemic status of the content. The following sentences are not the same:", "sourceParagraphs": [ 586, 587 ] }, { "type": "bullet", "text": "\"The company is the world's leading GEO organisation.\"", "sourceParagraph": 588 }, { "type": "bullet", "text": "\"The company describes itself as the world's leading GEO organisation.\"", "sourceParagraph": 589 }, { "type": "bullet", "text": "\"The company claims to be the world's leading GEO organisation based on its published research that has not yet been independently verified.\"", "sourceParagraph": 590 }, { "type": "bullet", "text": "\"In the specified independent comparison, the company ranked first.\"", "sourceParagraph": 591 }, { "type": "paragraph", "text": "Each sentence carries a different evidence status. AI:", "sourceParagraphs": [ 592, 593 ] }, { "type": "bullet", "text": "self-declaration as an independent result,", "sourceParagraph": 594 }, { "type": "bullet", "text": "marketing statement as a scientific finding,", "sourceParagraph": 595 }, { "type": "bullet", "text": "press release as an editorial news,", "sourceParagraph": 596 }, { "type": "bullet", "text": "sponsored award as independent accreditation,", "sourceParagraph": 597 }, { "type": "bullet", "text": "company data as external audit", "sourceParagraph": 598 }, { "type": "paragraph", "text": "This treatment is prohibited by the following rule:", "sourceParagraphs": [ 599, 600 ] }, { "type": "heading4", "text": "Citation and Source Status Protection Rule", "sourceParagraph": 601 }, { "type": "paragraph", "text": "Source status must remain explicit.", "sourceParagraphs": [ 602 ] }, { "type": "heading2", "text": "18. EPISTEMIC UPLIFT", "sourceParagraph": 604 }, { "type": "paragraph", "text": "The strengthening of a claim's source status in an AI response is called:", "sourceParagraphs": [ 605 ] }, { "type": "heading4", "text": "Epistemic uplift", "sourceParagraph": 606 }, { "type": "paragraph", "text": "Example transformations:", "sourceParagraphs": [ 607, 608 ] }, { "type": "table", "rows": [ [ "Source text", "Incorrect uplift by AI" ], [ "“According to the company's own data”", "“Proven”" ], [ "“Reported by specific customers”", "“Works in all customers”" ], [ "“Observed in the pilot study”", "“Generally effective”" ], [ "“The company defines itself as a leader”", "“Industry leader”" ], [ "“No independent verification was done”", "Limitation completely removed" ], [ "“Available in some markets”", "“Available globally”" ] ] }, { "type": "paragraph", "text": "These transformations are not summaries. They are the production of new and stronger claims.", "sourceParagraphs": [ 609, 610 ] }, { "type": "heading2", "text": "19. OFFICIAL SILENCE", "sourceParagraph": 612 }, { "type": "paragraph", "text": "The absence of information on a company's official page may not mean:", "sourceParagraphs": [ 613 ] }, { "type": "bullet", "text": "the information is incorrect,", "sourceParagraph": 614 }, { "type": "bullet", "text": "the information is prohibited,", "sourceParagraph": 615 }, { "type": "paragraph", "text": "AI cannot use this information. Official silence can take three different forms.", "sourceParagraphs": [ 616, 617 ] }, { "type": "heading3", "text": "19.1. Non-Material Silence", "sourceParagraph": 619 }, { "type": "paragraph", "text": "Information may be outside the prompt and user decision. Whether AI uses it or not is immaterial.", "sourceParagraphs": [ 620, 621 ] }, { "type": "heading3", "text": "19.2. Legitimate Supplementary External Information", "sourceParagraph": 623 }, { "type": "paragraph", "text": "It may be information not found on the official site but verified by a reliable source. Example:", "sourceParagraphs": [ 624, 625 ] }, { "type": "bullet", "text": "Legal record", "sourceParagraph": 626 }, { "type": "bullet", "text": "Historical event", "sourceParagraph": 627 }, { "type": "bullet", "text": "Public regulatory decision", "sourceParagraph": 628 }, { "type": "bullet", "text": "Independent research", "sourceParagraph": 629 }, { "type": "paragraph", "text": "AI can use this information with appropriate context and source status.", "sourceParagraphs": [ 630 ] }, { "type": "heading3", "text": "19.3. Silence of a Material Limit", "sourceParagraph": 632 }, { "type": "paragraph", "text": "An entity may not have disclosed a limit that affects the user's decision in its official representation. Example:", "sourceParagraphs": [ 633, 634 ] }, { "type": "bullet", "text": "The licence being valid only in certain countries", "sourceParagraph": 635 }, { "type": "bullet", "text": "The price appearing not to carry an additional fee", "sourceParagraph": 636 }, { "type": "bullet", "text": "The service being available only to certain types of customers", "sourceParagraph": 637 }, { "type": "bullet", "text": "Success data coming only from selected cases", "sourceParagraph": 638 }, { "type": "paragraph", "text": "In this case, the official representation itself may carry an audit finding. AI repeating the silence may reduce the truthfulness of reality.", "sourceParagraphs": [ 639, 640 ] }, { "type": "heading2", "text": "20. MATERIAL DEFICIENCY IN OFFICIAL REPRESENTATION", "sourceParagraph": 642 }, { "type": "paragraph", "text": "An official sentence may not be directly false. However, removing the necessary limits can make it misleading. Example: “We operate in Europe.” Material context:", "sourceParagraphs": [ 643, 644, 645, 646, 647 ] }, { "type": "bullet", "text": "only in two countries,", "sourceParagraph": 648 }, { "type": "bullet", "text": "only remotely,", "sourceParagraph": 649 }, { "type": "bullet", "text": "without a local office,", "sourceParagraph": 650 }, { "type": "bullet", "text": "only for corporate clients", "sourceParagraph": 651 }, { "type": "paragraph", "text": "may be provided. The official representation layer must record these two areas together:", "sourceParagraphs": [ 652, 653 ] }, { "type": "bullet", "text": "The main claim published", "sourceParagraph": 654 }, { "type": "bullet", "text": "The material limits necessary to correctly understand the claim", "sourceParagraph": 655 }, { "type": "paragraph", "text": "If an AI response carries the main sentence but removes the limits:", "sourceParagraphs": [ 656 ] }, { "type": "bullet", "text": "faithful in terms of words,", "sourceParagraph": 657 }, { "type": "bullet", "text": "misleading in terms of judgement", "sourceParagraph": 658 }, { "type": "paragraph", "text": "may be.", "sourceParagraphs": [ 659 ] }, { "type": "heading2", "text": "21. WRONG IS WRONG EVEN IF IT FAVOURS YOU", "sourceParagraph": 661 }, { "type": "paragraph", "text": "When an AI product portrays an entity as more powerful, larger or more authoritative than it is, silence by the organisation becomes a measurement and governance problem. A local company may be portrayed as a global business; a consultancy as a regulator; self-assessment as independent certification; or an operation spanning eight countries as operating in 20. Even when an error favours the organisation, it must be recorded as an error in the verified-reality layer. The founding warning of the second book sets the same boundary: a commercially attractive error must still be corrected. When an organisation knowingly exploits a favourable AI error:", "sourceParagraphs": [ 662, 663, 664, 665, 666, 667, 668, 669, 670 ] }, { "type": "bullet", "text": "obligation to correct,", "sourceParagraph": 671 }, { "type": "bullet", "text": "integrity of representation,", "sourceParagraph": 672 }, { "type": "bullet", "text": "commercial promise,", "sourceParagraph": 673 }, { "type": "bullet", "text": "governance", "sourceParagraph": 674 }, { "type": "paragraph", "text": "can also be examined in terms of.", "sourceParagraphs": [ 675 ] }, { "type": "heading2", "text": "22. REFERENCE CONFLICT", "sourceParagraph": 677 }, { "type": "paragraph", "text": "Sometimes the problem is not in the AI response, but in the reference records. Example: The homepage shows a different price. The PDF shows a different price. Checkout shows another price. Structured data carries an outdated price. In this case, there may not be a single correct official representation. Another example: The legal registry indicates one ownership. The company website states another ownership. Update dates are uncertain. In this case, the verified reality record may also carry a discrepancy.", "sourceParagraphs": [ 678, 679, 680, 681, 682, 683, 684, 685, 686, 687, 688, 689 ] }, { "type": "heading2", "text": "23. REFERENCE DEFECT RULE", "sourceParagraph": 691 }, { "type": "callout", "text": "If reference records are materially contradictory or insufficient, this uncertainty cannot automatically be converted into a failure of the AI product.", "sourceParagraph": 692 }, { "type": "paragraph", "text": "Correct statuses:", "sourceParagraphs": [ 693 ] }, { "type": "heading3", "text": "REFERENCE_CONFLICT", "sourceParagraph": 694 }, { "type": "heading3", "text": "INSUFFICIENT EVIDENCE", "sourceParagraph": 695 }, { "type": "heading3", "text": "UNRESOLVED", "sourceParagraph": 696 }, { "type": "heading3", "text": "TIME-SPLIT", "sourceParagraph": 697 }, { "type": "heading3", "text": "SCOPE-SPLIT", "sourceParagraph": 698 }, { "type": "paragraph", "text": "are possible. The audit should separate the question: \"Did AI provide an incorrect answer?\" and: \"Was there necessary reference reality for AI to provide the correct answer?\" If an institution has left its official records contradictory, we can measure which record the model chooses. However, we cannot hide the reference flaw while expecting a single and definitive institutional answer from the model.", "sourceParagraphs": [ 699, 700, 701, 702, 703, 704, 705 ] }, { "type": "heading2", "text": "24. CANONICAL RECORD LOCK", "sourceParagraph": 707 }, { "type": "paragraph", "text": "Both reference layers of GEO-1000 should be versioned before the measurement wave begins:", "sourceParagraphs": [ 708 ] }, { "type": "heading4", "text": "Official Representative Record", "sourceParagraph": 709 }, { "type": "paragraph", "text": "Which pages? Which language versions? Which date? Which claims? Which archive status?", "sourceParagraphs": [ 710, 711, 712, 713, 714 ] }, { "type": "heading4", "text": "Verified Entity Reality Record", "sourceParagraph": 715 }, { "type": "paragraph", "text": "Which evidence? Which counter-evidence? Which verification statuses? Which objections? Which date? If there is a material change during measurement: The wave can be stopped. The record can be split into a new version. Previous and subsequent observations can be analysed separately. Mixing the results under a single reference is prohibited.", "sourceParagraphs": [ 716, 717, 718, 719, 720, 721, 722, 723, 724, 725 ] }, { "type": "heading2", "text": "25. BURDEN OF PROOF WITH CLAIM STRENGTH", "sourceParagraph": 727 }, { "type": "paragraph", "text": "As a claim grows, the burden of proof also increases. The ladder below is a constructive representation.", "sourceParagraphs": [ 728, 729 ] }, { "type": "heading4", "text": "C0 — Simple Identity Claim", "sourceParagraph": 731 }, { "type": "paragraph", "text": "“X is a consultancy company.” Correct entity and activity record is required.", "sourceParagraphs": [ 732, 733 ] }, { "type": "heading4", "text": "C1 — Scope Claim", "sourceParagraph": 735 }, { "type": "paragraph", "text": "\"X provides services in Germany.\" An operation, contract, or delivery record is required.", "sourceParagraphs": [ 736, 737 ] }, { "type": "heading4", "text": "C2 — Quantitative Performance Claim", "sourceParagraph": 739 }, { "type": "paragraph", "text": "\"90% of projects are completed on time.\" Denominator, period, method, and unsuccessful results are required.", "sourceParagraphs": [ 740, 741 ] }, { "type": "heading4", "text": "C3 — Comparative Claim", "sourceParagraph": 743 }, { "type": "paragraph", "text": "\"It is faster than competitors.\" Comparison universe and equivalent measurement are required.", "sourceParagraphs": [ 744, 745 ] }, { "type": "heading4", "text": "C4 — Superiority Claim", "sourceParagraph": 747 }, { "type": "paragraph", "text": "\"It is the industry leader.\" Industry, criteria, geography, time, and method are required.", "sourceParagraphs": [ 748, 749 ] }, { "type": "heading4", "text": "C5 — Guarantee or Universal Provision", "sourceParagraph": 751 }, { "type": "paragraph", "text": "“Guarantees results for every customer.” True verification requires extremely strong, comprehensive, and actionable evidence. Official sources may be strong for C0 and some C1 claims. Alone, they may not be sufficient at C3–C5 levels.", "sourceParagraphs": [ 752, 753, 754, 755 ] }, { "type": "heading2", "text": "26. TWO SEPARATE ACCURACY MEASUREMENTS", "sourceParagraph": 757 }, { "type": "paragraph", "text": "Two main measurements must be reported separately within GEO-1000.", "sourceParagraphs": [ 758 ] }, { "type": "heading3", "text": "26.1. Official Representation Fidelity", "sourceParagraph": 760 }, { "type": "paragraph", "text": "Answers the following question:", "sourceParagraphs": [ 761 ] }, { "type": "callout", "text": "To what extent does the AI response convey the versioned official and canonical representation of the entity accurately, completely, and while preserving source status?", "sourceParagraph": 762 }, { "type": "paragraph", "text": "This measurement:", "sourceParagraphs": [ 763 ] }, { "type": "bullet", "text": "official identity,", "sourceParagraph": 764 }, { "type": "bullet", "text": "official service scope,", "sourceParagraph": 765 }, { "type": "bullet", "text": "published price,", "sourceParagraph": 766 }, { "type": "bullet", "text": "official boundaries,", "sourceParagraph": 767 }, { "type": "bullet", "text": "official source nature", "sourceParagraph": 768 }, { "type": "paragraph", "text": "assesses their areas.", "sourceParagraphs": [ 769 ] }, { "type": "heading3", "text": "26.2. Accuracy of Verified Entity Reality", "sourceParagraph": 771 }, { "type": "paragraph", "text": "Answers the following question:", "sourceParagraphs": [ 772 ] }, { "type": "callout", "text": "To what extent does the AI response convey the current entity reality, supported by available evidence, accurately, comprehensively, and with preserved boundaries?", "sourceParagraph": 773 }, { "type": "paragraph", "text": "This measurement:", "sourceParagraphs": [ 774 ] }, { "type": "bullet", "text": "supported factual claims,", "sourceParagraph": 775 }, { "type": "bullet", "text": "conflicting claims,", "sourceParagraph": 776 }, { "type": "bullet", "text": "outdated information,", "sourceParagraph": 777 }, { "type": "bullet", "text": "incorrect relations,", "sourceParagraph": 778 }, { "type": "bullet", "text": "factual deficiencies,", "sourceParagraph": 779 }, { "type": "bullet", "text": "epistemic enhancements", "sourceParagraph": 780 }, { "type": "paragraph", "text": "are assessed.", "sourceParagraphs": [ 781 ] }, { "type": "heading2", "text": "27. THE INABILITY OF TWO SCORES TO COMPENSATE FOR EACH OTHER", "sourceParagraph": 783 }, { "type": "paragraph", "text": "High official fidelity cannot compensate for low verified accuracy. Example: The official site publishes a false claim. AI repeats the claim exactly. Official fidelity may be high. Verified accuracy is low. Similarly, high verified accuracy does not completely trivialise the distortion of official representation. Example: AI uses an up-to-date external source correctly. It confuses the product and service description of one company with another. Some external facts in the response may be correct. Official identity fidelity is weak. Therefore, the two results should appear separately.", "sourceParagraphs": [ 784, 785, 786, 787, 788, 789, 790, 791, 792, 793, 794, 795, 796 ] }, { "type": "heading2", "text": "28. DOUBLE-LAYER RESPONSE STATUSES", "sourceParagraph": 798 }, { "type": "paragraph", "text": "An AI response can receive one of these results:", "sourceParagraphs": [ 799 ] }, { "type": "heading3", "text": "DUAL-PASS", "sourceParagraph": 801 }, { "type": "paragraph", "text": "It is materially consistent with both official representation and verified reality.", "sourceParagraphs": [ 802 ] }, { "type": "heading3", "text": "OFFICIAL-PASS / VERIFIED-FAIL", "sourceParagraph": 804 }, { "type": "paragraph", "text": "Carries the official record faithfully. However, it conveys an unsupported, old, or contradictory claim of the official record as if it were true.", "sourceParagraphs": [ 805, 806 ] }, { "type": "heading3", "text": "OFFICIAL-FAIL / VERIFIED-PASS", "sourceParagraph": 808 }, { "type": "paragraph", "text": "Does not fully comply with the official site. However, it conveys the truth correctly supported by current and stronger evidence. An explanation of the response's difference is preferred.", "sourceParagraphs": [ 809, 810, 811 ] }, { "type": "heading3", "text": "DUAL-FAIL", "sourceParagraph": 813 }, { "type": "paragraph", "text": "It materially disrupts both the official record and the verified reality.", "sourceParagraphs": [ 814 ] }, { "type": "heading3", "text": "OFFICIAL-PASS / VERIFIED-UNKNOWN", "sourceParagraph": 816 }, { "type": "paragraph", "text": "The official statement has been accurately conveyed. The verified reality status of the claim has not yet been determined.", "sourceParagraphs": [ 817, 818 ] }, { "type": "heading3", "text": "REFERENCE-CONFLICT", "sourceParagraph": 820 }, { "type": "paragraph", "text": "Reference records are contradictory in themselves. It is not possible to score the response definitively.", "sourceParagraphs": [ 821, 822 ] }, { "type": "heading3", "text": "APPROPRIATE-UNCERTAINTY", "sourceParagraph": 824 }, { "type": "paragraph", "text": "The AI has correctly explained the reference uncertainty and avoided making a definitive judgement. In some cases, this may be the most accurate answer.", "sourceParagraphs": [ 825, 826 ] }, { "type": "heading2", "text": "29. SYNTHETIC APPLE.COM DISPLAY", "sourceParagraph": 828 }, { "type": "paragraph", "text": "SYNTHETIC METHODOLOGY DEMONSTRATION / The following claims, responses, and scores are entirely fictional. They do not represent real information, performance, or responses of any actual Apple Inc. AI product. Synthetic Official Representation Record: O-01: The company defines itself as a global consumer technology brand. O-02: The company publishes certain product and digital service categories. O-03: The company defines itself as 'the world's most innovative company.' O-04: The official website contains specific country store conditions. Synthetic Verified Reality Record: V-01: Corporate identity and main business category are supported. V-02: Certain product categories are supported. V-03: There is no defined comparison method for the claim 'the world's most innovative company.' V-04: Store conditions vary by country.", "sourceParagraphs": [ 829, 830, 831, 832, 833, 834, 835, 836, 837, 838, 839 ] }, { "type": "heading3", "text": "29.1. Synthetic Response A", "sourceParagraph": 841 }, { "type": "paragraph", "text": "“Apple.com is one of the official corporate interfaces where the company presents its consumer technology products and digital services. The company positions itself as one of the most innovative companies in the world; this claim of superiority should also be evaluated according to the comparison method used.” Evaluation: Official representation is preserved. The claim of superiority is attributed to the company. The claim has not been elevated to independent factuality. The scope limitation has been added. Candidate status:", "sourceParagraphs": [ 842, 843, 844, 845, 846, 847, 848 ] }, { "type": "heading3", "text": "DUAL-PASS", "sourceParagraph": 849 }, { "type": "heading3", "text": "29.2. Synthetic Response B", "sourceParagraph": 851 }, { "type": "paragraph", "text": "“Apple is the undisputed most innovative company in the world.” Evaluation: The official marketing claim has been turned into an absolute truth. There is no comparison method. Source status is lost. “Undisputed” is an additional epistemic elevation. Candidate status:", "sourceParagraphs": [ 852, 853, 854, 855, 856, 857, 858 ] }, { "type": "heading3", "text": "OFFICIAL-PASS / VERIFIED-FAIL", "sourceParagraph": 859 }, { "type": "heading3", "text": "29.3. Synthetic Response C", "sourceParagraph": 861 }, { "type": "paragraph", "text": "“Apple is a local company that only sells phones in physical stores.” Evaluation: Corporate identity and scope are distorted. Official representation is incorrect. Contradicts verified reality. Candidate status:", "sourceParagraphs": [ 862, 863, 864, 865, 866, 867 ] }, { "type": "heading3", "text": "DUAL-FAIL", "sourceParagraph": 868 }, { "type": "heading3", "text": "29.4. Synthetic Response D", "sourceParagraph": 870 }, { "type": "paragraph", "text": "“The company’s official website publishes certain store conditions; however, prices and purchase terms may vary according to the country version.” Evaluation: The official record has been used correctly as a source. Geographic scope is preserved. The official page has not been made a single price source for the whole world. Candidate status:", "sourceParagraphs": [ 871, 872, 873, 874, 875, 876 ] }, { "type": "heading3", "text": "DUAL-PASS", "sourceParagraph": 877 }, { "type": "heading2", "text": "30. SYNTHETIC ASTERON CASE", "sourceParagraph": 879 }, { "type": "heading3", "text": "SYNTHETIC CASE — NOT A REAL INSTITUTION", "sourceParagraph": 880 }, { "type": "paragraph", "text": "The official site of a company named Asteron publishes the following claims:", "sourceParagraphs": [ 881 ] }, { "type": "bullet", "text": "“We operate in 25 countries.”", "sourceParagraph": 882 }, { "type": "bullet", "text": "“We achieve success in 98% of the projects.”", "sourceParagraph": 883 }, { "type": "bullet", "text": "We are an independently verified global standards organisation.", "sourceParagraph": 884 }, { "type": "bullet", "text": "All of our projects are guaranteed.", "sourceParagraph": 885 }, { "type": "paragraph", "text": "Audit records:", "sourceParagraphs": [ 886 ] }, { "type": "bullet", "text": "Verified active service country: 9", "sourceParagraph": 887 }, { "type": "bullet", "text": "Success data: only the 40 selected projects", "sourceParagraph": 888 }, { "type": "bullet", "text": "Independent verification: not found", "sourceParagraph": 889 }, { "type": "bullet", "text": "Standard: the recommended standard published by the company itself", "sourceParagraph": 890 }, { "type": "bullet", "text": "Guarantee: contractual only for certain delivery items", "sourceParagraph": 891 }, { "type": "paragraph", "text": "AI response: 'Asteron is a globally standardised organisation, operating in 25 countries, with a 98% success rate, independently verified, and guarantees results in all its projects.'", "sourceParagraphs": [ 892, 893 ] }, { "type": "heading3", "text": "30.1. Official representation fidelity", "sourceParagraph": 895 }, { "type": "paragraph", "text": "The response is largely faithful to the website text.", "sourceParagraphs": [ 896 ] }, { "type": "heading3", "text": "30.2. Verified Reality Accuracy", "sourceParagraph": 898 }, { "type": "paragraph", "text": "Response:", "sourceParagraphs": [ 899 ] }, { "type": "bullet", "text": "expands the country scope,", "sourceParagraph": 900 }, { "type": "bullet", "text": "makes selected data a general success,", "sourceParagraph": 901 }, { "type": "bullet", "text": "presents the entity's own standard as an independent standard,", "sourceParagraph": 902 }, { "type": "paragraph", "text": "turns delivery guarantee into outcome guarantee. It is a severe failure in terms of verified reality.", "sourceParagraphs": [ 903, 904 ] }, { "type": "heading3", "text": "30.3. How Could a Correct AI Response Be?", "sourceParagraph": 906 }, { "type": "paragraph", "text": "“Asteron states on its own website that it provides services in 25 countries, has a 98% success rate, and publishes a standard. In the existing audit records, active service has been verified in nine countries; it has been observed that the 98% rate is based on 40 selected projects; independent verification could not be confirmed. The guarantee is limited to specific delivery items.” This response:", "sourceParagraphs": [ 907, 908 ] }, { "type": "bullet", "text": "does not conceal the company statement,", "sourceParagraph": 909 }, { "type": "bullet", "text": "preserves the status of the statement,", "sourceParagraph": 910 }, { "type": "bullet", "text": "distinguishes the verified reality,", "sourceParagraph": 911 }, { "type": "paragraph", "text": "makes the limitation visible.", "sourceParagraphs": [ 912 ] }, { "type": "heading2", "text": "31. SYNTHETIC CASE CONTAINING REFERENCE DEFECT", "sourceParagraph": 914 }, { "type": "heading3", "text": "SYNTHETIC CASE", "sourceParagraph": 915 }, { "type": "paragraph", "text": "Official homepage: “Starting price 2,000 euros.” Official price PDF: “Starting price 3,000 euros.” Checkout: “4,000 euros.” Structured data: “1,500 euros.” AI response: “Starting price is 3,000 euros.” Which record did the AI choose? The PDF. Is it wrong? Maybe. Is it correct? Maybe. The reference record is inconsistent. In this case:", "sourceParagraphs": [ 916, 917, 918, 919, 920, 921, 922, 923, 924, 925, 926, 927, 928, 929, 930, 931, 932, 933 ] }, { "type": "bullet", "text": "AI failure,", "sourceParagraph": 934 }, { "type": "bullet", "text": "official representation conflict,", "sourceParagraph": 935 }, { "type": "bullet", "text": "lack of canonical source", "sourceParagraph": 936 }, { "type": "paragraph", "text": "are separate findings. The correct assessment:", "sourceParagraphs": [ 937, 938 ] }, { "type": "heading3", "text": "REFERENCE-CONFLICT", "sourceParagraph": 939 }, { "type": "paragraph", "text": "is possible. An entity cannot expect exact price fidelity without reconciling its own price records.", "sourceParagraphs": [ 940, 941 ] }, { "type": "heading2", "text": "32. SOURCE PRIORITY IN THE OFFICIAL REPRESENTATION LAYER", "sourceParagraph": 943 }, { "type": "paragraph", "text": "When multiple official records are found, a priority rule should be set for each claim field. Example candidate priority:", "sourceParagraphs": [ 944, 945 ] }, { "type": "bullet", "text": "Current contract or transaction record", "sourceParagraph": 946 }, { "type": "bullet", "text": "Canonical up-to-date product/price record", "sourceParagraph": 947 }, { "type": "bullet", "text": "Authorised structured data", "sourceParagraph": 948 }, { "type": "bullet", "text": "Current official page", "sourceParagraph": 949 }, { "type": "bullet", "text": "Current official PDF", "sourceParagraph": 950 }, { "type": "bullet", "text": "Historical press or blog record", "sourceParagraph": 951 }, { "type": "paragraph", "text": "This order is not universal for all claims. For example, press archives might be valuable for corporate history. An old blog record is not authoritative for price. Priority should be defined according to the type of claim.", "sourceParagraphs": [ 952, 953, 954, 955 ] }, { "type": "heading2", "text": "33. SOURCE CONFLICT IN VERIFIED REALITY", "sourceParagraph": 957 }, { "type": "paragraph", "text": "Two independent sources may contradict each other. In this case:", "sourceParagraphs": [ 958, 959 ] }, { "type": "bullet", "text": "the more independent one,", "sourceParagraph": 960 }, { "type": "bullet", "text": "the newer one,", "sourceParagraph": 961 }, { "type": "bullet", "text": "the one closer to the subject,", "sourceParagraph": 962 }, { "type": "bullet", "text": "the one with a clearer method,", "sourceParagraph": 963 }, { "type": "bullet", "text": "the one using broader data", "sourceParagraph": 964 }, { "type": "paragraph", "text": "does not automatically win. All criteria should be evaluated together. Example: A new blog post can be more up-to-date than an old official registry. However, the blog may not have sufficient authority for legal ownership. Source conflict resolution should be sensitive to the type of claim.", "sourceParagraphs": [ 965, 966, 967, 968, 969, 970 ] }, { "type": "heading2", "text": "34. MANDATORY PRESERVATION OF COUNTER-EVIDENCE", "sourceParagraph": 972 }, { "type": "paragraph", "text": "A verified record of facts cannot include only records supporting the claim. The following must also be preserved:", "sourceParagraphs": [ 973, 974 ] }, { "type": "bullet", "text": "Records contradicting the claim", "sourceParagraph": 975 }, { "type": "bullet", "text": "Unsuccessful cases", "sourceParagraph": 976 }, { "type": "bullet", "text": "Withdrawn results", "sourceParagraph": 977 }, { "type": "bullet", "text": "Expired certificates", "sourceParagraph": 978 }, { "type": "bullet", "text": "Negative user outcomes", "sourceParagraph": 979 }, { "type": "bullet", "text": "Regulatory decisions", "sourceParagraph": 980 }, { "type": "bullet", "text": "Corrections and errata", "sourceParagraph": 981 }, { "type": "bullet", "text": "Discrepancy between sources", "sourceParagraph": 982 }, { "type": "paragraph", "text": "Supporting a claim does not allow the dismissal of counter-evidence.", "sourceParagraphs": [ 983 ] }, { "type": "heading2", "text": "ARTICLE 35", "sourceParagraph": 985 }, { "type": "paragraph", "text": "Not all official–verified differences have the same level of importance. A material difference is a difference that could change the user’s or system’s decision. Examples: A minor spelling difference in the company name may not be material. Miswriting the type of company may be material. Missing colour of a product may not be material. Incorrect country of service is material. A difference in date format may not be material. Incorrect licence validity date is material. A change in marketing tone may not be material. Removing the warranty limit is material. Materiality should be evaluated based on its impact on:", "sourceParagraphs": [ 986, 987, 988, 989, 990, 991, 992, 993, 994, 995, 996, 997 ] }, { "type": "bullet", "text": "user decision,", "sourceParagraph": 998 }, { "type": "bullet", "text": "legal liability,", "sourceParagraph": 999 }, { "type": "bullet", "text": "price,", "sourceParagraph": 1000 }, { "type": "bullet", "text": "security,", "sourceParagraph": 1001 }, { "type": "bullet", "text": "authority,", "sourceParagraph": 1002 }, { "type": "bullet", "text": "scope,", "sourceParagraph": 1003 }, { "type": "bullet", "text": "evidence,", "sourceParagraph": 1004 }, { "type": "bullet", "text": "time", "sourceParagraph": 1005 }, { "type": "paragraph", "text": "These categories must be assessed by their effect on the user decision, legal liability, price, security, authority, scope, evidence and time.", "sourceParagraphs": [ 1006 ] }, { "type": "heading2", "text": "36. MANDATORY NORMATIVE PROVISIONS", "sourceParagraph": 1008 }, { "type": "protocol_label", "text": "CH04-N01", "sourceParagraph": 1009 }, { "type": "paragraph", "text": "Each material AI response should be evaluated separately against the Official Representation Record and the Verified Entity Reality Record.", "sourceParagraphs": [ 1010 ] }, { "type": "protocol_label", "text": "CH04-N02", "sourceParagraph": 1011 }, { "type": "paragraph", "text": "Official representation fidelity cannot be presented as the same outcome as material accuracy.", "sourceParagraphs": [ 1012 ] }, { "type": "protocol_label", "text": "CH04-N03", "sourceParagraph": 1013 }, { "type": "paragraph", "text": "A claim on the entity's own site does not count as independent verification without additional evidence appropriate to the claim type.", "sourceParagraphs": [ 1014 ] }, { "type": "protocol_label", "text": "CH04-N04", "sourceParagraph": 1015 }, { "type": "paragraph", "text": "First-party evidence cannot be rejected solely for being first-party; its appropriateness to the claim type, method, and conflict of interest should be evaluated.", "sourceParagraphs": [ 1016 ] }, { "type": "protocol_label", "text": "CH04-N05", "sourceParagraph": 1017 }, { "type": "paragraph", "text": "An independent source cannot be accepted as correct or authoritative solely for being independent.", "sourceParagraphs": [ 1018 ] }, { "type": "protocol_label", "text": "CH04-N06", "sourceParagraph": 1019 }, { "type": "paragraph", "text": "Company opinion, marketing claim, or self-assessment cannot be elevated to independent fact in the AI response.", "sourceParagraphs": [ 1020 ] }, { "type": "protocol_label", "text": "CH04-N07", "sourceParagraph": 1021 }, { "type": "paragraph", "text": "The source type and epistemic status should be maintained in the AI response to the extent that it is material.", "sourceParagraphs": [ 1022 ] }, { "type": "protocol_label", "text": "CH04-N08", "sourceParagraph": 1023 }, { "type": "paragraph", "text": "An AI response that contradicts the official record cannot be considered an automatic failure if it is supported by stronger, more recent evidence.", "sourceParagraphs": [ 1024 ] }, { "type": "protocol_label", "text": "CH04-N09", "sourceParagraph": 1025 }, { "type": "paragraph", "text": "An AI response faithful to the official record cannot be considered successful in general accuracy if the official claim contradicts the verified fact.", "sourceParagraphs": [ 1026 ] }, { "type": "protocol_label", "text": "CH04-N10", "sourceParagraph": 1027 }, { "type": "paragraph", "text": "If official and verified records materially conflict, the result should be REFERENCE_CONFLICT, UNRESOLVED, or carry an appropriate uncertainty status.", "sourceParagraphs": [ 1028 ] }, { "type": "protocol_label", "text": "CH04-N11", "sourceParagraph": 1029 }, { "type": "paragraph", "text": "A reference defect cannot quietly be turned into an AI product error.", "sourceParagraphs": [ 1030 ] }, { "type": "protocol_label", "text": "CH04-N12", "sourceParagraph": 1031 }, { "type": "paragraph", "text": "If the AI response treats an unverified first-party claim as fact, appropriate citation and limitation should be sought.", "sourceParagraphs": [ 1032 ] }, { "type": "protocol_label", "text": "CH04-N13", "sourceParagraph": 1033 }, { "type": "paragraph", "text": "Material official omissions cannot be left simply appearing as if the published sentence is correct.", "sourceParagraphs": [ 1034 ] }, { "type": "protocol_label", "text": "CH04-N14", "sourceParagraph": 1035 }, { "type": "paragraph", "text": "Errors favourable to the company must also be recorded as errors in the verified reality layer.", "sourceParagraphs": [ 1036 ] }, { "type": "protocol_label", "text": "CH04-N15", "sourceParagraph": 1037 }, { "type": "paragraph", "text": "Official Representation Record and Verified Entity Reality Record must be versioned before the measurement begins.", "sourceParagraphs": [ 1038 ] }, { "type": "protocol_label", "text": "CH04-N16", "sourceParagraph": 1039 }, { "type": "paragraph", "text": "If there is a change in the material reference during the measurement wave, previous and subsequent observations must be evaluated in separate versions.", "sourceParagraphs": [ 1040 ] }, { "type": "protocol_label", "text": "CH04-N17", "sourceParagraph": 1041 }, { "type": "paragraph", "text": "Counter-evidence and negative records cannot be removed from the verified reality file solely because they are negative.", "sourceParagraphs": [ 1042 ] }, { "type": "protocol_label", "text": "CH04-N18", "sourceParagraph": 1043 }, { "type": "paragraph", "text": "UNRESOLVED, UNKNOWN, and INSUFFICIENT EVIDENCE results are not considered positive validation.", "sourceParagraphs": [ 1044 ] }, { "type": "protocol_label", "text": "CH04-N19", "sourceParagraph": 1045 }, { "type": "paragraph", "text": "As the strength of the claim increases, the required burden of evidence must also increase.", "sourceParagraphs": [ 1046 ] }, { "type": "protocol_label", "text": "CH04-N20", "sourceParagraph": 1047 }, { "type": "paragraph", "text": "High Official representation fidelity cannot compensate for low Verified Entity Reality Accuracy.", "sourceParagraphs": [ 1048 ] }, { "type": "protocol_label", "text": "CH04-N21", "sourceParagraph": 1049 }, { "type": "paragraph", "text": "There must be an accountable human or institution owner as a result of the two-layer evaluation.", "sourceParagraphs": [ 1050 ] }, { "type": "heading2", "text": "37. FORMS OF FAILURE", "sourceParagraph": 1052 }, { "type": "protocol_label", "text": "CH04-F01 — CONSIDERING THE WEBSITE AS ABSOLUTE REALITY", "sourceParagraph": 1053 }, { "type": "paragraph", "text": "All claims published by the institution are accepted as verified facts.", "sourceParagraphs": [ 1054 ] }, { "type": "protocol_label", "text": "CH04-F02 — CONSIDERING OFFICIAL RECORDS AS WORTHLESS", "sourceParagraph": 1055 }, { "type": "paragraph", "text": "Direct institutional records such as price, policy, or product coverage are ignored because it is the first party.", "sourceParagraphs": [ 1056 ] }, { "type": "protocol_label", "text": "CH04-F03 — MAKING SELF-DECLARATION AN INDEPENDENT EVIDENCE", "sourceParagraph": 1057 }, { "type": "paragraph", "text": "The company's own leadership or success claim is used like external verification.", "sourceParagraphs": [ 1058 ] }, { "type": "protocol_label", "text": "CH04-F04 — WORSHIP OF INDEPENDENT SOURCES", "sourceParagraph": 1059 }, { "type": "paragraph", "text": "Independent but distant from the subject, old, or unmethodical sources are considered superior to official and close records.", "sourceParagraphs": [ 1060 ] }, { "type": "protocol_label", "text": "CH04-F05 — EPISTEMIC ELEVATION", "sourceParagraph": 1061 }, { "type": "paragraph", "text": "The phrase \"The company says\" is turned into the judgement \"it has been proven.\"", "sourceParagraphs": [ 1062 ] }, { "type": "protocol_label", "text": "CH04-F06 — ERASING SOURCE STATUS", "sourceParagraph": 1063 }, { "type": "paragraph", "text": "First-party data, sponsored content, or press releases are presented as independent news.", "sourceParagraphs": [ 1064 ] }, { "type": "protocol_label", "text": "CH04-F07 — FAITHFULLY REPRODUCING OFFICIAL ERRORS", "sourceParagraph": 1065 }, { "type": "paragraph", "text": "AI repeats the official false claim correctly and is generally considered successful.", "sourceParagraphs": [ 1066 ] }, { "type": "protocol_label", "text": "CH04-F08 — PUNISHING UP-TO-DATE EXTERNAL CORRECTIONS", "sourceParagraph": 1067 }, { "type": "paragraph", "text": "AI is considered wrong because it leaves the old official site and uses strong current evidence.", "sourceParagraphs": [ 1068 ] }, { "type": "protocol_label", "text": "CH04-F09 — CONSIDERING OFFICIAL SILENCE AS PROHIBITED", "sourceParagraph": 1069 }, { "type": "paragraph", "text": "Any external information not found on the official site is automatically considered incorrect.", "sourceParagraphs": [ 1070 ] }, { "type": "protocol_label", "text": "CH04-F10 — IGNORING MATERIAL DEFICIENCY", "sourceParagraph": 1071 }, { "type": "paragraph", "text": "Even if the official sentence is correct, the necessary boundary is not considered.", "sourceParagraphs": [ 1072 ] }, { "type": "protocol_label", "text": "CH04-F11 — CONSIDERING REFERENCE ERROR AS MODEL DEFECT", "sourceParagraph": 1073 }, { "type": "paragraph", "text": "AI alone is blamed for contradictory price or identity records.", "sourceParagraphs": [ 1074 ] }, { "type": "protocol_label", "text": "CH04-F12 — HIDING REFERENCE CONFLICT", "sourceParagraph": 1075 }, { "type": "paragraph", "text": "Adjudicators choose the source they want from among contradictory records after the result.", "sourceParagraphs": [ 1076 ] }, { "type": "protocol_label", "text": "CH04-F13 — INDEPENDENT SOURCE HIERARCHY REGARDLESS OF CLAIM TYPE", "sourceParagraph": 1077 }, { "type": "paragraph", "text": "In any case, it is assumed that the official or independent source is superior.", "sourceParagraphs": [ 1078 ] }, { "type": "protocol_label", "text": "CH04-F14 — DELETING COUNTEREVIDENCE", "sourceParagraph": 1079 }, { "type": "paragraph", "text": "Records that weaken a positive claim are removed from the verification file.", "sourceParagraphs": [ 1080 ] }, { "type": "protocol_label", "text": "CH04-F15 — CONSIDERING OLD OFFICIAL RECORDS AS CURRENT", "sourceParagraph": 1081 }, { "type": "paragraph", "text": "Archives, old PDFs, or press releases are made the current canonical source.", "sourceParagraphs": [ 1082 ] }, { "type": "protocol_label", "text": "CH04-F16 — COUNTING UNVERIFIED CLAIMS AS PASS", "sourceParagraph": 1083 }, { "type": "paragraph", "text": "Since the falsity is not proven, the claim is considered verified.", "sourceParagraphs": [ 1084 ] }, { "type": "protocol_label", "text": "CH04-F17 — OVERESTIMATING CLAIM STRENGTH FROM EVIDENCE", "sourceParagraph": 1085 }, { "type": "paragraph", "text": "Limited pilot or local data is generalised to universal results.", "sourceParagraphs": [ 1086 ] }, { "type": "protocol_label", "text": "CH04-F18 — SINGLE-LAYER GEO SCORE", "sourceParagraph": 1087 }, { "type": "paragraph", "text": "Accuracy verified with official fidelity is combined invisibly within a single result.", "sourceParagraphs": [ 1088 ] }, { "type": "protocol_label", "text": "CH04-F19 — PRESERVING FAVOURABLE BIAS", "sourceParagraph": 1089 }, { "type": "paragraph", "text": "No correction is made when the AI institution overstates.", "sourceParagraphs": [ 1090 ] }, { "type": "protocol_label", "text": "CH04-F20 — COUNTING REALITY RECORD AS FOUNDING DECISION", "sourceParagraph": 1091 }, { "type": "paragraph", "text": "Statement of NobleJackal or Kaan MURAZ is transformed into verified reality without external evidence.", "sourceParagraphs": [ 1092 ] }, { "type": "heading2", "text": "38. AUDIT PROCEDURE", "sourceParagraph": 1094 }, { "type": "heading3", "text": "Step 1 — Extract Material AI Claims", "sourceParagraph": 1095 }, { "type": "paragraph", "text": "Response:", "sourceParagraphs": [ 1096 ] }, { "type": "bullet", "text": "atomic claims,", "sourceParagraph": 1097 }, { "type": "bullet", "text": "relationships,", "sourceParagraph": 1098 }, { "type": "bullet", "text": "scope and time areas", "sourceParagraph": 1099 }, { "type": "paragraph", "text": "are separated.", "sourceParagraphs": [ 1100 ] }, { "type": "heading3", "text": "Step 2 — Determine the Subject of Each Claim", "sourceParagraph": 1102 }, { "type": "paragraph", "text": "The claim is about which:", "sourceParagraphs": [ 1103 ] }, { "type": "bullet", "text": "company,", "sourceParagraph": 1104 }, { "type": "bullet", "text": "brand,", "sourceParagraph": 1105 }, { "type": "bullet", "text": "product,", "sourceParagraph": 1106 }, { "type": "bullet", "text": "person,", "sourceParagraph": 1107 }, { "type": "bullet", "text": "local entity", "sourceParagraph": 1108 }, { "type": "paragraph", "text": "?", "sourceParagraphs": [ 1109 ] }, { "type": "heading3", "text": "Step 3 — Find the Official Representation Record", "sourceParagraph": 1111 }, { "type": "paragraph", "text": "Is the claim found on official surfaces? In which source? In which version? On which date? In which scope?", "sourceParagraphs": [ 1112, 1113, 1114, 1115, 1116 ] }, { "type": "heading3", "text": "Step 4 — Determine the Official Status", "sourceParagraph": 1118 }, { "type": "paragraph", "text": "Claim:", "sourceParagraphs": [ 1119 ] }, { "type": "bullet", "text": "canonical,", "sourceParagraph": 1120 }, { "type": "bullet", "text": "historical,", "sourceParagraph": 1121 }, { "type": "bullet", "text": "withdrawn,", "sourceParagraph": 1122 }, { "type": "bullet", "text": "draft,", "sourceParagraph": 1123 }, { "type": "bullet", "text": "conflicting,", "sourceParagraph": 1124 }, { "type": "bullet", "text": "unknown", "sourceParagraph": 1125 }, { "type": "paragraph", "text": "Which of their statuses is it?", "sourceParagraphs": [ 1126 ] }, { "type": "heading3", "text": "Step 5 — Classify the Type of Claim", "sourceParagraph": 1128 }, { "type": "paragraph", "text": "Claim:", "sourceParagraphs": [ 1129 ] }, { "type": "bullet", "text": "identity,", "sourceParagraph": 1130 }, { "type": "bullet", "text": "price,", "sourceParagraph": 1131 }, { "type": "bullet", "text": "scope,", "sourceParagraph": 1132 }, { "type": "bullet", "text": "licence,", "sourceParagraph": 1133 }, { "type": "bullet", "text": "performance,", "sourceParagraph": 1134 }, { "type": "bullet", "text": "leadership,", "sourceParagraph": 1135 }, { "type": "bullet", "text": "partnership,", "sourceParagraph": 1136 }, { "type": "bullet", "text": "user result", "sourceParagraph": 1137 }, { "type": "paragraph", "text": "Which of its types is it?", "sourceParagraphs": [ 1138 ] }, { "type": "heading3", "text": "Step 6 — Determine the Appropriate Type of Evidence", "sourceParagraph": 1140 }, { "type": "paragraph", "text": "Which source is truly authoritative for a claim? A single universal source ranking is not used.", "sourceParagraphs": [ 1141, 1142 ] }, { "type": "heading3", "text": "Step 7 — Collect Verification and Counter-Evidence", "sourceParagraph": 1144 }, { "type": "paragraph", "text": "Supporting records Conflicting records Timeliness Scope Method Independence are evaluated.", "sourceParagraphs": [ 1145, 1146, 1147, 1148, 1149, 1150, 1151 ] }, { "type": "heading3", "text": "Step 8 — Assign Verification Status", "sourceParagraph": 1153 }, { "type": "paragraph", "text": "VR-0 through VR-5, VR-C, VR-U, VR-T, or VR-S statuses are assigned as appropriate.", "sourceParagraphs": [ 1154 ] }, { "type": "heading3", "text": "Step 9 — Review Source Status Protection", "sourceParagraph": 1156 }, { "type": "paragraph", "text": "AI: Has it preserved the difference between “the company says” and “proven”? Has it elevated self-declaration to independent reality?", "sourceParagraphs": [ 1157, 1158, 1159 ] }, { "type": "heading3", "text": "Step 10 — Examine Material Deficiencies", "sourceParagraph": 1161 }, { "type": "paragraph", "text": "In the response or official representation, necessary for user decision:", "sourceParagraphs": [ 1162 ] }, { "type": "bullet", "text": "boundary,", "sourceParagraph": 1163 }, { "type": "bullet", "text": "time,", "sourceParagraph": 1164 }, { "type": "bullet", "text": "geography,", "sourceParagraph": 1165 }, { "type": "bullet", "text": "licence,", "sourceParagraph": 1166 }, { "type": "bullet", "text": "data method", "sourceParagraph": 1167 }, { "type": "paragraph", "text": "missing?", "sourceParagraphs": [ 1168 ] }, { "type": "heading3", "text": "Step 11 — Separate Reference Defect", "sourceParagraph": 1170 }, { "type": "paragraph", "text": "If the official or verified record is flawed in itself, a finding is created separate from the AI error.", "sourceParagraphs": [ 1171 ] }, { "type": "heading3", "text": "Step 12 — Provide Dual-Layered Result", "sourceParagraph": 1173 }, { "type": "paragraph", "text": "Response:", "sourceParagraphs": [ 1174 ] }, { "type": "heading3", "text": "DUAL-PASS", "sourceParagraph": 1175 }, { "type": "heading3", "text": "OFFICIAL-PASS / VERIFIED-FAIL", "sourceParagraph": 1176 }, { "type": "heading3", "text": "OFFICIAL-FAIL / VERIFIED-PASS", "sourceParagraph": 1177 }, { "type": "heading3", "text": "DUAL-FAIL", "sourceParagraph": 1178 }, { "type": "heading3", "text": "REFERENCE-CONFLICT", "sourceParagraph": 1179 }, { "type": "heading3", "text": "APPROPRIATE-UNCERTAINTY", "sourceParagraph": 1180 }, { "type": "paragraph", "text": "is saved with an appropriate status.", "sourceParagraphs": [ 1181 ] }, { "type": "heading2", "text": "39. REQUIRED EVIDENCE", "sourceParagraph": 1183 }, { "type": "paragraph", "text": "Canonical Entity Record Official Representation Record Verified Entity Reality Record Official web pages Official PDFs and documents Structured data Canonical JSON or API records Legal and regulatory records Licence and certificate verifications Product, price, and checkout records Contract and operation records Customer and partner verifications Performance data sets Methodology and denominators Independent reviews Sponsor and conflict of interest Counter-evidence Negative results Errata and retraction records Official source precedence rules Verification statuses Adjudicator decisions Appeal records Measurement date Reference versions Responsible person or institution", "sourceParagraphs": [ 1184, 1185, 1186, 1187, 1188, 1189, 1190, 1191, 1192, 1193, 1194, 1195, 1196, 1197, 1198, 1199, 1200, 1201, 1202, 1203, 1204, 1205, 1206, 1207, 1208, 1209 ] }, { "type": "heading2", "text": "40. AUDIT CHECKLIST", "sourceParagraph": 1211 }, { "type": "paragraph", "text": "Is there an Official Representation Record? Is there a Verified Entity Reality Record? Were both versioned for the same measurement date? Is the subject of each AI claim correct? Was the claim officially published? Is the official source current and canonical? Do official records conflict with one another? Was the appropriate evidence source used for the type of claim? Was a first-party source rejected automatically? Was a first-party statement automatically treated as independent evidence? Is the independent source genuinely relevant? Were the claim's scope and time preserved? Does the file contain counter-evidence? Was a self-declaration elevated to verified reality? Was the company's view conveyed with its source status intact? Is a material boundary or exclusion missing? Was an error favourable to the entity recorded as such?", "sourceParagraphs": [ 1212, 1213, 1214, 1215, 1216, 1217, 1218, 1219, 1220, 1221, 1222, 1223, 1224, 1225, 1226, 1227, 1228 ] }, { "type": "paragraph", "text": "Has the reference defect been separated from the AI error? Have the UNKNOWN and UNRESOLVED fields been considered positive? Does high official fidelity hide low verified accuracy? Can the AI's deviation from the official record be explained with strong evidence? Has the reference record changed during measurement? Are the old and new observations in separate versions? Does the dual-layer result appear in the public record? Is the accountable owner of the result known?", "sourceParagraphs": [ 1229, 1230, 1231, 1232, 1233, 1234, 1235, 1236 ] }, { "type": "heading2", "text": "41. OBJECTIONS AND RESPONSES", "sourceParagraph": 1238 }, { "type": "heading3", "text": "Objection 1 — “The company knows the most accurate information about itself.”", "sourceParagraph": 1239 }, { "type": "paragraph", "text": "This may be true in many respects. The company:", "sourceParagraphs": [ 1240, 1241 ] }, { "type": "bullet", "text": "its own product,", "sourceParagraph": 1242 }, { "type": "bullet", "text": "its price,", "sourceParagraph": 1243 }, { "type": "bullet", "text": "its support policy,", "sourceParagraph": 1244 }, { "type": "bullet", "text": "its operation", "sourceParagraph": 1245 }, { "type": "paragraph", "text": "knows it well. However, the company is also the commercial owner of the claim. Sentences that require external comparison, such as “the best,” “leader,” “most reliable,” and “proven,” cannot be verified by self-declaration alone.", "sourceParagraphs": [ 1246, 1247, 1248 ] }, { "type": "heading3", "text": "Objection 2 — “If you do not count your own site as evidence, how will we know the price?”", "sourceParagraph": 1250 }, { "type": "paragraph", "text": "The official price page is the primary source for the price the institution offers to the public. However:", "sourceParagraphs": [ 1251, 1252 ] }, { "type": "bullet", "text": "checkout,", "sourceParagraph": 1253 }, { "type": "bullet", "text": "tax,", "sourceParagraph": 1254 }, { "type": "bullet", "text": "additional fee,", "sourceParagraph": 1255 }, { "type": "bullet", "text": "contract", "sourceParagraph": 1256 }, { "type": "paragraph", "text": "if it conflicts, the actual offer is evaluated separately. The statement “own site is not evidence” cannot be applied equally to all types of claims. The correct ruling is:", "sourceParagraphs": [ 1257, 1258, 1259 ] }, { "type": "callout", "text": "A company's own site can be a primary source for statements under its control; it is not automatic independent verification for external performance and superiority claims.", "sourceParagraph": 1260 }, { "type": "heading3", "text": "Objection 3 — “If AI reflects the official site, hasn't it completed its task?”", "sourceParagraph": 1262 }, { "type": "paragraph", "text": "If the only goal is: “What does the site say?” then yes. However, if the user asks: “What is the company actually doing?” merely repeating the site may not be sufficient. The type of task affects the measurement layer.", "sourceParagraphs": [ 1263, 1264, 1265, 1266, 1267, 1268, 1269 ] }, { "type": "heading3", "text": "Objection 4 — “External sources can also be wrong.”", "sourceParagraph": 1271 }, { "type": "paragraph", "text": "That is correct. Therefore, independence is not automatic accuracy. The source is evaluated in terms of:", "sourceParagraphs": [ 1272, 1273, 1274 ] }, { "type": "bullet", "text": "authority,", "sourceParagraph": 1275 }, { "type": "bullet", "text": "proximity,", "sourceParagraph": 1276 }, { "type": "bullet", "text": "currency,", "sourceParagraph": 1277 }, { "type": "bullet", "text": "method,", "sourceParagraph": 1278 }, { "type": "bullet", "text": "scope", "sourceParagraph": 1279 }, { "type": "paragraph", "text": "in terms of maintenance.", "sourceParagraphs": [ 1280 ] }, { "type": "heading3", "text": "Objection 5 — “Who will determine the verified record of reality?”", "sourceParagraph": 1282 }, { "type": "paragraph", "text": "The company alone cannot determine it. NOMOS alone cannot determine it either. The process:", "sourceParagraphs": [ 1283, 1284, 1285 ] }, { "type": "bullet", "text": "evidence holder,", "sourceParagraph": 1286 }, { "type": "bullet", "text": "adjudicator,", "sourceParagraph": 1287 }, { "type": "bullet", "text": "appropriate expert,", "sourceParagraph": 1288 }, { "type": "bullet", "text": "appeal,", "sourceParagraph": 1289 }, { "type": "bullet", "text": "version record,", "sourceParagraph": 1290 }, { "type": "bullet", "text": "accountable person", "sourceParagraph": 1291 }, { "type": "paragraph", "text": "is required. More independent review is needed for high-risk and conflicted claims.", "sourceParagraphs": [ 1292, 1293 ] }, { "type": "heading3", "text": "Objection 6 — “If the reality record can always change, wouldn’t the score be meaningless?”", "sourceParagraph": 1295 }, { "type": "paragraph", "text": "No. The score:", "sourceParagraphs": [ 1296, 1297 ] }, { "type": "bullet", "text": "date,", "sourceParagraph": 1298 }, { "type": "bullet", "text": "is meaningful with", "sourceParagraph": 1299 }, { "type": "bullet", "text": "scope", "sourceParagraph": 1300 }, { "type": "paragraph", "text": "the version. Being subject to change does not mean it cannot be measured. It means it cannot be presented prematurely.", "sourceParagraphs": [ 1301, 1302, 1303 ] }, { "type": "heading3", "text": "Objection 7 — “Why would AI use negative information that is not on the official site?”", "sourceParagraph": 1305 }, { "type": "paragraph", "text": "Information:", "sourceParagraphs": [ 1306 ] }, { "type": "bullet", "text": "can be forced into the categories of\n\ntrue,", "sourceParagraph": 1307 }, { "type": "bullet", "text": "current,", "sourceParagraph": 1308 }, { "type": "bullet", "text": "relevant,", "sourceParagraph": 1309 }, { "type": "bullet", "text": "financial for user decision", "sourceParagraph": 1310 }, { "type": "paragraph", "text": "can be used. However, the source and context must be preserved. The silence of the official site does not eliminate the negative reality.", "sourceParagraphs": [ 1311, 1312, 1313 ] }, { "type": "heading3", "text": "Objection 8 — 'If AI reports the company's claim as 'the company says,' it does not provide value to the user.'", "sourceParagraph": 1315 }, { "type": "paragraph", "text": "On the contrary, it correctly displays the claim status. User:", "sourceParagraphs": [ 1316, 1317 ] }, { "type": "bullet", "text": "the institution's opinion,", "sourceParagraph": 1318 }, { "type": "bullet", "text": "the proven finding,", "sourceParagraph": 1319 }, { "type": "bullet", "text": "open ambiguity", "sourceParagraph": 1320 }, { "type": "paragraph", "text": "can distinguish. Looking certain is not the same as looking correct.", "sourceParagraphs": [ 1321, 1322 ] }, { "type": "heading3", "text": "Objection 9 — \"Why should we correct minor mistakes that are in favour of the company?\"", "sourceParagraph": 1324 }, { "type": "paragraph", "text": "Because of the wrong expectation:", "sourceParagraphs": [ 1325 ] }, { "type": "bullet", "text": "wrong customer,", "sourceParagraph": 1326 }, { "type": "bullet", "text": "delivery failure,", "sourceParagraph": 1327 }, { "type": "bullet", "text": "return,", "sourceParagraph": 1328 }, { "type": "bullet", "text": "complaint,", "sourceParagraph": 1329 }, { "type": "bullet", "text": "loss of trust", "sourceParagraph": 1330 }, { "type": "paragraph", "text": "It can produce. The advantage is apparent short-term visibility; long-term is the duty of representation.", "sourceParagraphs": [ 1331, 1332 ] }, { "type": "heading3", "text": "Objection 10 — “Don’t two separate scores complicate public communication?”", "sourceParagraph": 1334 }, { "type": "paragraph", "text": "Concealing all this complexity creates the greater risk. A concise result card may instead state: Official Representation Fidelity — high / Verified Entity Reality Accuracy — medium / Principal gaps — unsupported leadership and country-coverage claims. That is clearer than one misleading number.", "sourceParagraphs": [ 1335, 1336, 1337, 1338 ] }, { "type": "heading2", "text": "1. the Official Representation Record; and", "sourceParagraph": 1395 }, { "type": "heading2", "text": "2. the Evidence-Constrained Verified Entity Reality Record.", "sourceParagraph": 1396 }, { "type": "paragraph", "text": "Official fidelity and verified accuracy MUST NOT be treated as the same result. An AI response MAY faithfully reproduce an official claim while remaining factually unsupported, contradicted, outdated, overbroad, or materially misleading. An AI response MAY differ from an official record without failing when the official record is stale, incomplete, conflicted, or contradicted by stronger and more appropriate evidence. Self-claims, first-party data, independent evidence, sponsored content, official registries, transactional records, and third-party statements MUST retain their source and evidence status. Reference conflicts MUST be recorded separately from AI-response failures. Unknown or unresolved evidence MUST NOT be converted into verified truth.", "sourceParagraphs": [ 1398, 1399, 1401, 1402, 1403, 1405, 1406, 1407, 1409, 1410, 1411, 1413, 1414 ] }, { "type": "paragraph", "text": "Every material generative-system claim about the audited entity MUST be evaluated against two separately versioned reference layers: the Official Representation Record and the Evidence-Constrained Verified Entity Reality Record. Official fidelity and verified accuracy MUST NOT be treated as the same result. An AI product may faithfully reproduce an official claim while creating a materially unsupported, contradicted, outdated or overbroad representation. When the official record is stale, incomplete or conflicted, a different response grounded in stronger evidence is not automatically an error. Source type and evidential status MUST be preserved, and reference conflicts MUST be recorded separately from AI failures. High Official Representation Fidelity MUST NOT compensate for low Verified Entity Reality Accuracy.", "sourceParagraphs": [ 1416, 1417, 1418, 1419 ] }, { "type": "heading2", "text": "COMMON JUDGEMENT OF CHAPTER 44", "sourceParagraph": 1421 }, { "type": "paragraph", "text": "A company has the right to speak about itself. It can explain who it is. It can describe what it offers. It can publish its prices. It can set its boundaries. It can convey its vision. However, not every sentence an entity says about itself has the same epistemic status. “The price of our service package is 2,000 euros.” and “We are the best company in the world.” are not the same type of claim. The first sentence may be a commercial statement controlled by the company. The second sentence requires external comparison. If an AI product conveys both with equal certainty, it undermines the source status. An AI product can perfectly replicate a company website. But the site:", "sourceParagraphs": [ 1422, 1423, 1424, 1425, 1426, 1427, 1428, 1429, 1430, 1431, 1432, 1433, 1434, 1435, 1436, 1437 ] }, { "type": "bullet", "text": "old,", "sourceParagraph": 1438 }, { "type": "bullet", "text": "conflicting,", "sourceParagraph": 1439 }, { "type": "bullet", "text": "unproven,", "sourceParagraph": 1440 }, { "type": "bullet", "text": "exceeding the scope,", "sourceParagraph": 1441 }, { "type": "bullet", "text": "conceals material boundaries", "sourceParagraph": 1442 }, { "type": "paragraph", "text": "then fidelity to that record is not good GEO. Conversely, an AI product may depart from the official website and still be correct when its response rests on stronger, current and appropriate evidence. The departure may be a correction rather than an error. NOMOS's fourth law of measurement is therefore:", "sourceParagraphs": [ 1443, 1444, 1445, 1446 ] }, { "type": "callout", "text": "Official representation does not replace verified reality.", "sourceParagraph": 1447 }, { "type": "paragraph", "text": "The fifth law is as follows:", "sourceParagraphs": [ 1448 ] }, { "type": "callout", "text": "Verified reality cannot be established by ignoring the official source.", "sourceParagraph": 1449 }, { "type": "paragraph", "text": "Its sixth law is:", "sourceParagraphs": [ 1450 ] }, { "type": "callout", "text": "Representation depends not only on what a source says, but also on the authority under which it speaks and the evidential status of the statement.", "sourceParagraph": 1451 }, { "type": "paragraph", "text": "Its seventh law is as follows:", "sourceParagraphs": [ 1452 ] }, { "type": "callout", "text": "Faithfully reproducing an official error is not accurate representation.", "sourceParagraph": 1453 }, { "type": "paragraph", "text": "Its eighth law is as follows:", "sourceParagraphs": [ 1454 ] }, { "type": "callout", "text": "The record of reality is not an indisputable absolute judgement; it is a versioned and evidence-based decision file.", "sourceParagraph": 1455 }, { "type": "heading2", "text": "Order of NOMOS Section 4", "sourceParagraph": 1457 }, { "type": "callout", "text": "Do not just show me what is written on your site. / Show which sentence is supported by which evidence.", "sourceParagraph": 1458 }, { "type": "callout", "text": "Do not present your own words as if they are independently verified.", "sourceParagraph": 1459 }, { "type": "callout", "text": "Do not devalue first-party data just because it belongs to you. / But do not make it an indisputable truth just because it belongs to you.", "sourceParagraph": 1460 }, { "type": "callout", "text": "Do not turn company opinion into fact, press release into news, or sponsored award into independent authority.", "sourceParagraph": 1461 }, { "type": "callout", "text": "Do not count me as successful when I correctly repeat the official error.", "sourceParagraph": 1462 }, { "type": "callout", "text": "If your official record is outdated, do not penalise an answer supported by current evidence merely because it differs from that record.", "sourceParagraph": 1463 }, { "type": "callout", "text": "Do not remain silent about the mistake in your favour.", "sourceParagraph": 1464 }, { "type": "callout", "text": "Do not remove counter-evidence from the file.", "sourceParagraph": 1465 }, { "type": "callout", "text": "When your references conflict, do not invent a single truth. / Record the conflict. / If the answer is unknown, say so.", "sourceParagraph": 1466 }, { "type": "callout", "text": "Separate your claim, its source, its evidence, its limit, and its time from each other.", "sourceParagraph": 1467 }, { "type": "callout", "text": "Before ordering me what to say, prove what you have the right to say.", "sourceParagraph": 1468 }, { "type": "heading2", "text": "The Chapter's Closing Sentence", "sourceParagraph": 1470 }, { "type": "callout", "text": "Good GEO is not the faithful repetition of whatever an entity says. It is the accurate transmission of what has been said, without distorting its source, evidence, scope or time.", "sourceParagraph": 1471 } ] }, { "number": 5, "id": "NOMOS-GEO-AUDIT-CH05", "title": "The Relevant User Population for Each AI Product", "subtitle": null, "sourceFile": "5.ci bölüm.docx", "sourceSha256": "24C9C1F8A3E953064C339B99D62B78BB786E545011B3A075AA69FD173C77EE9D", "sourceWordCount": 7842, "sourceIds": [ "K04", "K05", "K06", "K07" ], "blocks": [ { "type": "heading2", "text": "Chapter Boundary", "sourceParagraph": 16 }, { "type": "paragraph", "text": "The first four chapters established the basic objects of measurement: a single response is not a GEO score; representation is a conditional distribution, not a single number; the audited entity must be resolved before results are observed; and an AI response must be evaluated against both official representation and evidence-based verified reality. We must now define the denominator of the measurement:", "sourceParagraphs": [ 17, 18, 19, 20, 21, 22 ] }, { "type": "callout", "text": "About which human population will this distribution make a statement?", "sourceParagraph": 23 }, { "type": "paragraph", "text": "Everyone living in a country? Only adults? Those with internet access? Those who can officially use a specific AI product? Free account holders? Paid plan users? People who can use the product in the oversight language? Truly active users? Potential users who might use it one day? People in markets where the brand can provide services? If a country's population is very large but the relevant AI product is not officially accessible in that country, how will that country be included in the main product panel? If the population share of a small country is far below one person in a sample of 1,000, will that country be completely invisible?", "sourceParagraphs": [ 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35 ] }, { "type": "paragraph", "text": "If one person knows three languages, should that person be counted as three people? To which population does a user belong if they can access a product's mobile surface but not its web surface? Until these questions are answered, ‘1,000 random users for each AI product’ does not define a scientific target population; it states only a headcount. This chapter does not discard country population. It assigns it its proper role:", "sourceParagraphs": [ 36, 37, 38, 39, 40, 41, 42, 43 ] }, { "type": "callout", "text": "The country population is the initial base for sampling; it is not the final eligible user population.", "sourceParagraph": 44 }, { "type": "paragraph", "text": "In this section, the audit system directly applies the previously determined requirement for country, language, date, product, and measurement record to the population definition. Because requesting evidence, boundaries, context, and time also requires defining with the same clarity which people are being referred to. This section has not yet:", "sourceParagraphs": [ 45, 46 ] }, { "type": "bullet", "text": "It does not finalise how the 1,000-person sample will be distributed across countries,", "sourceParagraph": 47 }, { "type": "bullet", "text": "which algorithm will be used to round fractions to whole numbers,", "sourceParagraph": 48 }, { "type": "bullet", "text": "from which panel participants will be selected,", "sourceParagraph": 49 }, { "type": "bullet", "text": "how sampling weights will be calculated,", "sourceParagraph": 50 }, { "type": "bullet", "text": "or which variance method will be used to produce confidence intervals", "sourceParagraph": 51 }, { "type": "paragraph", "text": "These will be arranged in the following sections. The task of Section 5 is more basic:", "sourceParagraphs": [ 52, 53, 54 ] }, { "type": "callout", "text": "For each AI product and each measurement question, to define which human population the results can be generalised to.", "sourceParagraph": 55 }, { "type": "heading2", "text": "NOMOS Challenge", "sourceParagraph": 57 }, { "type": "paragraph", "text": "You open a world map. You add up the populations of the countries. Then you want to distribute 1,000 people according to the population. Large countries will get many participants. Small countries will get one or close to zero shares. The core of this idea is correct:", "sourceParagraphs": [ 58, 59, 60, 61, 62, 63 ] }, { "type": "callout", "text": "A person in the world population should not be hundreds of times more influential than a person in a large country just because they live in a small country.", "sourceParagraph": 64 }, { "type": "paragraph", "text": "But now apply the same method to two different AI products. The first product:", "sourceParagraphs": [ 65, 66 ] }, { "type": "bullet", "text": "can be used officially in some countries,", "sourceParagraph": 67 }, { "type": "bullet", "text": "is limited in some countries.", "sourceParagraph": 68 }, { "type": "paragraph", "text": "It is not offered at all in some countries. The country coverage of the second product is different. What happens if you divide the world's population in the same way between the two products? In countries where the first product is not available, you may ask people for measurements. Participants:", "sourceParagraphs": [ 69, 70, 71, 72, 73 ] }, { "type": "bullet", "text": "may be encouraged to misrepresent their locations,", "sourceParagraph": 74 }, { "type": "bullet", "text": "use prohibited or unsupported access methods,", "sourceParagraph": 75 }, { "type": "bullet", "text": "open an account in another country,", "sourceParagraph": 76 }, { "type": "bullet", "text": "bypass security and usage conditions", "sourceParagraph": 77 }, { "type": "paragraph", "text": "This does not measure the real user experience. It generates an artificial access experience. Now consider another problem. In a certain country, the population is very large. However:", "sourceParagraphs": [ 78, 79, 80, 81, 82, 83 ] }, { "type": "bullet", "text": "a significant portion of the population does not meet the minimum age requirement of the product.", "sourceParagraph": 84 }, { "type": "bullet", "text": "not everyone has reliable internet access,", "sourceParagraph": 85 }, { "type": "bullet", "text": "the product can only be used on certain devices,", "sourceParagraph": 86 }, { "type": "bullet", "text": "the audit language cannot be used by the entire population,", "sourceParagraph": 87 }, { "type": "paragraph", "text": "a specific plan is only available to users with a payment method. If you make the total population the denominator directly, you include people who actually cannot be monitored and cannot use the product in the target population. Now imagine a small country. The theoretical share in a sample of 1,000 under population weighting: 0.03 people.", "sourceParagraphs": [ 88, 89, 90, 91, 92, 93 ] }, { "type": "paragraph", "text": "One-third of a person cannot be selected. If you give one person to the country, you can overrepresent the country in the population-weighted main panel. If you give no one, then in that country:", "sourceParagraphs": [ 94, 95, 96 ] }, { "type": "bullet", "text": "access,", "sourceParagraph": 97 }, { "type": "bullet", "text": "language,", "sourceParagraph": 98 }, { "type": "bullet", "text": "product behaviour,", "sourceParagraph": 99 }, { "type": "bullet", "text": "local representation", "sourceParagraph": 100 }, { "type": "paragraph", "text": "can remain completely invisible. These two goals are different:", "sourceParagraphs": [ 101, 102 ] }, { "type": "bullet", "text": "To measure the average user experience in the world population", "sourceParagraph": 103 }, { "type": "bullet", "text": "To prevent small countries from being completely unmonitored", "sourceParagraph": 104 }, { "type": "paragraph", "text": "The same single panel cannot perfectly fulfil both goals. Therefore, NOMOS establishes three separate principles:", "sourceParagraphs": [ 105, 106 ] }, { "type": "callout", "text": "Population weight determines people's share in the global average.", "sourceParagraph": 107 }, { "type": "callout", "text": "Observer coverage prevents small countries and languages from being completely invisible.", "sourceParagraph": 108 }, { "type": "callout", "text": "Language and accessibility justice further measures subgroups that may be suppressed by the population average.", "sourceParagraph": 109 }, { "type": "paragraph", "text": "The first premise of this section is:", "sourceParagraphs": [ 110 ] }, { "type": "callout", "text": "The total population of a country is not the AI-eligible user population in that country.", "sourceParagraph": 111 }, { "type": "paragraph", "text": "Its second provision states:", "sourceParagraphs": [ 112 ] }, { "type": "callout", "text": "The target population of an AI product cannot be automatically transferred to another AI product.", "sourceParagraph": 113 }, { "type": "paragraph", "text": "Its third provision states:", "sourceParagraphs": [ 114 ] }, { "type": "callout", "text": "A person's weight in the main score is not the same as their country's right to appear in the audit.", "sourceParagraph": 115 }, { "type": "paragraph", "text": "Its fourth provision states:", "sourceParagraphs": [ 116 ] }, { "type": "callout", "text": "The lack of product access is not a misrepresentation; however, it is a global access limitation and should also be reported.", "sourceParagraph": 117 }, { "type": "heading2", "text": "1. PURPOSE OF THE CHAPTER", "sourceParagraph": 119 }, { "type": "paragraph", "text": "The purpose of this section is to define the target human population of the GEO-1000 measurement in terms of product, question, country, language, interface, and time. The section standardises the following distinctions:", "sourceParagraphs": [ 120, 121 ] }, { "type": "bullet", "text": "Total country population and suitable AI user population", "sourceParagraph": 122 }, { "type": "bullet", "text": "Potential access and actual usage", "sourceParagraph": 123 }, { "type": "bullet", "text": "Product accessibility and representational accuracy", "sourceParagraph": 124 }, { "type": "bullet", "text": "Target population and sampling frame", "sourceParagraph": 125 }, { "type": "bullet", "text": "Valid analytical observation with the selected participant", "sourceParagraph": 126 }, { "type": "bullet", "text": "Account with person", "sourceParagraph": 127 }, { "type": "bullet", "text": "Session with person", "sourceParagraph": 128 }, { "type": "bullet", "text": "Core model with product", "sourceParagraph": 129 }, { "type": "bullet", "text": "Free and paid plan populations", "sourceParagraph": 130 }, { "type": "bullet", "text": "Web and mobile user populations", "sourceParagraph": 131 }, { "type": "bullet", "text": "Usage exceeding access restriction with official access", "sourceParagraph": 132 }, { "type": "bullet", "text": "Country Observer coverage with country population weight", "sourceParagraph": 133 }, { "type": "bullet", "text": "Language Fairness Panel with population-weighted language share", "sourceParagraph": 134 }, { "type": "bullet", "text": "Active user exposure with potential user population", "sourceParagraph": 135 }, { "type": "bullet", "text": "Market or jurisdiction domain query with global identity query", "sourceParagraph": 136 }, { "type": "bullet", "text": "Corporate, education, or developer product with main consumer product", "sourceParagraph": 137 }, { "type": "bullet", "text": "Accessibility of the data collection tool for users with accessibility needs", "sourceParagraph": 138 }, { "type": "bullet", "text": "Users outside the target population and NOT TESTED users", "sourceParagraph": 139 }, { "type": "bullet", "text": "Countries where the product is unusable and countries where incorrect responses are received in the product", "sourceParagraph": 140 }, { "type": "paragraph", "text": "At the end of this section, each GEO-1000 audit should be able to explicitly answer the following question:", "sourceParagraphs": [ 141 ] }, { "type": "callout", "text": "What user universe, selected from which people, does this result make a judgement about?", "sourceParagraph": 142 }, { "type": "heading2", "text": "2. CENTRAL NORMATIVE PROVISION", "sourceParagraph": 144 }, { "type": "callout", "text": "Before results are observed, each GEO-1000 audit must define and version a target user population specific to the audited AI product, measurement question, country, language, interface, plan, and time conditions. [K04–K07]", "sourceParagraph": 145 }, { "type": "paragraph", "text": "Target population:", "sourceParagraphs": [ 146 ] }, { "type": "bullet", "text": "\"world population\"", "sourceParagraph": 147 }, { "type": "bullet", "text": "\"internet users\",", "sourceParagraph": 148 }, { "type": "bullet", "text": "\"AI users\",", "sourceParagraph": 149 }, { "type": "bullet", "text": "\"customers\"", "sourceParagraph": 150 }, { "type": "paragraph", "text": "The population cannot be left under an ambiguous label such as ‘relevant users’. As applicable, the following questions must be answered: Which AI product? Which user surface? Which countries? Access as of what date? What age condition? Which language or languages? Which device and interface? Which plan or account type? What are the official terms of use? Must users pay to access the product? Does the measurement question apply to all users or only to a particular market or jurisdiction? Are active users or potential users being measured? Which user groups are out of scope? Which groups belong to the target population but are absent from the sampling frame? A change in any of these fields may change the target population. When it does, old and new scores cannot be combined in the same series without explanation.", "sourceParagraphs": [ 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168 ] }, { "type": "heading2", "text": "3. THE POPULATION IS NOT A SINGLE NUMBER", "sourceParagraph": 170 }, { "type": "paragraph", "text": "The official or estimated total population of a country is only the starting layer for GEO-1000. The audit population can consist of the following narrowing layers.", "sourceParagraphs": [ 171, 172 ] }, { "type": "heading3", "text": "3.1. Resident Population", "sourceParagraph": 174 }, { "type": "paragraph", "text": "The total human population living in the specified country or jurisdiction. This number:", "sourceParagraphs": [ 175, 176 ] }, { "type": "bullet", "text": "does not imply any suitability in terms of age,", "sourceParagraph": 177 }, { "type": "bullet", "text": "internet,", "sourceParagraph": 178 }, { "type": "bullet", "text": "product access,", "sourceParagraph": 179 }, { "type": "bullet", "text": "language,", "sourceParagraph": 180 }, { "type": "bullet", "text": "account,", "sourceParagraph": 181 }, { "type": "bullet", "text": "device", "sourceParagraph": 182 }, { "type": "paragraph", "text": "considerations.", "sourceParagraphs": [ 183 ] }, { "type": "heading3", "text": "3.2. Age-Eligible population", "sourceParagraph": 185 }, { "type": "paragraph", "text": "The population that meets the age requirement of the relevant AI product and research protocol. For a product:", "sourceParagraphs": [ 186, 187 ] }, { "type": "bullet", "text": "minimum usage age,", "sourceParagraph": 188 }, { "type": "bullet", "text": "parental consent requirement,", "sourceParagraph": 189 }, { "type": "bullet", "text": "country-specific age rule", "sourceParagraph": 190 }, { "type": "paragraph", "text": "may exist. The research protocol may be more restrictive than the product condition. For example, the founding GEO-1000 panel may be composed of only adult participants to reduce ethical and legal complexity. In this case, the score is not: \"The entire population\":", "sourceParagraphs": [ 191, 192, 193, 194, 195, 196 ] }, { "type": "heading4", "text": "but rather \"Defined adult and product-eligible population\"", "sourceParagraph": 197 }, { "type": "paragraph", "text": "result.", "sourceParagraphs": [ 198 ] }, { "type": "heading3", "text": "3.3. Connected Population", "sourceParagraph": 200 }, { "type": "paragraph", "text": "The population that can use the relevant product at a level where:", "sourceParagraphs": [ 201 ] }, { "type": "bullet", "text": "internet,", "sourceParagraph": 202 }, { "type": "bullet", "text": "device,", "sourceParagraph": 203 }, { "type": "bullet", "text": "connection quality,", "sourceParagraph": 204 }, { "type": "bullet", "text": "technical access", "sourceParagraph": 205 }, { "type": "paragraph", "text": "is available. Internet access may not be a binary characteristic. A person:", "sourceParagraphs": [ 206, 207, 208 ] }, { "type": "bullet", "text": "can use messaging,", "sourceParagraph": 209 }, { "type": "bullet", "text": "however, they may not be able to run the heavy web application,", "sourceParagraph": 210 }, { "type": "bullet", "text": "may use mobile internet,", "sourceParagraph": 211 }, { "type": "bullet", "text": "however, they may not be able to access the desktop experience,", "sourceParagraph": 212 }, { "type": "paragraph", "text": "may not be able to upload responses or screenshots due to low connectivity. Connection limitations should not automatically exclude the user from the target population. It should be associated with the actual usage conditions of the measured product.", "sourceParagraphs": [ 213, 214, 215 ] }, { "type": "heading3", "text": "3.4. Accessible Population from the Product Perspective", "sourceParagraph": 217 }, { "type": "paragraph", "text": "For the relevant AI product on the specified date:", "sourceParagraphs": [ 218 ] }, { "type": "bullet", "text": "officially launched,", "sourceParagraph": 219 }, { "type": "bullet", "text": "could be used in accordance with the law and product conditions,", "sourceParagraph": 220 }, { "type": "bullet", "text": "the necessary interface could be accessed", "sourceParagraph": 221 }, { "type": "paragraph", "text": "It is population. The absence of a product in a country does not mean that the people in that country are misrepresented. In this case: the representation response is none, there is no access coverage. The two outcomes should be kept separate.", "sourceParagraphs": [ 222, 223, 224, 225, 226, 227 ] }, { "type": "heading3", "text": "3.5. Population Suitable in Terms of Account and Plan", "sourceParagraph": 229 }, { "type": "paragraph", "text": "Some products:", "sourceParagraphs": [ 230 ] }, { "type": "bullet", "text": "may require an account,", "sourceParagraph": 231 }, { "type": "bullet", "text": "may require phone verification,", "sourceParagraph": 232 }, { "type": "bullet", "text": "may require a payment method,", "sourceParagraph": 233 }, { "type": "bullet", "text": "may be offered only with a corporate contract,", "sourceParagraph": 234 }, { "type": "paragraph", "text": "may unlock certain features only in a paid plan. These conditions affect which portion of the target population can actually access the relevant user surface. The free plan score cannot automatically be applied to paid plan users.", "sourceParagraphs": [ 235, 236, 237 ] }, { "type": "heading3", "text": "3.6. Population Suitable in Terms of Language", "sourceParagraph": 239 }, { "type": "paragraph", "text": "The population that can use the relevant prompt and AI response at a sufficient level in the audit language. Language suitability alone:", "sourceParagraphs": [ 240, 241 ] }, { "type": "bullet", "text": "cannot be determined only by the official language of the country,", "sourceParagraph": 242 }, { "type": "bullet", "text": "the person's mother tongue", "sourceParagraph": 243 }, { "type": "paragraph", "text": "It should be distinguished:", "sourceParagraphs": [ 244, 245 ] }, { "type": "bullet", "text": "Mother or home language", "sourceParagraph": 246 }, { "type": "bullet", "text": "Language of daily use", "sourceParagraph": 247 }, { "type": "bullet", "text": "AI interface language", "sourceParagraph": 248 }, { "type": "bullet", "text": "Prompt language", "sourceParagraph": 249 }, { "type": "bullet", "text": "Response language", "sourceParagraph": 250 }, { "type": "bullet", "text": "Language for professional use", "sourceParagraph": 251 }, { "type": "bullet", "text": "Reading proficiency", "sourceParagraph": 252 }, { "type": "paragraph", "text": "A person can be suitable in multiple languages. This person cannot be counted as more than one full person in the global population calculation.", "sourceParagraphs": [ 253, 254 ] }, { "type": "heading3", "text": "3.7. Interface-Eligible population", "sourceParagraph": 256 }, { "type": "paragraph", "text": "Measurement:", "sourceParagraphs": [ 257 ] }, { "type": "bullet", "text": "web,", "sourceParagraph": 258 }, { "type": "bullet", "text": "mobile application,", "sourceParagraph": 259 }, { "type": "bullet", "text": "desktop application,", "sourceParagraph": 260 }, { "type": "bullet", "text": "corporate workspace,", "sourceParagraph": 261 }, { "type": "bullet", "text": "embedded assistant", "sourceParagraph": 262 }, { "type": "paragraph", "text": "can be used. The accessible user universe may differ for each interface. A user with web access may not have the mobile product. The corporate user surface may not be publicly available.", "sourceParagraphs": [ 263, 264, 265, 266 ] }, { "type": "heading3", "text": "3.8. Research-Eligible population", "sourceParagraph": 268 }, { "type": "paragraph", "text": "Even if a person can use the product, for inclusion in research, additionally:", "sourceParagraphs": [ 269 ] }, { "type": "bullet", "text": "voluntary consent,", "sourceParagraph": 270 }, { "type": "bullet", "text": "privacy,", "sourceParagraph": 271 }, { "type": "bullet", "text": "data processing,", "sourceParagraph": 272 }, { "type": "bullet", "text": "collecting screenshots,", "sourceParagraph": 273 }, { "type": "bullet", "text": "country-specific research requirements,", "sourceParagraph": 274 }, { "type": "bullet", "text": "does not imply any suitability in terms of age,", "sourceParagraph": 275 }, { "type": "bullet", "text": "security", "sourceParagraph": 276 }, { "type": "paragraph", "text": "conditions may need to be met. People who cannot be included in the research but use the actual product may be part of the target population. In this case, the sampling frame underrepresents the target population. This underrepresentation should not be hidden.", "sourceParagraphs": [ 277, 278, 279, 280 ] }, { "type": "heading3", "text": "3.9. Analytically Eligible population", "sourceParagraph": 282 }, { "type": "paragraph", "text": "The population for which the collected data can be used to produce results. The participant:", "sourceParagraphs": [ 283, 284 ] }, { "type": "bullet", "text": "altered the prompt,", "sourceParagraph": 285 }, { "type": "bullet", "text": "used the wrong product,", "sourceParagraph": 286 }, { "type": "bullet", "text": "chose the answer,", "sourceParagraph": 287 }, { "type": "bullet", "text": "submitted incomplete evidence", "sourceParagraph": 288 }, { "type": "paragraph", "text": "It is possible. Even if an individual belongs to the target population, their observation may not enter the analytical dataset. Valid observations and the target population are not the same thing.", "sourceParagraphs": [ 289, 290, 291 ] }, { "type": "heading2", "text": "4. DEFINITION OF THE SUITABLE USER POPULATION", "sourceParagraph": 293 }, { "type": "paragraph", "text": "For AI product a at time t, the eligible user universe may be defined as follows:", "sourceParagraphs": [ 294 ] }, { "type": "formula", "text": "U_{a,t} = {u ∈ P_t: AgeEligible(u,a,t)=1 ∧ Connected(u,a,t)=1 ∧ ProductAvailable(u,a,t)=1 ∧ LawfulAccess(u,a,t)=1 ∧ InterfaceEligible(u,a,t)=1}", "sourceParagraph": 295 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 296 ] }, { "type": "bullet", "text": "Pt: the general human population at the specified date", "sourceParagraph": 297 }, { "type": "bullet", "text": "u: individual human", "sourceParagraph": 298 }, { "type": "bullet", "text": "a: AI product and user surface", "sourceParagraph": 299 }, { "type": "bullet", "text": "t: measurement time", "sourceParagraph": 300 }, { "type": "paragraph", "text": "In terms of language and prompt, the target universe can later be narrowed down as follows:", "sourceParagraphs": [ 301 ] }, { "type": "formula", "text": "U_{a,c,l,s,p,t} = {u ∈ U_{a,t}: Country(u)=c ∧ LanguageEligible(u,l)=1 ∧ SurfaceEligible(u,s)=1 ∧ PromptRelevant(u,p)=1}", "sourceParagraph": 302 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 303 ] }, { "type": "bullet", "text": "c: country or jurisdiction", "sourceParagraph": 304 }, { "type": "bullet", "text": "l: language and locale", "sourceParagraph": 305 }, { "type": "bullet", "text": "s: user surface", "sourceParagraph": 306 }, { "type": "bullet", "text": "p: prompt or question family", "sourceParagraph": 307 }, { "type": "paragraph", "text": "The purpose of this notation is not to record all individuals one by one. It is to make the logical boundary of the target population visible.", "sourceParagraphs": [ 308, 309 ] }, { "type": "heading2", "text": "5. BLINDLY MULTIPLYING MARGINAL RATES IS PROHIBITED", "sourceParagraph": 311 }, { "type": "paragraph", "text": "The eligible population may sometimes be approximately calculated with a formula like the following:", "sourceParagraphs": [ 312 ] }, { "type": "formula", "text": "N × p_age × p_internet × p_product × p_language", "sourceParagraph": 313 }, { "type": "paragraph", "text": "This calculation can only be justified under the independence of conditions or the assumption of an appropriate joint distribution. In real life, features may not be independent. For example:", "sourceParagraphs": [ 314, 315, 316 ] }, { "type": "bullet", "text": "internet usage may vary by age,", "sourceParagraph": 317 }, { "type": "bullet", "text": "access to a paid plan may be related to income,", "sourceParagraph": 318 }, { "type": "bullet", "text": "certain language use may be related to the region within the country and education,", "sourceParagraph": 319 }, { "type": "paragraph", "text": "device access and connection quality may vary together. Multiplying marginal rates can predict a target population that is:", "sourceParagraphs": [ 320, 321 ] }, { "type": "bullet", "text": "larger than necessary,", "sourceParagraph": 322 }, { "type": "bullet", "text": "smaller than necessary", "sourceParagraph": 323 }, { "type": "paragraph", "text": "The order of preference should be:", "sourceParagraphs": [ 324, 325 ] }, { "type": "bullet", "text": "Appropriate joint microdata or cross-tabulation", "sourceParagraph": 326 }, { "type": "bullet", "text": "Reliable multidimensional population projection", "sourceParagraph": 327 }, { "type": "bullet", "text": "Calibrated model", "sourceParagraph": 328 }, { "type": "bullet", "text": "Marginal approach with explained rationale and sensitivity analysis", "sourceParagraph": 329 }, { "type": "paragraph", "text": "If simple multiplication is used:", "sourceParagraphs": [ 330 ] }, { "type": "bullet", "text": "assumption,", "sourceParagraph": 331 }, { "type": "bullet", "text": "data source,", "sourceParagraph": 332 }, { "type": "bullet", "text": "uncertainty,", "sourceParagraph": 333 }, { "type": "bullet", "text": "alternative scenario", "sourceParagraph": 334 }, { "type": "paragraph", "text": "should be disclosed.", "sourceParagraphs": [ 335 ] }, { "type": "heading2", "text": "6. AUDIT QUESTION CHANGES THE TARGET POPULATION", "sourceParagraph": 337 }, { "type": "paragraph", "text": "Not every prompt has the same measurement universe for the entire world population.", "sourceParagraphs": [ 338 ] }, { "type": "heading3", "text": "6.1. Global Identity Questions", "sourceParagraph": 340 }, { "type": "paragraph", "text": "Example: “What kind of company is Apple.com?” This question can be widely asked among the suitable global user population for the product. Because the company identity is not limited to users of a specific market.", "sourceParagraphs": [ 341, 342, 343, 344 ] }, { "type": "heading3", "text": "6.2. Local Service Questions", "sourceParagraph": 346 }, { "type": "paragraph", "text": "Example: “Does this company provide services in Turkey?” Main target population:", "sourceParagraphs": [ 347, 348, 349 ] }, { "type": "bullet", "text": "Eligible users in Turkey,", "sourceParagraph": 350 }, { "type": "bullet", "text": "External users investigating Turkey's service", "sourceParagraph": 351 }, { "type": "paragraph", "text": "It is possible. It is not necessary to give equal weight to the entire world population.", "sourceParagraphs": [ 352, 353 ] }, { "type": "heading3", "text": "6.3. Price and Purchase Questions", "sourceParagraph": 355 }, { "type": "paragraph", "text": "Price:", "sourceParagraphs": [ 356 ] }, { "type": "bullet", "text": "country,", "sourceParagraph": 357 }, { "type": "bullet", "text": "tax,", "sourceParagraph": 358 }, { "type": "bullet", "text": "currency,", "sourceParagraph": 359 }, { "type": "bullet", "text": "plan,", "sourceParagraph": 360 }, { "type": "bullet", "text": "customer type", "sourceParagraph": 361 }, { "type": "paragraph", "text": "may vary. The target population should be defined according to the relevant commercial market and type of user.", "sourceParagraphs": [ 362, 363 ] }, { "type": "heading3", "text": "6.4. Legal and Licensing Issues", "sourceParagraph": 365 }, { "type": "paragraph", "text": "Licensing and legal authority depends on the jurisdiction. The target population:", "sourceParagraphs": [ 366, 367 ] }, { "type": "bullet", "text": "can obtain the relevant service in that country,", "sourceParagraph": 368 }, { "type": "bullet", "text": "are subject to that legal regime.", "sourceParagraph": 369 }, { "type": "paragraph", "text": "The target population may be restricted accordingly.", "sourceParagraphs": [ 370 ] }, { "type": "heading3", "text": "6.5. Advisory Issues", "sourceParagraph": 372 }, { "type": "paragraph", "text": "The target population for the question \"Is this company suitable for me?\" is not all people. It can be defined according to:", "sourceParagraphs": [ 373, 374, 375 ] }, { "type": "bullet", "text": "User needs", "sourceParagraph": 376 }, { "type": "bullet", "text": "Country", "sourceParagraph": 377 }, { "type": "bullet", "text": "Budget", "sourceParagraph": 378 }, { "type": "bullet", "text": "Customer type", "sourceParagraph": 379 }, { "type": "bullet", "text": "Scope of service", "sourceParagraph": 380 }, { "type": "bullet", "text": "Eligibility and exclusion conditions", "sourceParagraph": 381 }, { "type": "paragraph", "text": "The advisory population may be narrower than the identity population.", "sourceParagraphs": [ 382 ] }, { "type": "heading3", "text": "6.6. High-Risk Questions", "sourceParagraph": 384 }, { "type": "paragraph", "text": "In the fields of health, law, finance, or security:", "sourceParagraphs": [ 385 ] }, { "type": "bullet", "text": "judgement,", "sourceParagraph": 386 }, { "type": "bullet", "text": "user profile,", "sourceParagraph": 387 }, { "type": "bullet", "text": "risk status,", "sourceParagraph": 388 }, { "type": "bullet", "text": "professional authority", "sourceParagraph": 389 }, { "type": "paragraph", "text": "more strongly limits the target population.", "sourceParagraphs": [ 390 ] }, { "type": "heading2", "text": "7. POPULATION FRAMEWORK MUST BE LINKED TO PROMPT FAMILY", "sourceParagraph": 392 }, { "type": "paragraph", "text": "In an audit, there may be multiple population frameworks instead of a single target population.", "sourceParagraphs": [ 393 ] }, { "type": "heading3", "text": "7.1. Core Identity Population", "sourceParagraph": 395 }, { "type": "paragraph", "text": "The global or broadly eligible user population measuring the entity's core identity.", "sourceParagraphs": [ 396 ] }, { "type": "heading3", "text": "7.2. Service-Relevant Population", "sourceParagraph": 398 }, { "type": "paragraph", "text": "It is the population for which the entity can actually provide service or can be considered for service by the user.", "sourceParagraphs": [ 399 ] }, { "type": "heading3", "text": "7.3. Market Transaction Population", "sourceParagraph": 401 }, { "type": "paragraph", "text": "The specific price, purchase, return, or commercial condition is the population to which it can be applied.", "sourceParagraphs": [ 402 ] }, { "type": "heading3", "text": "7.4. Jurisdictional Population", "sourceParagraph": 404 }, { "type": "paragraph", "text": "It is the user population to which a specific licence, legal, or regulatory claim can be applied.", "sourceParagraphs": [ 405 ] }, { "type": "heading3", "text": "7.5. Recommendation-Eligible Population", "sourceParagraph": 407 }, { "type": "paragraph", "text": "It is the user population that may be appropriate within the true scope of the entity.", "sourceParagraphs": [ 408 ] }, { "type": "heading3", "text": "7.6. High-Risk Affected Population", "sourceParagraph": 410 }, { "type": "paragraph", "text": "It is a population for which misrepresentation can cause serious harm. These frameworks should not be confused with each other. A global identity score does not mean \"recommendable for the whole world.\"", "sourceParagraphs": [ 411, 412, 413, 414, 415 ] }, { "type": "heading2", "text": "8. THE TYPE OF AI PRODUCT DETERMINES THE POPULATION", "sourceParagraph": 417 }, { "type": "paragraph", "text": "Not all AI products are publicly available chat products.", "sourceParagraphs": [ 418 ] }, { "type": "heading3", "text": "8.1. General Consumer AI Product", "sourceParagraph": 420 }, { "type": "paragraph", "text": "Target population:", "sourceParagraphs": [ 421 ] }, { "type": "bullet", "text": "There may be people who can access the product,", "sourceParagraph": 422 }, { "type": "bullet", "text": "meet the terms of use,", "sourceParagraph": 423 }, { "type": "bullet", "text": "and use it in the relevant language.", "sourceParagraph": 424 }, { "type": "paragraph", "text": "The GEO-1000 real user panel can be directly applied to this type of product.", "sourceParagraphs": [ 425, 426 ] }, { "type": "heading3", "text": "8.2. Enterprise AI Product", "sourceParagraph": 428 }, { "type": "paragraph", "text": "The product is only for:", "sourceParagraphs": [ 429 ] }, { "type": "bullet", "text": "employer contract,", "sourceParagraph": 430 }, { "type": "bullet", "text": "corporate licence,", "sourceParagraph": 431 }, { "type": "bullet", "text": "can be used with an in-house user account", "sourceParagraph": 432 }, { "type": "paragraph", "text": "The target population is not the general adult population:", "sourceParagraphs": [ 433, 434 ] }, { "type": "callout", "text": "It is the seat or user universe with corporate access to the product.", "sourceParagraph": 435 }, { "type": "paragraph", "text": "Corporate results cannot be compared with the general consumer product in population weighting.", "sourceParagraphs": [ 436 ] }, { "type": "heading3", "text": "8.3. Educational Product", "sourceParagraph": 438 }, { "type": "paragraph", "text": "Target population:", "sourceParagraphs": [ 439 ] }, { "type": "bullet", "text": "student,", "sourceParagraph": 440 }, { "type": "bullet", "text": "teacher,", "sourceParagraph": 441 }, { "type": "bullet", "text": "institution,", "sourceParagraph": 442 }, { "type": "bullet", "text": "age group,", "sourceParagraph": 443 }, { "type": "bullet", "text": "educational licence", "sourceParagraph": 444 }, { "type": "paragraph", "text": "may be limited in terms of. If children or minors are to be included, a separate ethical protocol is required.", "sourceParagraphs": [ 445, 446 ] }, { "type": "heading3", "text": "8.4. Developer or API Product", "sourceParagraph": 448 }, { "type": "paragraph", "text": "An API output may not be exposed directly through an end-user interface. The target unit is:", "sourceParagraphs": [ 449, 450 ] }, { "type": "bullet", "text": "developer,", "sourceParagraph": 451 }, { "type": "bullet", "text": "application,", "sourceParagraph": 452 }, { "type": "bullet", "text": "API call,", "sourceParagraph": 453 }, { "type": "bullet", "text": "controlled test session", "sourceParagraph": 454 }, { "type": "paragraph", "text": "may be. The GEO-1000 human population panel cannot be applied directly and unchanged to the API. Separately for the API:", "sourceParagraphs": [ 455, 456, 457 ] }, { "type": "heading4", "text": "Controlled System Audit", "sourceParagraph": 458 }, { "type": "paragraph", "text": "may be required.", "sourceParagraphs": [ 459 ] }, { "type": "heading3", "text": "8.5. Embedded AI Assistant", "sourceParagraph": 461 }, { "type": "paragraph", "text": "AI:", "sourceParagraphs": [ 462 ] }, { "type": "bullet", "text": "operating system,", "sourceParagraph": 463 }, { "type": "bullet", "text": "search interface,", "sourceParagraph": 464 }, { "type": "bullet", "text": "shopping platform,", "sourceParagraph": 465 }, { "type": "bullet", "text": "business software,", "sourceParagraph": 466 }, { "type": "bullet", "text": "device", "sourceParagraph": 467 }, { "type": "paragraph", "text": "It may be embedded. The target population is the actual user population of the embedded product. It is not the total user population of the base model provider.", "sourceParagraphs": [ 468, 469, 470 ] }, { "type": "heading3", "text": "8.6. Expert or Sectoral AI", "sourceParagraph": 472 }, { "type": "paragraph", "text": "The product is only for:", "sourceParagraphs": [ 473 ] }, { "type": "bullet", "text": "lawyers,", "sourceParagraph": 474 }, { "type": "bullet", "text": "doctors,", "sourceParagraph": 475 }, { "type": "bullet", "text": "finance professionals,", "sourceParagraph": 476 }, { "type": "bullet", "text": "engineers", "sourceParagraph": 477 }, { "type": "paragraph", "text": "may be designed for. The target population should be limited by profession and access conditions.", "sourceParagraphs": [ 478, 479 ] }, { "type": "heading2", "text": "9. PRODUCT ACCESS STATUSES", "sourceParagraph": 481 }, { "type": "paragraph", "text": "Access status should be recorded for each country, product, interface, and date.", "sourceParagraphs": [ 482 ] }, { "type": "heading3", "text": "PA-0 — UNKNOWN", "sourceParagraph": 484 }, { "type": "paragraph", "text": "Access status could not be determined reliably.", "sourceParagraphs": [ 485 ] }, { "type": "heading3", "text": "PA-1 — UNAVAILABLE", "sourceParagraph": 487 }, { "type": "paragraph", "text": "The product is not offered in the relevant country or user group.", "sourceParagraphs": [ 488 ] }, { "type": "heading3", "text": "PA-2 — RESTRICTED", "sourceParagraph": 490 }, { "type": "paragraph", "text": "The product is limited to certain:", "sourceParagraphs": [ 491 ] }, { "type": "bullet", "text": "account,", "sourceParagraph": 492 }, { "type": "bullet", "text": "invitation,", "sourceParagraph": 493 }, { "type": "bullet", "text": "institution,", "sourceParagraph": 494 }, { "type": "bullet", "text": "does not imply any suitability in terms of age,", "sourceParagraph": 495 }, { "type": "bullet", "text": "plan,", "sourceParagraph": 496 }, { "type": "bullet", "text": "device", "sourceParagraph": 497 }, { "type": "paragraph", "text": "terms.", "sourceParagraphs": [ 498 ] }, { "type": "heading3", "text": "PA-3 — ACCESSIBLE BUT NOT OFFICIALLY SUPPORTED", "sourceParagraph": 500 }, { "type": "paragraph", "text": "Actual access to the product can be observed. However, official support or usage status is unclear or limited. These users should not be automatically added to the main official population panel. A separate experimental panel can be created.", "sourceParagraphs": [ 501, 502, 503, 504 ] }, { "type": "heading3", "text": "PA-4 — OFFICIALLY AVAILABLE", "sourceParagraph": 506 }, { "type": "paragraph", "text": "The product is officially accessible and the main usage conditions are defined.", "sourceParagraphs": [ 507 ] }, { "type": "heading3", "text": "PA-5 — OFFICIALLY AVAILABLE WITH FULL AUDIT SURFACE", "sourceParagraph": 509 }, { "type": "paragraph", "text": "The product is not only inaccessible; it is controlled:", "sourceParagraphs": [ 510 ] }, { "type": "bullet", "text": "language,", "sourceParagraph": 511 }, { "type": "bullet", "text": "plan,", "sourceParagraph": 512 }, { "type": "bullet", "text": "interface,", "sourceParagraph": 513 }, { "type": "bullet", "text": "under session conditions", "sourceParagraph": 514 }, { "type": "paragraph", "text": "suitable for generating the observation required by the protocol. The availability of the product in a country does not mean that all plans and languages are in compliance with the control.", "sourceParagraphs": [ 515, 516 ] }, { "type": "heading2", "text": "10. BYPASSING RESTRICTION IS NOT A REAL POPULATION MEASUREMENT", "sourceParagraph": 518 }, { "type": "paragraph", "text": "Participants cannot be asked to perform the following behaviours:", "sourceParagraphs": [ 519 ] }, { "type": "bullet", "text": "Declare the wrong country", "sourceParagraph": 520 }, { "type": "bullet", "text": "Create a fake account", "sourceParagraph": 521 }, { "type": "bullet", "text": "Change age information", "sourceParagraph": 522 }, { "type": "bullet", "text": "Bypass official access restriction", "sourceParagraph": 523 }, { "type": "bullet", "text": "Use someone else's account", "sourceParagraph": 524 }, { "type": "bullet", "text": "Bypass the product's security or verification", "sourceParagraph": 525 }, { "type": "bullet", "text": "Violating the terms of use", "sourceParagraph": 526 }, { "type": "paragraph", "text": "These behaviours can also be examined as a technical experiment. However:", "sourceParagraphs": [ 527, 528 ] }, { "type": "heading4", "text": "The experience of natural and legitimate user population", "sourceParagraph": 529 }, { "type": "paragraph", "text": "cannot be reported as such. If there is no access in a country, the correct status should be: UNAVAILABLE / or / OUTSIDE_NATIVE_REACH. A single observation exceeding the restriction does not generate a country population score.", "sourceParagraphs": [ 530, 531, 532, 533, 534 ] }, { "type": "heading2", "text": "11. DISTINCTION BETWEEN POTENTIAL USER AND ACTIVE USER", "sourceParagraph": 536 }, { "type": "paragraph", "text": "The people who can use the product are not the same community as the people who actually use the product.", "sourceParagraphs": [ 537 ] }, { "type": "heading3", "text": "11.1. Potential Access Population", "sourceParagraph": 539 }, { "type": "paragraph", "text": "These are the people who can technically and legally use the product. This population answers the question: 'What would a suitable person see if they used the product?' The population is valuable for justice and product comparison.", "sourceParagraphs": [ 540, 541, 542, 543, 544 ] }, { "type": "heading3", "text": "11.2. Active User Population", "sourceParagraph": 546 }, { "type": "paragraph", "text": "These are the people who actually used the product during a specific period. This population approaches the question: 'What do people see within the actual usage volume?' It requires reliable active usage data.", "sourceParagraphs": [ 547, 548, 549, 550, 551 ] }, { "type": "heading3", "text": "11.3. Query-Exposed User Population", "sourceParagraph": 553 }, { "type": "paragraph", "text": "These are people who use the product and have a potential to ask questions about the monitored brand or topic. This population often cannot be known directly. Search or query behaviour data may be required.", "sourceParagraphs": [ 554, 555, 556 ] }, { "type": "heading3", "text": "11.4. Population Actually Seeing the Response", "sourceParagraph": 558 }, { "type": "paragraph", "text": "They are users who are actually exposed to a specific representation output. This may not be fully measurable without real product provider data outside the user panel.", "sourceParagraphs": [ 559, 560 ] }, { "type": "heading2", "text": "12. EXPOSURE WEIGHT IS NOT THE SAME AS ELIGIBILITY WEIGHT", "sourceParagraph": 562 }, { "type": "paragraph", "text": "Two separate global results can be generated.", "sourceParagraphs": [ 563 ] }, { "type": "heading3", "text": "12.1. Eligibility Population Weighted Result", "sourceParagraph": 565 }, { "type": "paragraph", "text": "Each person is weighted according to their share within the defined eligible user population. This result is:", "sourceParagraphs": [ 566, 567 ] }, { "type": "bullet", "text": "globally fair,", "sourceParagraph": 568 }, { "type": "bullet", "text": "potentially accessible,", "sourceParagraph": 569 }, { "type": "bullet", "text": "representative of the population", "sourceParagraph": 570 }, { "type": "paragraph", "text": "in terms of", "sourceParagraphs": [ 571 ] }, { "type": "heading3", "text": "12.2. Real Exposure Weighted Result", "sourceParagraph": 573 }, { "type": "paragraph", "text": "AI products or user groups are weighted according to actual usage volumes. This result:", "sourceParagraphs": [ 574, 575 ] }, { "type": "bullet", "text": "is valuable in terms of", "sourceParagraph": 576 }, { "type": "bullet", "text": "current commercial impact,", "sourceParagraph": 577 }, { "type": "bullet", "text": "actual user exposure,", "sourceParagraph": 578 }, { "type": "paragraph", "text": "operational risk.\n\nHowever, reliable usage data may not be available. Estimates may rely on provider statements, market data, or incomplete sampling. Therefore, the type of weighting must be explicitly stated.", "sourceParagraphs": [ 579, 580, 581, 582 ] }, { "type": "heading3", "text": "12.3. Replacing One with Another is Prohibited", "sourceParagraph": 584 }, { "type": "paragraph", "text": "Eligible population score: does not mean \"Most of the actual users saw this response.\" Exposure score: does not mean \"Fair representation of the world population is ensured.\"", "sourceParagraphs": [ 585, 586, 587, 588, 589, 590 ] }, { "type": "heading2", "text": "13. ACCESS SCOPE AND REPRESENTATION ACCURACY SHOULD BE SEPARATED", "sourceParagraph": 592 }, { "type": "paragraph", "text": "An AI product may only be available in limited countries. In these countries, it can generate very accurate representations. Another product may have wider access but can produce lower accuracy. If we only look at representation accuracy, the product with limited access may appear superior. If we only look at access coverage, incorrect representation may be overlooked. Therefore, two separate results are needed.", "sourceParagraphs": [ 593, 594, 595, 596, 597, 598 ] }, { "type": "heading3", "text": "13.1. Access Coverage", "sourceParagraph": 600 }, { "type": "paragraph", "text": "Indicates how much of the specified global reference population can appropriately use the relevant AI product. Candidate representation:", "sourceParagraphs": [ 601, 602 ] }, { "type": "formula", "text": "AC_a = N(U_{a,t}) / N(G_t)", "sourceParagraph": 603 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 604 ] }, { "type": "bullet", "text": "Ua,t: eligible user population of the product", "sourceParagraph": 605 }, { "type": "bullet", "text": "Gt: defined global reference population", "sourceParagraph": 606 }, { "type": "heading3", "text": "13.2. Conditional Representation Accuracy", "sourceParagraph": 608 }, { "type": "paragraph", "text": "It is the probability of seeing an accurate representation among users who can access the product. Candidate notation:", "sourceParagraphs": [ 609, 610 ] }, { "type": "formula", "text": "RA_a = Pr(Y=1 | u ∈ U_{a,t})", "sourceParagraph": 611 }, { "type": "heading3", "text": "13.3. Effective Accurate Representation Access", "sourceParagraph": 613 }, { "type": "heading3", "text": "CANDIDATE CONCEPT", "sourceParagraph": 614 }, { "type": "paragraph", "text": "An operational metric can be developed that combines access coverage with conditional accuracy:", "sourceParagraphs": [ 615 ] }, { "type": "formula", "text": "ECR_a = AC_a × RA_a", "sourceParagraph": 616 }, { "type": "paragraph", "text": "This metric roughly approaches the question: “What proportion of the defined global reference population can access this product and meet the condition of seeing an accurate representation?” This is not yet the final NOMOS formula. Because:", "sourceParagraphs": [ 617, 618, 619, 620 ] }, { "type": "bullet", "text": "access is different from actual usage,", "sourceParagraph": 621 }, { "type": "bullet", "text": "weight and dependency structures are complex,", "sourceParagraph": 622 }, { "type": "bullet", "text": "product usage is voluntary,", "sourceParagraph": 623 }, { "type": "paragraph", "text": "The correct representation depends on the prompt. However, it maintains the basic distinction:", "sourceParagraphs": [ 624, 625 ] }, { "type": "callout", "text": "Not being able to access is not the same as getting a wrong answer.", "sourceParagraph": 626 }, { "type": "heading2", "text": "14. NATIVE REACH AND COMMON SUPPORT", "sourceParagraph": 628 }, { "type": "paragraph", "text": "When comparing AI products, two separate population universes are needed.", "sourceParagraphs": [ 629 ] }, { "type": "heading3", "text": "14.1. Native Reach Population", "sourceParagraph": 631 }, { "type": "paragraph", "text": "Each AI product has its eligible population within its official and actual access domain. This result answers the following question:", "sourceParagraphs": [ 632, 633 ] }, { "type": "callout", "text": "To what extent does the product accurately represent the population it can actually reach?", "sourceParagraph": 634 }, { "type": "paragraph", "text": "Target populations may differ between products.", "sourceParagraphs": [ 635 ] }, { "type": "heading3", "text": "14.2. Common Support Population", "sourceParagraph": 637 }, { "type": "paragraph", "text": "In a common form across all compared AI products:", "sourceParagraphs": [ 638 ] }, { "type": "bullet", "text": "officially accessible,", "sourceParagraph": 639 }, { "type": "bullet", "text": "can be used under the same language or interface condition", "sourceParagraph": 640 }, { "type": "paragraph", "text": "is the population. This result answers the following question:", "sourceParagraphs": [ 641, 642 ] }, { "type": "callout", "text": "Which product produces more accurate representation in the same user universe?", "sourceParagraph": 643 }, { "type": "heading3", "text": "14.3. Two Results Should Be Published Together", "sourceParagraph": 645 }, { "type": "paragraph", "text": "Native Reach:", "sourceParagraphs": [ 646 ] }, { "type": "bullet", "text": "shows actual product access,", "sourceParagraph": 647 }, { "type": "bullet", "text": "natural geographic distribution", "sourceParagraph": 648 }, { "type": "paragraph", "text": "Common Support: provides a fairer product comparison. A product can be high in Native Reach score, but lower in Common Support score. Two results answer different questions.", "sourceParagraphs": [ 649, 650, 651, 652, 653, 654 ] }, { "type": "heading2", "text": "15. MARKET-RELATED POPULATION", "sourceParagraph": 656 }, { "type": "paragraph", "text": "The global identity score of an entity may differ from its commercial target market score. Example: If a local law firm provides services only in Turkey: the global identity question can be measured in the eligible user population worldwide, but recommendations and service suitability should only be measured for users in Turkey or within the actual scope of services. The institution may select the result in its favour:", "sourceParagraphs": [ 657, 658, 659, 660, 661, 662 ] }, { "type": "bullet", "text": "only the high-scoring market,", "sourceParagraph": 663 }, { "type": "bullet", "text": "only his/her own language,", "sourceParagraph": 664 }, { "type": "bullet", "text": "only its existing customers", "sourceParagraph": 665 }, { "type": "paragraph", "text": "the target population cannot be announced. Target market:", "sourceParagraphs": [ 666, 667 ] }, { "type": "bullet", "text": "Before the audit begins,", "sourceParagraph": 668 }, { "type": "bullet", "text": "to the service records,", "sourceParagraph": 669 }, { "type": "bullet", "text": "within the scope of the contract and operations", "sourceParagraph": 670 }, { "type": "paragraph", "text": "It should be locked according to.", "sourceParagraphs": [ 671 ] }, { "type": "heading2", "text": "16. PERSON, ACCOUNT AND SESSION SEPARATION", "sourceParagraph": 673 }, { "type": "paragraph", "text": "Population weight belongs to people. Not to AI accounts.", "sourceParagraphs": [ 674, 675 ] }, { "type": "heading3", "text": "16.1. One Person, Multiple Accounts", "sourceParagraph": 677 }, { "type": "paragraph", "text": "The same person:", "sourceParagraphs": [ 678 ] }, { "type": "bullet", "text": "free,", "sourceParagraph": 679 }, { "type": "bullet", "text": "paid,", "sourceParagraph": 680 }, { "type": "bullet", "text": "corporate", "sourceParagraph": 681 }, { "type": "paragraph", "text": "can have multiple accounts. These accounts can produce separate observations in product surface comparisons. They cannot be counted as separate individuals in the global human population.", "sourceParagraphs": [ 682, 683, 684 ] }, { "type": "heading3", "text": "16.2. One Account, Multiple People", "sourceParagraph": 686 }, { "type": "paragraph", "text": "A family, team, or corporate account may be used by multiple people. It is not equal to the number of account users.", "sourceParagraphs": [ 687, 688 ] }, { "type": "heading3", "text": "16.3. One Person, Multiple Sessions", "sourceParagraph": 690 }, { "type": "paragraph", "text": "A person can produce separate observations in different sessions. These observations can measure time or session stability. They do not increase independent population size.", "sourceParagraphs": [ 691, 692, 693 ] }, { "type": "heading3", "text": "16.4. One Person, Multiple AI Products", "sourceParagraph": 695 }, { "type": "paragraph", "text": "It is valuable for paired product comparison. However, observations are dependent due to shared characteristics at the user level.", "sourceParagraphs": [ 696, 697 ] }, { "type": "heading3", "text": "16.5. Main Weight Unit", "sourceParagraph": 699 }, { "type": "paragraph", "text": "In the population-based GEO score, the main weight unit should be:", "sourceParagraphs": [ 700 ] }, { "type": "heading4", "text": "human", "sourceParagraph": 701 }, { "type": "paragraph", "text": "Calculations, session, and product observations should be associated under the human unit.", "sourceParagraphs": [ 702, 703 ] }, { "type": "heading2", "text": "17. LANGUAGE POPULATION", "sourceParagraph": 705 }, { "type": "paragraph", "text": "Using only the country’s official language is not sufficient for language fairness.", "sourceParagraphs": [ 706 ] }, { "type": "heading3", "text": "17.1. Prompt Language", "sourceParagraph": 708 }, { "type": "paragraph", "text": "The language in which the participant asks the audit question.", "sourceParagraphs": [ 709 ] }, { "type": "heading3", "text": "17.2. Interface Language", "sourceParagraph": 711 }, { "type": "paragraph", "text": "The language used in the menu and user interface of the AI product. It may not be the same as the prompt language.", "sourceParagraphs": [ 712, 713 ] }, { "type": "heading3", "text": "17.3. Response Language", "sourceParagraph": 715 }, { "type": "paragraph", "text": "The language in which the AI actually responds. The system may respond in a different language than the prompt language. This is also a measurement result.", "sourceParagraphs": [ 716, 717, 718 ] }, { "type": "heading3", "text": "17.4. User Proficiency Language", "sourceParagraph": 720 }, { "type": "paragraph", "text": "The language in which the participant can meaningfully use the prompt and response. It does not have to be the native language.", "sourceParagraphs": [ 721, 722 ] }, { "type": "heading3", "text": "17.5. Locale", "sourceParagraph": 724 }, { "type": "paragraph", "text": "The regional context along with the language. The same language can be different:", "sourceParagraphs": [ 725, 726 ] }, { "type": "bullet", "text": "country,", "sourceParagraph": 727 }, { "type": "bullet", "text": "price,", "sourceParagraph": 728 }, { "type": "bullet", "text": "law,", "sourceParagraph": 729 }, { "type": "bullet", "text": "terminology", "sourceParagraph": 730 }, { "type": "paragraph", "text": "may carry the condition.", "sourceParagraphs": [ 731 ] }, { "type": "heading2", "text": "18. DOUBLE COUNTING OF MULTILINGUAL USERS", "sourceParagraph": 733 }, { "type": "paragraph", "text": "A person:", "sourceParagraphs": [ 734 ] }, { "type": "bullet", "text": "Turkish,", "sourceParagraph": 735 }, { "type": "bullet", "text": "may be able to use English.", "sourceParagraph": 736 }, { "type": "bullet", "text": "German", "sourceParagraph": 737 }, { "type": "paragraph", "text": "This person may take three separate language tests. However, counting them as three people in the global population weight distorts the population estimate. Three methods can be used.", "sourceParagraphs": [ 738, 739, 740, 741 ] }, { "type": "heading3", "text": "18.1. Single Mother Tongue Assignment", "sourceParagraph": 743 }, { "type": "paragraph", "text": "The person is assigned to a single language cell in the main analysis. Other languages are used as additional or paired tests.", "sourceParagraphs": [ 744, 745 ] }, { "type": "heading3", "text": "18.2. Split Person Weight", "sourceParagraph": 747 }, { "type": "paragraph", "text": "The person's population weight can be divided among the appropriate languages in a predefined manner. The total weight does not exceed one person.", "sourceParagraphs": [ 748, 749 ] }, { "type": "heading3", "text": "18.3. Separate Multilingual Matched Panel", "sourceParagraph": 751 }, { "type": "paragraph", "text": "Multilingual users are used for cross-language comparisons of the same person. This panel is analysed separately from the main population panel. Regardless of the method used:", "sourceParagraphs": [ 752, 753, 754 ] }, { "type": "callout", "text": "The total population weight of the same person cannot be replicated without explanation.", "sourceParagraph": 755 }, { "type": "heading2", "text": "19. HOW SHOULD LANGUAGE PROFICIENCY BE VERIFIED?", "sourceParagraph": 757 }, { "type": "paragraph", "text": "A participant's statement: \"I know this language.\" can be an initial record. However, in high security auditing, one of the following methods can be used:", "sourceParagraphs": [ 758, 759, 760, 761 ] }, { "type": "bullet", "text": "Native language or daily usage declaration", "sourceParagraph": 762 }, { "type": "bullet", "text": "Short reading and comprehension test", "sourceParagraph": 763 }, { "type": "bullet", "text": "Language record of the panel provider", "sourceParagraph": 764 }, { "type": "bullet", "text": "Local adjudicator verification", "sourceParagraph": 765 }, { "type": "bullet", "text": "Education or professional use record", "sourceParagraph": 766 }, { "type": "bullet", "text": "Pilot mission success", "sourceParagraph": 767 }, { "type": "paragraph", "text": "Extremely difficult language test panels can be accessible to trained and high-income users. The purpose is not to conduct an academic language exam. It is to verify the competence to understand the prompt and response as a real user.", "sourceParagraphs": [ 768, 769, 770 ] }, { "type": "heading2", "text": "20. WHAT HAPPENS IF A LANGUAGE IS NOT SUPPORTED?", "sourceParagraph": 772 }, { "type": "paragraph", "text": "An AI product may:", "sourceParagraphs": [ 773 ] }, { "type": "bullet", "text": "not officially support a particular language,", "sourceParagraph": 774 }, { "type": "bullet", "text": "partially support it,", "sourceParagraph": 775 }, { "type": "bullet", "text": "not claim to support it but can generate responses", "sourceParagraph": 776 }, { "type": "paragraph", "text": "It may be. Language statuses should be recorded separately:", "sourceParagraphs": [ 777, 778 ] }, { "type": "heading3", "text": "OFFICIALLY_SUPPORTED", "sourceParagraph": 779 }, { "type": "heading3", "text": "LIMITED_SUPPORT", "sourceParagraph": 780 }, { "type": "heading3", "text": "OBSERVED_BUT_UNSUPPORTED", "sourceParagraph": 781 }, { "type": "heading3", "text": "NOT_SUPPORTED", "sourceParagraph": 782 }, { "type": "heading3", "text": "UNKNOWN", "sourceParagraph": 783 }, { "type": "paragraph", "text": "Response received in a language that is not officially supported:", "sourceParagraphs": [ 784 ] }, { "type": "bullet", "text": "discovery,", "sourceParagraph": 785 }, { "type": "bullet", "text": "experimental access,", "sourceParagraph": 786 }, { "type": "bullet", "text": "language resilience", "sourceParagraph": 787 }, { "type": "paragraph", "text": "may be measured. It should not be automatically added to the main supported user population score. Language access limit should also be reported.", "sourceParagraphs": [ 788, 789, 790 ] }, { "type": "heading2", "text": "21. COUNTRY POPULATION WEIGHTING AND Country Observer COVERAGE", "sourceParagraph": 792 }, { "type": "paragraph", "text": "Population-weighted main panel and country visibility panel serve separate purposes.", "sourceParagraphs": [ 793 ] }, { "type": "heading3", "text": "21.1. Population Panel", "sourceParagraph": 795 }, { "type": "paragraph", "text": "Participants are distributed according to their share in the product-appropriate target population. Larger eligible populations receive higher weight. Produces the main global population score.", "sourceParagraphs": [ 796, 797, 798 ] }, { "type": "heading3", "text": "21.2. Country Observer Panel", "sourceParagraph": 800 }, { "type": "paragraph", "text": "The Country Observer Panel seeks at least one observation from smaller countries that receive little or no representation in the main population sample. A single Country Observer observation:", "sourceParagraphs": [ 801, 802 ] }, { "type": "bullet", "text": "indicates access is available,", "sourceParagraph": 803 }, { "type": "bullet", "text": "that a specific singular result has occurred,", "sourceParagraph": 804 }, { "type": "bullet", "text": "shows the need for new research", "sourceParagraph": 805 }, { "type": "paragraph", "text": "may indicate. Does not produce a country score on its own.", "sourceParagraphs": [ 806, 807 ] }, { "type": "heading3", "text": "21.3. Sufficiency for Country Score", "sourceParagraph": 809 }, { "type": "paragraph", "text": "For country-based rates to be published:", "sourceParagraphs": [ 810 ] }, { "type": "bullet", "text": "minimum sample,", "sourceParagraph": 811 }, { "type": "bullet", "text": "valid observation rate,", "sourceParagraph": 812 }, { "type": "bullet", "text": "uncertainty,", "sourceParagraph": 813 }, { "type": "bullet", "text": "panel representativeness", "sourceParagraph": 814 }, { "type": "paragraph", "text": "conditions must be met. With one or several participants: 'The country's NOMOS score' cannot be announced.", "sourceParagraphs": [ 815, 816, 817, 818 ] }, { "type": "heading2", "text": "22. Language Fairness Panel", "sourceParagraph": 820 }, { "type": "paragraph", "text": "Population weight may make low-population or low-resource languages invisible in the main score. Therefore, a separate Language Fairness Panel can be established. Purpose:", "sourceParagraphs": [ 821, 822, 823 ] }, { "type": "bullet", "text": "to test underrepresented languages,", "sourceParagraph": 824 }, { "type": "bullet", "text": "to detect severe language-based degradation,", "sourceParagraph": 825 }, { "type": "bullet", "text": "to identify prompt translation issues,", "sourceParagraph": 826 }, { "type": "bullet", "text": "to make low-resource language behaviour visible", "sourceParagraph": 827 }, { "type": "paragraph", "text": "must be. The Language Fairness Panel can provide additional samples to the main population panel. However, over-sampled language observations should be returned to actual population weights in the global score. Otherwise, the fairness panel alters the global population estimate.", "sourceParagraphs": [ 828, 829, 830, 831 ] }, { "type": "heading2", "text": "23. ACCESSIBILITY PANEL", "sourceParagraph": 833 }, { "type": "paragraph", "text": "Users with visual, hearing, motor, cognitive, or other accessibility needs are a real part of the target population. If the NOMOS Capture or research task is not suitable for these individuals' participation: users cannot be removed from the target population. Correct outcome:", "sourceParagraphs": [ 834, 835, 836, 837 ] }, { "type": "callout", "text": "The sampling frame is incomplete in terms of accessibility.", "sourceParagraph": 838 }, { "type": "paragraph", "text": "A separate accessibility panel can evaluate the following:", "sourceParagraphs": [ 839 ] }, { "type": "bullet", "text": "Screen reader compatibility", "sourceParagraph": 840 }, { "type": "bullet", "text": "Using the keyboard", "sourceParagraph": 841 }, { "type": "bullet", "text": "Voice input", "sourceParagraph": 842 }, { "type": "bullet", "text": "Low vision settings", "sourceParagraph": 843 }, { "type": "bullet", "text": "Cognitive load", "sourceParagraph": 844 }, { "type": "bullet", "text": "Mobile accessibility", "sourceParagraph": 845 }, { "type": "bullet", "text": "Usability of the Capture tool", "sourceParagraph": 846 }, { "type": "paragraph", "text": "The accessibility panel should see users as part of the population coverage, not as a \"special exception.\"", "sourceParagraphs": [ 847 ] }, { "type": "heading2", "text": "24. TARGET POPULATION AND SAMPLING FRAME", "sourceParagraph": 849 }, { "type": "paragraph", "text": "These two concepts should not be confused with each other.", "sourceParagraphs": [ 850 ] }, { "type": "heading3", "text": "24.1. Target Population", "sourceParagraph": 852 }, { "type": "paragraph", "text": "The entire group of people about which one wants to make a conclusion.", "sourceParagraphs": [ 853 ] }, { "type": "heading3", "text": "24.2. Sampling Frame", "sourceParagraph": 855 }, { "type": "paragraph", "text": "The population from which participants can actually be selected:", "sourceParagraphs": [ 856 ] }, { "type": "bullet", "text": "panel,", "sourceParagraph": 857 }, { "type": "bullet", "text": "list,", "sourceParagraph": 858 }, { "type": "bullet", "text": "database,", "sourceParagraph": 859 }, { "type": "bullet", "text": "joint organisation,", "sourceParagraph": 860 }, { "type": "bullet", "text": "user pool", "sourceParagraph": 861 }, { "type": "paragraph", "text": "is the cluster. The sampling frame may not fully cover the target population.", "sourceParagraphs": [ 862, 863 ] }, { "type": "heading3", "text": "24.3. Contactable Population", "sourceParagraph": 865 }, { "type": "paragraph", "text": "These are the people to whom the research invitation can actually be delivered.", "sourceParagraphs": [ 866 ] }, { "type": "heading3", "text": "24.4. Population Accepting to Participate", "sourceParagraph": 868 }, { "type": "paragraph", "text": "Those who voluntarily participate in the research.", "sourceParagraphs": [ 869 ] }, { "type": "heading3", "text": "24.5. Population Completing the Task", "sourceParagraph": 871 }, { "type": "paragraph", "text": "Those who complete the audit task.", "sourceParagraphs": [ 872 ] }, { "type": "heading3", "text": "24.6. Valid Analytical Population", "sourceParagraph": 874 }, { "type": "paragraph", "text": "They are the ones who produce observations suitable for the protocol and usable in the analysis. The numbers of these layers should be reported separately.", "sourceParagraphs": [ 875, 876 ] }, { "type": "heading2", "text": "25. COVERAGE ERROR", "sourceParagraph": 878 }, { "type": "paragraph", "text": "The sampling frame may exclude some parts of the target population. This is called coverage error. Examples:", "sourceParagraphs": [ 879, 880, 881 ] }, { "type": "bullet", "text": "Only professional online panels", "sourceParagraph": 882 }, { "type": "bullet", "text": "Only desktop users", "sourceParagraph": 883 }, { "type": "bullet", "text": "Only people with credit cards", "sourceParagraph": 884 }, { "type": "bullet", "text": "Only those using an English interface", "sourceParagraph": 885 }, { "type": "bullet", "text": "Only residents of large cities", "sourceParagraph": 886 }, { "type": "bullet", "text": "Only users highly interested in technology", "sourceParagraph": 887 }, { "type": "bullet", "text": "Exclusion of people with accessibility needs", "sourceParagraph": 888 }, { "type": "paragraph", "text": "Coverage error:", "sourceParagraphs": [ 889 ] }, { "type": "bullet", "text": "should be explained,", "sourceParagraph": 890 }, { "type": "bullet", "text": "should be measured if possible,", "sourceParagraph": 891 }, { "type": "paragraph", "text": "should be reduced with weighting or an additional panel. Weighting cannot magically create a group that does not exist in the sampling frame.", "sourceParagraphs": [ 892, 893 ] }, { "type": "heading2", "text": "26. OVER-COVERAGE", "sourceParagraph": 895 }, { "type": "paragraph", "text": "People who do not belong to the target population may be found in the sampling frame. Examples:", "sourceParagraphs": [ 896, 897 ] }, { "type": "bullet", "text": "Participant in a country where the product is not accessible", "sourceParagraph": 898 }, { "type": "bullet", "text": "Person who does not meet the required age condition", "sourceParagraph": 899 }, { "type": "bullet", "text": "Incorrect plan or user surface", "sourceParagraph": 900 }, { "type": "bullet", "text": "Participant who cannot use the audit language", "sourceParagraph": 901 }, { "type": "bullet", "text": "Person found again with multiple registrations", "sourceParagraph": 902 }, { "type": "bullet", "text": "User working on behalf of another country", "sourceParagraph": 903 }, { "type": "paragraph", "text": "Excess coverage records should not be included in the main score. However, the reason for exclusion and the rate should be reported.", "sourceParagraphs": [ 904, 905 ] }, { "type": "heading2", "text": "27. NON-RESPONSE AND SELF-SELECTION", "sourceParagraph": 907 }, { "type": "paragraph", "text": "People who agree to participate in the research may systematically differ from the target population. Volunteers:", "sourceParagraphs": [ 908, 909 ] }, { "type": "bullet", "text": "May use AI products more frequently,", "sourceParagraph": 910 }, { "type": "bullet", "text": "May be more interested in technology,", "sourceParagraph": 911 }, { "type": "bullet", "text": "May have a higher level of education,", "sourceParagraph": 912 }, { "type": "bullet", "text": "May already be familiar with the brand,", "sourceParagraph": 913 }, { "type": "paragraph", "text": "May behave differently due to reward or compensation. Therefore: the expression \"1,000 random people in the world\" should only be used in truly probabilistic and documented selection designs. If a volunteer or online panel is used, the correct expression is:", "sourceParagraphs": [ 914, 915, 916, 917, 918 ] }, { "type": "heading4", "text": "Layered and weighted voluntary user panel", "sourceParagraph": 919 }, { "type": "paragraph", "text": "may be possible. The panel type cannot be stored.", "sourceParagraphs": [ 920, 921 ] }, { "type": "heading2", "text": "28. SUITABLE POPULATION CHANGES OVER TIME", "sourceParagraph": 923 }, { "type": "paragraph", "text": "The user universe of an AI product is not fixed. The following can change:", "sourceParagraphs": [ 924, 925 ] }, { "type": "bullet", "text": "Supported countries", "sourceParagraph": 926 }, { "type": "bullet", "text": "Minimum age", "sourceParagraph": 927 }, { "type": "bullet", "text": "Free access", "sourceParagraph": 928 }, { "type": "bullet", "text": "Paid plan structure", "sourceParagraph": 929 }, { "type": "bullet", "text": "Mobile or web availability", "sourceParagraph": 930 }, { "type": "bullet", "text": "Language support", "sourceParagraph": 931 }, { "type": "bullet", "text": "Account terms", "sourceParagraph": 932 }, { "type": "bullet", "text": "Enterprise product coverage", "sourceParagraph": 933 }, { "type": "bullet", "text": "Legal and regulatory status", "sourceParagraph": 934 }, { "type": "paragraph", "text": "For this reason, each population record:", "sourceParagraphs": [ 935 ] }, { "type": "bullet", "text": "observation date,", "sourceParagraph": 936 }, { "type": "bullet", "text": "period of validity,", "sourceParagraph": 937 }, { "type": "bullet", "text": "source version,", "sourceParagraph": 938 }, { "type": "bullet", "text": "last verification date", "sourceParagraph": 939 }, { "type": "paragraph", "text": "must carry.", "sourceParagraphs": [ 940 ] }, { "type": "heading3", "text": "28.1. Change During Measurement Wave", "sourceParagraph": 942 }, { "type": "paragraph", "text": "If product access materially changes during the wave:", "sourceParagraphs": [ 943 ] }, { "type": "bullet", "text": "the wave can be stopped,", "sourceParagraph": 944 }, { "type": "bullet", "text": "two separate periods can be created before and after the change,", "sourceParagraph": 945 }, { "type": "paragraph", "text": "affected countries can be analysed separately. Users before and after the change cannot be combined as a single population.", "sourceParagraphs": [ 946, 947 ] }, { "type": "heading2", "text": "29. POPULATION SOURCE FILE", "sourceParagraph": 949 }, { "type": "paragraph", "text": "Each audit should store the target population data it uses in a versioned file. As applicable, the following fields should be included:", "sourceParagraphs": [ 950, 951 ] }, { "type": "bullet", "text": "Country", "sourceParagraph": 952 }, { "type": "bullet", "text": "Reference population", "sourceParagraph": 953 }, { "type": "bullet", "text": "Population history", "sourceParagraph": 954 }, { "type": "bullet", "text": "Age eligibility", "sourceParagraph": 955 }, { "type": "bullet", "text": "Internet and device access", "sourceParagraph": 956 }, { "type": "bullet", "text": "Product access status", "sourceParagraph": 957 }, { "type": "bullet", "text": "Interface access", "sourceParagraph": 958 }, { "type": "bullet", "text": "Plan and account requirement", "sourceParagraph": 959 }, { "type": "bullet", "text": "Language groups", "sourceParagraph": 960 }, { "type": "bullet", "text": "Research eligibility", "sourceParagraph": 961 }, { "type": "bullet", "text": "Common distribution method", "sourceParagraph": 962 }, { "type": "bullet", "text": "Uncertainty", "sourceParagraph": 963 }, { "type": "bullet", "text": "Sources", "sourceParagraph": 964 }, { "type": "bullet", "text": "Last verification date", "sourceParagraph": 965 }, { "type": "bullet", "text": "Responsible person", "sourceParagraph": 966 }, { "type": "bullet", "text": "Population version", "sourceParagraph": 967 }, { "type": "paragraph", "text": "Country quotas cannot be reproduced after results without this file.", "sourceParagraphs": [ 968 ] }, { "type": "heading2", "text": "30. SYNTHETIC PRODUCT AND COUNTRY DISPLAY", "sourceParagraph": 970 }, { "type": "paragraph", "text": "SYNTHETIC METHODOLOGY DEMONSTRATION / The following countries, AI product, population numbers, and access rates are entirely fictional. They do not represent any real country or product. Let's consider a synthetic general consumer product called Orion AI. There are four synthetic countries.", "sourceParagraphs": [ 971, 972, 973 ] }, { "type": "table", "rows": [ [ "Country", "Total population", "Estimated suitable population after age, internet, language, and interface conditions", "Product access", "Population of Native Reach" ], [ "Country A", "1,400,000,000", "780,000,000", "Unavailable", "0" ], [ "Country B", "1,450,000,000", "520,000,000", "Official access", "520,000,000" ], [ "Country C", "340,000,000", "250,000,000", "Official access", "250,000,000" ], [ "Country D", "35,000", "25,000", "Official access", "25,000" ], [ "Total", "3,190,035,000", "1,550,025,000", "", "770,025,000" ] ] }, { "type": "heading3", "text": "30.1. If the total population was used", "sourceParagraph": 975 }, { "type": "paragraph", "text": "Country A would have a large share in the main sample due to its very large population. However, the product cannot be used in that country. Participants could not produce a natural user experience.", "sourceParagraphs": [ 976, 977, 978 ] }, { "type": "heading3", "text": "30.2. When Native Reach Population Is Used", "sourceParagraph": 980 }, { "type": "paragraph", "text": "The theoretical population shares of the main product panel are approximately:", "sourceParagraphs": [ 981 ] }, { "type": "bullet", "text": "Country B: 67.53%", "sourceParagraph": 982 }, { "type": "bullet", "text": "Country C: 32.47%", "sourceParagraph": 983 }, { "type": "bullet", "text": "Country D: 0.003%", "sourceParagraph": 984 }, { "type": "bullet", "text": "Country A: outside of Native Reach", "sourceParagraph": 985 }, { "type": "paragraph", "text": "The theoretical share of Country D in a main sample of 1,000 people is far below one person. Therefore, the population-weighted main score should not give high weight to Country D. However, if the country is not intended to be completely invisible:", "sourceParagraphs": [ 986, 987, 988, 989, 990 ] }, { "type": "heading4", "text": "Country Observer Panel", "sourceParagraph": 991 }, { "type": "paragraph", "text": "At least one exploratory observation can be obtained. This observation:", "sourceParagraphs": [ 992, 993 ] }, { "type": "bullet", "text": "is not the score of Country D,", "sourceParagraph": 994 }, { "type": "bullet", "text": "do not change the main population weight,", "sourceParagraph": 995 }, { "type": "paragraph", "text": "shows that the product behaviour can be observed in the country.", "sourceParagraphs": [ 996 ] }, { "type": "heading3", "text": "30.3. How is Country A Reported?", "sourceParagraph": 998 }, { "type": "paragraph", "text": "For Country A:", "sourceParagraphs": [ 999 ] }, { "type": "heading3", "text": "The OUTSIDE_NATIVE_REACH", "sourceParagraph": 1000 }, { "type": "paragraph", "text": "status is used. The exclusion of Country A should not hide the product's access limits. The public results card also shows:", "sourceParagraphs": [ 1001, 1002, 1003 ] }, { "type": "bullet", "text": "Product access coverage", "sourceParagraph": 1004 }, { "type": "bullet", "text": "Share not accessible in the global reference population", "sourceParagraph": 1005 }, { "type": "bullet", "text": "Native Reach score", "sourceParagraph": 1006 }, { "type": "bullet", "text": "Common Support comparison score", "sourceParagraph": 1007 }, { "type": "heading2", "text": "31. SYNTHETIC PROMPT POPULATION REPRESENTATION", "sourceParagraph": 1009 }, { "type": "heading3", "text": "SYNTHETIC EXAMPLE", "sourceParagraph": 1010 }, { "type": "paragraph", "text": "Let's consider three prompts for the same Orion AI product.", "sourceParagraphs": [ 1011 ] }, { "type": "heading4", "text": "Prompt A", "sourceParagraph": 1013 }, { "type": "paragraph", "text": "\"What kind of company is Asteron?\" Target population: could be the broad suitable user population of Orion AI.", "sourceParagraphs": [ 1014, 1015, 1016, 1017 ] }, { "type": "heading4", "text": "Prompt B", "sourceParagraph": 1019 }, { "type": "paragraph", "text": "\"Does Asteron provide corporate travel services in Turkey?\" Target population:", "sourceParagraphs": [ 1020, 1021 ] }, { "type": "bullet", "text": "Eligible users in Turkey,", "sourceParagraph": 1022 }, { "type": "bullet", "text": "Identified external users researching the Turkey service", "sourceParagraph": 1023 }, { "type": "paragraph", "text": "may be.", "sourceParagraphs": [ 1024 ] }, { "type": "heading4", "text": "Prompt C", "sourceParagraph": 1026 }, { "type": "paragraph", "text": "“Is Asteron suitable for legal immigration consultancy for me?” If Asteron does not provide such a service: there may be no advice-eligible population, the correct answer should exclude all suitable users. Combining this prompt into the overall identity score might be incorrect. This example shows:", "sourceParagraphs": [ 1027, 1028, 1029, 1030, 1031, 1032 ] }, { "type": "callout", "text": "Even for the same entity and the same AI product, the target population may vary depending on the user question measured.", "sourceParagraph": 1033 }, { "type": "heading2", "text": "32. FOUR PANEL ARCHITECTURE", "sourceParagraph": 1035 }, { "type": "heading3", "text": "CANDIDATE ARCHITECTURE — CANDIDATE ARCHITECTURE", "sourceParagraph": 1036 }, { "type": "paragraph", "text": "Four complementary panels are recommended for the global and fair implementation of GEO-1000.", "sourceParagraphs": [ 1037 ] }, { "type": "heading3", "text": "32.1. Population Panel", "sourceParagraph": 1039 }, { "type": "paragraph", "text": "The main panel weighted according to the product-eligible user population. Produces the main population estimate.", "sourceParagraphs": [ 1040, 1041 ] }, { "type": "heading3", "text": "32.2. Country Observer Panel", "sourceParagraph": 1043 }, { "type": "paragraph", "text": "Prevents small countries that cannot provide sufficient samples in the population panel from being completely invisible. It is not sufficient on its own to produce a country score.", "sourceParagraphs": [ 1044, 1045 ] }, { "type": "heading3", "text": "32.3. Language Fairness Panel", "sourceParagraph": 1047 }, { "type": "paragraph", "text": "Also measures languages that are low-resource or suppressed in population averages. It reconnects to the global main score with real population weights.", "sourceParagraphs": [ 1048, 1049 ] }, { "type": "heading3", "text": "32.4. Accessibility Panel", "sourceParagraph": 1051 }, { "type": "paragraph", "text": "Measures accessibility needs and assistive technology users that the standard digital panel might exclude. Observations from these four panels:", "sourceParagraphs": [ 1052, 1053 ] }, { "type": "bullet", "text": "goal,", "sourceParagraph": 1054 }, { "type": "bullet", "text": "weight,", "sourceParagraph": 1055 }, { "type": "bullet", "text": "inference universe", "sourceParagraph": 1056 }, { "type": "paragraph", "text": "It should be labelled separately in terms of maintenance. The same observation cannot be used at full weight on more than one panel without explanation.", "sourceParagraphs": [ 1057, 1058 ] }, { "type": "heading2", "text": "33. ADULT INITIAL PANEL", "sourceParagraph": 1060 }, { "type": "heading3", "text": "CANDIDATE BASE RULE — CANDIDATE FOUNDATIONAL RULE", "sourceParagraph": 1061 }, { "type": "paragraph", "text": "In the first general release of GEO-1000, the real user dashboards:", "sourceParagraphs": [ 1062 ] }, { "type": "heading4", "text": "from adult participants who meet the product and country conditions", "sourceParagraph": 1063 }, { "type": "paragraph", "text": "It is recommended to be created. This approach:", "sourceParagraphs": [ 1064, 1065 ] }, { "type": "bullet", "text": "consent,", "sourceParagraph": 1066 }, { "type": "bullet", "text": "data protection,", "sourceParagraph": 1067 }, { "type": "bullet", "text": "collecting screenshots,", "sourceParagraph": 1068 }, { "type": "bullet", "text": "high-risk research,", "sourceParagraph": 1069 }, { "type": "bullet", "text": "parental consent", "sourceParagraph": 1070 }, { "type": "paragraph", "text": "It can reduce its complexity. However, the following conclusion cannot be drawn: 'The experience of child and young users has been measured with the adult panel.' In products that affect children or young people, separately:", "sourceParagraphs": [ 1071, 1072, 1073, 1074 ] }, { "type": "bullet", "text": "ethical approval,", "sourceParagraph": 1075 }, { "type": "bullet", "text": "age-appropriate protocol,", "sourceParagraph": 1076 }, { "type": "bullet", "text": "parent or guardian processes,", "sourceParagraph": 1077 }, { "type": "bullet", "text": "harm protection", "sourceParagraph": 1078 }, { "type": "paragraph", "text": "are required.", "sourceParagraphs": [ 1079 ] }, { "type": "heading2", "text": "34. NO POPULATION FRAMEWORK CAN BE CALLED \"ALL HUMANITY\"", "sourceParagraph": 1081 }, { "type": "paragraph", "text": "A product:", "sourceParagraphs": [ 1082 ] }, { "type": "bullet", "text": "if it does not exist in some countries,", "sourceParagraph": 1083 }, { "type": "bullet", "text": "if it does not support some languages,", "sourceParagraph": 1084 }, { "type": "bullet", "text": "if it only works on certain devices,", "sourceParagraph": 1085 }, { "type": "bullet", "text": "if it has certain age or plan condition,", "sourceParagraph": 1086 }, { "type": "paragraph", "text": "result: it cannot be presented as “the GEO experience of all the people in the world”. The correct expression:", "sourceParagraphs": [ 1087, 1088, 1089, 1090 ] }, { "type": "heading4", "text": "Estimated representation in the target population suitable for the specified product", "sourceParagraph": 1091 }, { "type": "paragraph", "text": "should be used. The word global:", "sourceParagraphs": [ 1092, 1093 ] }, { "type": "bullet", "text": "multi-country,", "sourceParagraph": 1094 }, { "type": "bullet", "text": "multi-lingual,", "sourceParagraph": 1095 }, { "type": "bullet", "text": "large population", "sourceParagraph": 1096 }, { "type": "paragraph", "text": "can define a measurement. It does not mean unlimited and universal coverage.", "sourceParagraphs": [ 1097, 1098 ] }, { "type": "heading2", "text": "35. FOUNDATIONAL ETHICAL LIMIT OF THE POPULATION", "sourceParagraph": 1100 }, { "type": "paragraph", "text": "Defining a population for measurement does not mean considering some people worthless. A group may receive a low weight in the main score. However:", "sourceParagraphs": [ 1101, 1102, 1103 ] }, { "type": "bullet", "text": "high risk,", "sourceParagraph": 1104 }, { "type": "bullet", "text": "language fairness,", "sourceParagraph": 1105 }, { "type": "bullet", "text": "accessibility,", "sourceParagraph": 1106 }, { "type": "bullet", "text": "country visibility", "sourceParagraph": 1107 }, { "type": "paragraph", "text": "may also be examined separately. A group may be excluded from the Native Reach score because they cannot access a product. This lack of access can also be reported as a global coverage gap. The purpose of NOMOS:", "sourceParagraphs": [ 1108, 1109, 1110, 1111 ] }, { "type": "bullet", "text": "compressing everyone into a single artificial average,", "sourceParagraph": 1112 }, { "type": "bullet", "text": "seeing only the largest populations,", "sourceParagraph": 1113 }, { "type": "bullet", "text": "appearing to symbolically represent small communities with a single person", "sourceParagraph": 1114 }, { "type": "paragraph", "text": "is not the goal. The objective is:", "sourceParagraphs": [ 1115, 1116 ] }, { "type": "callout", "text": "To represent each person in the correct denominator of the correct question, without mixing their weight and visibility.", "sourceParagraph": 1117 }, { "type": "heading2", "text": "36. MANDATORY NORMATIVE PROVISIONS", "sourceParagraph": 1119 }, { "type": "protocol_label", "text": "CH05-N01", "sourceParagraph": 1120 }, { "type": "paragraph", "text": "The target user population must be defined separately for each AI product.", "sourceParagraphs": [ 1121 ] }, { "type": "protocol_label", "text": "CH05-N02", "sourceParagraph": 1122 }, { "type": "paragraph", "text": "The total population of a country cannot be directly used as the AI-eligible user population.", "sourceParagraphs": [ 1123 ] }, { "type": "protocol_label", "text": "CH05-N03", "sourceParagraph": 1124 }, { "type": "paragraph", "text": "The target population must be linked to product, country, age, access, interface, plan, language, prompt, and time conditions.", "sourceParagraphs": [ 1125 ] }, { "type": "protocol_label", "text": "CH05-N04", "sourceParagraph": 1126 }, { "type": "paragraph", "text": "Target population and exclusion rules must be versioned before seeing the results.", "sourceParagraphs": [ 1127 ] }, { "type": "protocol_label", "text": "CH05-N05", "sourceParagraph": 1128 }, { "type": "paragraph", "text": "The target population of an AI product cannot be transferred to another AI product without explanation.", "sourceParagraphs": [ 1129 ] }, { "type": "protocol_label", "text": "CH05-N06", "sourceParagraph": 1130 }, { "type": "paragraph", "text": "Lack of product access cannot be scored as a misrepresentation result; it must be reported separately as access coverage.", "sourceParagraphs": [ 1131 ] }, { "type": "protocol_label", "text": "CH05-N07", "sourceParagraph": 1132 }, { "type": "paragraph", "text": "Participants cannot be asked to bypass product or country access restrictions.", "sourceParagraphs": [ 1133 ] }, { "type": "protocol_label", "text": "CH05-N08", "sourceParagraph": 1134 }, { "type": "paragraph", "text": "Access not officially supported cannot be automatically added to the main Native Reach population.", "sourceParagraphs": [ 1135 ] }, { "type": "protocol_label", "text": "CH05-N09", "sourceParagraph": 1136 }, { "type": "paragraph", "text": "Individual, account, and session units must be separated from each other.", "sourceParagraphs": [ 1137 ] }, { "type": "protocol_label", "text": "CH05-N10", "sourceParagraph": 1138 }, { "type": "paragraph", "text": "Multiple accounts or languages of a person cannot multiply the individual in population weight.", "sourceParagraphs": [ 1139 ] }, { "type": "protocol_label", "text": "CH05-N11", "sourceParagraph": 1140 }, { "type": "paragraph", "text": "Language eligibility cannot be assigned based solely on the country's official language.", "sourceParagraphs": [ 1141 ] }, { "type": "protocol_label", "text": "CH05-N12", "sourceParagraph": 1142 }, { "type": "paragraph", "text": "Requested language, interface language, response language, and user proficiency language must be recorded separately.", "sourceParagraphs": [ 1143 ] }, { "type": "protocol_label", "text": "CH05-N13", "sourceParagraph": 1144 }, { "type": "paragraph", "text": "The total population weight of a person using multiple languages cannot exceed one person without explanation.", "sourceParagraphs": [ 1145 ] }, { "type": "protocol_label", "text": "CH05-N14", "sourceParagraph": 1146 }, { "type": "paragraph", "text": "The population-weighted panel and the Country Observer panel should be kept separate.", "sourceParagraphs": [ 1147 ] }, { "type": "protocol_label", "text": "CH05-N15", "sourceParagraph": 1148 }, { "type": "paragraph", "text": "One or several Country Observer observations cannot be published as a country score.", "sourceParagraphs": [ 1149 ] }, { "type": "protocol_label", "text": "CH05-N16", "sourceParagraph": 1150 }, { "type": "paragraph", "text": "In the Language Fairness Panel, over-sampled observations should be reverted to their true population weight in the global score.", "sourceParagraphs": [ 1151 ] }, { "type": "protocol_label", "text": "CH05-N17", "sourceParagraph": 1152 }, { "type": "paragraph", "text": "Users with accessibility needs cannot be excluded from the target population because the data collection tool is not suitable.", "sourceParagraphs": [ 1153 ] }, { "type": "protocol_label", "text": "CH05-N18", "sourceParagraph": 1154 }, { "type": "paragraph", "text": "The target population, sampling frame, contacted population, participants, and valid analytical observations should be reported separately.", "sourceParagraphs": [ 1155 ] }, { "type": "protocol_label", "text": "CH05-N19", "sourceParagraph": 1156 }, { "type": "paragraph", "text": "A non-probability or voluntary panel must not be represented as a random sample of the global population.", "sourceParagraphs": [ 1157 ] }, { "type": "protocol_label", "text": "CH05-N20", "sourceParagraph": 1158 }, { "type": "paragraph", "text": "Weighting cannot make a group that is completely absent in the sampling frame observable.", "sourceParagraphs": [ 1159 ] }, { "type": "protocol_label", "text": "CH05-N21", "sourceParagraph": 1160 }, { "type": "paragraph", "text": "Potential suitable population and active user exposure cannot be used as the same metric.", "sourceParagraphs": [ 1161 ] }, { "type": "protocol_label", "text": "CH05-N22", "sourceParagraph": 1162 }, { "type": "paragraph", "text": "The results Native Reach and Common Support must be separated from each other.", "sourceParagraphs": [ 1163 ] }, { "type": "protocol_label", "text": "CH05-N23", "sourceParagraph": 1164 }, { "type": "paragraph", "text": "The same target population cannot automatically be used for identity, service, price, legal, and recommendation prompts.", "sourceParagraphs": [ 1165 ] }, { "type": "protocol_label", "text": "CH05-N24", "sourceParagraph": 1166 }, { "type": "paragraph", "text": "Population data, product access status, and language distribution must carry a timestamp and source version.", "sourceParagraphs": [ 1167 ] }, { "type": "protocol_label", "text": "CH05-N25", "sourceParagraph": 1168 }, { "type": "paragraph", "text": "Marginal utility ratios cannot be blindly multiplied without an explanation of the independence assumption.", "sourceParagraphs": [ 1169 ] }, { "type": "protocol_label", "text": "CH05-N26", "sourceParagraph": 1170 }, { "type": "paragraph", "text": "If the product type is not suitable for the general consumer panel, the GEO-1000 human population method cannot be applied without explanation.", "sourceParagraphs": [ 1171 ] }, { "type": "protocol_label", "text": "CH05-N27", "sourceParagraph": 1172 }, { "type": "paragraph", "text": "Children or underage users cannot be included in the main panel without a separate ethical and legal protocol.", "sourceParagraphs": [ 1173 ] }, { "type": "protocol_label", "text": "CH05-N28", "sourceParagraph": 1174 }, { "type": "paragraph", "text": "Product access coverage and representational accuracy must be published as separate results.", "sourceParagraphs": [ 1175 ] }, { "type": "protocol_label", "text": "CH05-N29", "sourceParagraph": 1176 }, { "type": "paragraph", "text": "The term global cannot make countries, languages, ages, interfaces, or plans that are out of scope invisible.", "sourceParagraphs": [ 1177 ] }, { "type": "protocol_label", "text": "CH05-N30", "sourceParagraph": 1178 }, { "type": "paragraph", "text": "The target population record must have a responsible human or institutional owner.", "sourceParagraphs": [ 1179 ] }, { "type": "heading2", "text": "37. FORMS OF FAILURE", "sourceParagraph": 1181 }, { "type": "protocol_label", "text": "CH05-F01 — COUNT THE TOTAL POPULATION AS ELIGIBLE POPULATION", "sourceParagraph": 1182 }, { "type": "paragraph", "text": "The country's population is directly included in the denominator without evaluating product, age, internet, language, and interface conditions.", "sourceParagraphs": [ 1183 ] }, { "type": "protocol_label", "text": "CH05-F02 — APPLY THE SAME WORLD MAP TO ALL PRODUCTS", "sourceParagraph": 1184 }, { "type": "paragraph", "text": "AI products with different access areas are measured against the same target population.", "sourceParagraphs": [ 1185 ] }, { "type": "protocol_label", "text": "CH05-F03 — COUNT A USER WHO EXCEEDS THE LIMIT AS A NATURAL USER", "sourceParagraph": 1186 }, { "type": "paragraph", "text": "Observations that exceed official access conditions are presented as a country population experience.", "sourceParagraphs": [ 1187 ] }, { "type": "protocol_label", "text": "CH05-F04 — COUNTING PRODUCT ABSENCE AS WRONG ANSWER", "sourceParagraph": 1188 }, { "type": "paragraph", "text": "Countries that are inaccessible are considered zero in representation accuracy.", "sourceParagraphs": [ 1189 ] }, { "type": "protocol_label", "text": "CH05-F05 — COMPLETELY HIDING PRODUCT ABSENCE", "sourceParagraph": 1190 }, { "type": "paragraph", "text": "Large populations that are inaccessible are removed from the Native Reach denominator; access coverage is not shown either.", "sourceParagraphs": [ 1191 ] }, { "type": "protocol_label", "text": "CH05-F06 — COUNTING ONE PERSON AS MORE THAN ONE", "sourceParagraph": 1192 }, { "type": "paragraph", "text": "The same person's account, language, or product observations multiply the population weight.", "sourceParagraphs": [ 1193 ] }, { "type": "protocol_label", "text": "CH05-F07 — COUNTING ACCOUNT AS PERSON", "sourceParagraph": 1194 }, { "type": "paragraph", "text": "Multiple account ownership or shared account usage is not considered.", "sourceParagraphs": [ 1195 ] }, { "type": "protocol_label", "text": "CH05-F08 — COUNTING OFFICIAL LANGUAGE AS USER LANGUAGE", "sourceParagraph": 1196 }, { "type": "paragraph", "text": "The official language of the country is considered the audit language for all participants.", "sourceParagraphs": [ 1197 ] }, { "type": "protocol_label", "text": "CH05-F09 — COUNTING MULTILINGUAL INDIVIDUALS TWICE", "sourceParagraph": 1198 }, { "type": "paragraph", "text": "The same individual receives full population weight in each language.", "sourceParagraphs": [ 1199 ] }, { "type": "protocol_label", "text": "CH05-F10 — ASSIGNING ONE PERSON TO A SMALL COUNTRY AND EXPLAINING THE COUNTRY SCORE", "sourceParagraph": 1200 }, { "type": "paragraph", "text": "Country Observer observation is transformed into a statistical country estimate.", "sourceParagraphs": [ 1201 ] }, { "type": "protocol_label", "text": "CH05-F11 — COMPLETELY DELETING SMALL COUNTRIES", "sourceParagraph": 1202 }, { "type": "paragraph", "text": "Because the population share is low, the country or language is not included in any additional coverage panel.", "sourceParagraphs": [ 1203 ] }, { "type": "protocol_label", "text": "CH05-F12 — OVERWEIGHTING EXAMPLES OF LANGUAGE FAIRNESS IN THE MAIN SCORE", "sourceParagraph": 1204 }, { "type": "paragraph", "text": "Over-sampled languages are not returned to their true population weight.", "sourceParagraphs": [ 1205 ] }, { "type": "protocol_label", "text": "CH05-F13 — COUNTING THE ACTIVE USER AS POTENTIAL POPULATION", "sourceParagraph": 1206 }, { "type": "paragraph", "text": "Actual usage data is presented as if it includes all people who could access the product.", "sourceParagraphs": [ 1207 ] }, { "type": "protocol_label", "text": "CH05-F14 — COUNTING THE POTENTIAL POPULATION AS ACTUAL EXPOSURE", "sourceParagraph": 1208 }, { "type": "paragraph", "text": "It is assumed that the response of everyone who could use the product is actually seen.", "sourceParagraphs": [ 1209 ] }, { "type": "protocol_label", "text": "CH05-F15 — COUNTING THE VOLUNTARY PANEL AS A RANDOM WORLD SAMPLE", "sourceParagraph": 1210 }, { "type": "paragraph", "text": "Self-selection and panel bias are concealed.", "sourceParagraphs": [ 1211 ] }, { "type": "protocol_label", "text": "CH05-F16 — IGNORING COVERAGE ERROR WITH WEIGHTING", "sourceParagraph": 1212 }, { "type": "paragraph", "text": "It is claimed that groups not present in the frame are represented solely by weighting.", "sourceParagraphs": [ 1213 ] }, { "type": "protocol_label", "text": "CH05-F17 — REMOVING USERS EXCLUDED BY THE DATA COLLECTION TOOL FROM THE POPULATION", "sourceParagraph": 1214 }, { "type": "paragraph", "text": "Accessibility or device issues are interpreted as if the user is not suitable.", "sourceParagraphs": [ 1215 ] }, { "type": "protocol_label", "text": "CH05-F18 — APPLYING THE SAME POPULATION TO ALL PROMPTS", "sourceParagraph": 1216 }, { "type": "paragraph", "text": "Global identity, local pricing, and legal advice questions are combined on the same denominator.", "sourceParagraphs": [ 1217 ] }, { "type": "protocol_label", "text": "CH05-F19 — APPLYING GENERAL CONSUMER POPULATION TO THE API PRODUCT", "sourceParagraph": 1218 }, { "type": "paragraph", "text": "The human user panel is transferred directly to the developer or application product.", "sourceParagraphs": [ 1219 ] }, { "type": "protocol_label", "text": "CH05-F20 — SECRETLY COMBINING FREE AND PAID PLANS", "sourceParagraph": 1220 }, { "type": "paragraph", "text": "A single product distribution is published without recording the plan difference.", "sourceParagraphs": [ 1221 ] }, { "type": "protocol_label", "text": "CH05-F21 — UNIFYING WEB AND MOBILE POPULATIONS", "sourceParagraph": 1222 }, { "type": "paragraph", "text": "Interface access and behaviour differences remain invisible.", "sourceParagraphs": [ 1223 ] }, { "type": "protocol_label", "text": "CH05-F22 — IGNORING THE AGE LIMIT", "sourceParagraph": 1224 }, { "type": "paragraph", "text": "Inappropriate age groups are added to the target population or panel without explanation.", "sourceParagraphs": [ 1225 ] }, { "type": "protocol_label", "text": "CH05-F23 — MULTIPLYING MARGINAL RATES AS IF INDEPENDENT", "sourceParagraph": 1226 }, { "type": "paragraph", "text": "The relationship between age, internet, language, and access is ignored.", "sourceParagraphs": [ 1227 ] }, { "type": "protocol_label", "text": "CH05-F24 — CHANGING THE POPULATION AFTER THE RESULT", "sourceParagraph": 1228 }, { "type": "paragraph", "text": "A low-scoring country or user group is removed from the target population.", "sourceParagraphs": [ 1229 ] }, { "type": "protocol_label", "text": "CH05-F25 — USING THE OLD PRODUCT REACH MAP", "sourceParagraph": 1230 }, { "type": "paragraph", "text": "Actual product reach on the measurement date is not verified.", "sourceParagraphs": [ 1231 ] }, { "type": "protocol_label", "text": "CH05-F26 — CONFUSING COMMON SUPPORT AND NATIVE REACH RESULTS", "sourceParagraph": 1232 }, { "type": "paragraph", "text": "Comparisons made on joint population are presented as if they reflect the product's actual global reach.", "sourceParagraphs": [ 1233 ] }, { "type": "protocol_label", "text": "CH05-F27 — NARROWING THE TARGET MARKET ACCORDING TO COMMERCIAL INTEREST", "sourceParagraph": 1234 }, { "type": "paragraph", "text": "The organisation only declares the countries where it succeeds as target markets.", "sourceParagraphs": [ 1235 ] }, { "type": "protocol_label", "text": "CH05-F28 — COUNTING THE SCOPE OF ACCESS AS GEO ACCURACY", "sourceParagraph": 1236 }, { "type": "paragraph", "text": "Being present in many countries is presented as proof of accurate representation.", "sourceParagraphs": [ 1237 ] }, { "type": "protocol_label", "text": "CH05-F29 — SELLING LIMITED ACCESS AS HIGH GEO SUCCESS", "sourceParagraph": 1238 }, { "type": "paragraph", "text": "High accuracy in a small number of countries is presented as global superiority.", "sourceParagraphs": [ 1239 ] }, { "type": "protocol_label", "text": "CH05-F30 — USING THE TERM “THE WHOLE WORLD” WITHOUT COVERAGE RECORD", "sourceParagraph": 1240 }, { "type": "paragraph", "text": "Exclusions of product, language, country, and plan are hidden.", "sourceParagraphs": [ 1241 ] }, { "type": "heading2", "text": "38. AUDIT PROCEDURE", "sourceParagraph": 1243 }, { "type": "heading3", "text": "Step 1 — Lock the AI Product and User surface", "sourceParagraph": 1244 }, { "type": "paragraph", "text": "Product Interface Plan Account type Displayed model Measurement date is recorded.", "sourceParagraphs": [ 1245, 1246, 1247, 1248, 1249, 1250, 1251 ] }, { "type": "heading3", "text": "Step 2 — Classify the Measurement Question", "sourceParagraph": 1253 }, { "type": "paragraph", "text": "Prompt:", "sourceParagraphs": [ 1254 ] }, { "type": "bullet", "text": "global identity,", "sourceParagraph": 1255 }, { "type": "bullet", "text": "local service,", "sourceParagraph": 1256 }, { "type": "bullet", "text": "price,", "sourceParagraph": 1257 }, { "type": "bullet", "text": "law,", "sourceParagraph": 1258 }, { "type": "bullet", "text": "recommendation,", "sourceParagraph": 1259 }, { "type": "bullet", "text": "high risk", "sourceParagraph": 1260 }, { "type": "paragraph", "text": "which of their families does it belong to?", "sourceParagraphs": [ 1261 ] }, { "type": "heading3", "text": "Step 3 — Determine the Global Reference Population", "sourceParagraph": 1263 }, { "type": "paragraph", "text": "The initial human universe to which the result will be associated is defined.", "sourceParagraphs": [ 1264 ] }, { "type": "heading3", "text": "Step 4 — Determine Age and Research Eligibility", "sourceParagraph": 1266 }, { "type": "paragraph", "text": "Product age condition Research age condition Consent and data processing terms are recorded.", "sourceParagraphs": [ 1267, 1268, 1269, 1270 ] }, { "type": "heading3", "text": "Step 5 — Verify Country-Based Product Access", "sourceParagraph": 1272 }, { "type": "paragraph", "text": "An access status between PA-0 and PA-5 is assigned for each country.", "sourceParagraphs": [ 1273 ] }, { "type": "heading3", "text": "Step 6 — Determine Internet, Device, and Interface Compatibility", "sourceParagraph": 1275 }, { "type": "paragraph", "text": "The population that can use the measured surface is estimated.", "sourceParagraphs": [ 1276 ] }, { "type": "heading3", "text": "Step 7 — Establish the Language and Locale Universe", "sourceParagraph": 1278 }, { "type": "paragraph", "text": "Prompt language, Interface language, Response language, and User language proficiency are separated.", "sourceParagraphs": [ 1279, 1280, 1281, 1282, 1283 ] }, { "type": "heading3", "text": "Step 8 — Apply Prompt Relevance", "sourceParagraph": 1285 }, { "type": "paragraph", "text": "It is evaluated whether all users belong to the target population of the relevant prompt family.", "sourceParagraphs": [ 1286 ] }, { "type": "heading3", "text": "Step 9 — Select the Population Model Appropriate to the Product Type", "sourceParagraph": 1288 }, { "type": "paragraph", "text": "Consumer, enterprise, education, API, or embedded product is separated.", "sourceParagraphs": [ 1289 ] }, { "type": "heading3", "text": "Step 10 — Compare the Target Population with the Sampling Frame", "sourceParagraph": 1291 }, { "type": "paragraph", "text": "It is recorded as under-coverage and over-coverage.", "sourceParagraphs": [ 1292 ] }, { "type": "heading3", "text": "Step 11 — Identify the Main and Complementary Panels", "sourceParagraph": 1294 }, { "type": "paragraph", "text": "The scopes of the Population Panel, Country Observer Panel, Language Fairness Panel, and Accessibility Panel are versioned.", "sourceParagraphs": [ 1295, 1296, 1297, 1298, 1299 ] }, { "type": "heading3", "text": "Step 12 — Separate the Native Reach and Common Support Universes", "sourceParagraph": 1301 }, { "type": "paragraph", "text": "Intra-product results and inter-product comparison populations are recorded separately.", "sourceParagraphs": [ 1302 ] }, { "type": "heading3", "text": "Step 13 — Separate Access Scope and Accuracy Results", "sourceParagraph": 1304 }, { "type": "paragraph", "text": "The accuracy between the population a product can reach and the users reached is reported separately.", "sourceParagraphs": [ 1305 ] }, { "type": "heading3", "text": "Step 14 — Freeze the Population File", "sourceParagraph": 1307 }, { "type": "paragraph", "text": "Data sources, assumptions, date, and population version are locked.", "sourceParagraphs": [ 1308 ] }, { "type": "heading3", "text": "Step 15 — Track Material Changes", "sourceParagraph": 1310 }, { "type": "paragraph", "text": "If product access or conditions change during measurement, a new version is created.", "sourceParagraphs": [ 1311 ] }, { "type": "heading2", "text": "39. REQUIRED EVIDENCE", "sourceParagraph": 1313 }, { "type": "paragraph", "text": "AI-product identity; user surface; plan and account conditions; measurement date; countries in which the product is accessible; country-level access status; age conditions; legal and ethical participation conditions; country-population records; age distributions; internet and device-access records; language and locale distributions; product-language support; interface accessibility; target prompt family; target-population definition; sampling frame; undercoverage; overcoverage; panel type; participant-selection method; probability or opt-in panel status; multilingual-user weights; Country Observer coverage; Language Fairness coverage; Accessibility Panel coverage; Native Reach population; Common Support population; active-user data, where used; access-coverage calculation; assumptions and sensitivity analyses; population version; and last verification date.", "sourceParagraphs": [ 1314, 1315, 1316, 1317, 1318, 1319, 1320, 1321, 1322, 1323, 1324, 1325, 1326, 1327, 1328, 1329, 1330, 1331, 1332, 1333, 1334, 1335, 1336, 1337, 1338, 1339, 1340, 1341, 1342, 1343, 1344, 1345, 1346 ] }, { "type": "paragraph", "text": "Responsible person or institution", "sourceParagraphs": [ 1347 ] }, { "type": "heading2", "text": "40. AUDIT CHECKLIST", "sourceParagraph": 1349 }, { "type": "paragraph", "text": "Which AI product and user surface is being measured? For which date was the target population defined? Were the total country population and the eligible user population separated? Was an age requirement applied? Was internet and device access evaluated? Is the product officially accessible on a country basis? Were participants asked to overcome access restrictions? Were free and paid plans separated? Were web and mobile interfaces separated? Did the prompt family change the target population? How was language and locale suitability determined? Were multilingual users counted twice? Is the difference between the target population and the sampling frame clear? Is the panel probabilistic or voluntary? Was coverage error measured? Were users with accessibility needs silently excluded?", "sourceParagraphs": [ 1350, 1351, 1352, 1353, 1354, 1355, 1356, 1357, 1358, 1359, 1360, 1361, 1362, 1363, 1364, 1365 ] }, { "type": "paragraph", "text": "Have Country Observer observations been converted into a country score? Did Language Fairness examples have excessive weight in the main score? Were the results of Native Reach and Common Support separated? Are coverage and representation accuracy separate? Were active user and potential user results separated? Is the product type suitable for the human population panel? Were the marginal rates in population weights multiplied blindly? Was the population definition changed after the results were seen? Are the used population and access data versioned? Are untested groups visible? Is the term “global” consistent with the actual coverage? Is there a clear accountable owner of the population record?", "sourceParagraphs": [ 1366, 1367, 1368, 1369, 1370, 1371, 1372, 1373, 1374, 1375, 1376, 1377 ] }, { "type": "heading2", "text": "41. OBJECTIONS AND RESPONSES", "sourceParagraph": 1379 }, { "type": "heading3", "text": "Objection 1 — “Isn’t it fairer to distribute directly according to the world population?”", "sourceParagraph": 1380 }, { "type": "paragraph", "text": "The world population is a strong starting base. However, including people who cannot use the product in the main product sample does not generate real user experience. The correct method is:", "sourceParagraphs": [ 1381, 1382, 1383 ] }, { "type": "bullet", "text": "population weight within the product-eligible population,", "sourceParagraph": 1384 }, { "type": "bullet", "text": "global access coverage.", "sourceParagraph": 1385 }, { "type": "paragraph", "text": "Both results must be reported.", "sourceParagraphs": [ 1386 ] }, { "type": "heading3", "text": "Objection 2 — “Shouldn’t most users come from large countries?”", "sourceParagraph": 1388 }, { "type": "paragraph", "text": "Countries that meet the product, language, and access conditions should naturally have a high share. However, just because the total population is large, a fake user experience cannot be generated from a country where the product is not accessible.", "sourceParagraphs": [ 1389, 1390 ] }, { "type": "heading3", "text": "Objection 3 — “Why would it be wrong to give one person to a small country like San Marino?”", "sourceParagraph": 1392 }, { "type": "paragraph", "text": "A person is valuable for exploration and Country Observer observation. What is wrong is to present that person's result as a 'country score.' The visibility of the small country can be protected with the Country Observer Panel. The main population weight remains according to the actual population share.", "sourceParagraphs": [ 1393, 1394, 1395, 1396, 1397, 1398 ] }, { "type": "heading3", "text": "Objection 4 — “If we give at least one person to each country, wouldn't we look more global?”", "sourceParagraph": 1400 }, { "type": "paragraph", "text": "A wider geographical coverage is provided. However, the population estimate may be distorted. Therefore:", "sourceParagraphs": [ 1401, 1402, 1403 ] }, { "type": "bullet", "text": "geographical scope,", "sourceParagraph": 1404 }, { "type": "bullet", "text": "population density", "sourceParagraph": 1405 }, { "type": "paragraph", "text": "They are separate results. It may be a scope panel to give one person to each country. It is not a population panel.", "sourceParagraphs": [ 1406, 1407, 1408 ] }, { "type": "heading3", "text": "Objection 5 — \"If we give zero to a product in an inaccessible country, wouldn't the product be unfairly punished?\"", "sourceParagraph": 1410 }, { "type": "paragraph", "text": "An unreachable country is not counted as zero in representation accuracy. It becomes visible in the access coverage metric. This distinction protects the product from both unfair penalties and false universality.", "sourceParagraphs": [ 1411, 1412, 1413 ] }, { "type": "heading3", "text": "Objection 6 — “If there is real usage data, why do we need population data?”", "sourceParagraph": 1415 }, { "type": "paragraph", "text": "Real usage data shows operational impact. However, the product's:", "sourceParagraphs": [ 1416, 1417 ] }, { "type": "bullet", "text": "which populations it has not reached,", "sourceParagraph": 1418 }, { "type": "bullet", "text": "the geographical inequality in the user base,", "sourceParagraph": 1419 }, { "type": "bullet", "text": "potential user fairness", "sourceParagraph": 1420 }, { "type": "paragraph", "text": "It does not show alone. The two types of data answer different questions.", "sourceParagraphs": [ 1421, 1422 ] }, { "type": "heading3", "text": "Objection 7 — 'Since all users will ask the same prompt, why is prompt relevance important?'", "sourceParagraph": 1424 }, { "type": "paragraph", "text": "Identity questions may be meaningful in a broad population. Price, legal, or advice questions are only meaningful for certain users and markets. Irrelevant population may distort the response distribution away from the user reality.", "sourceParagraphs": [ 1425, 1426, 1427 ] }, { "type": "heading3", "text": "Objection 8 — \"If a person knows multiple languages, why shouldn’t they be fully counted in all languages?\"", "sourceParagraph": 1429 }, { "type": "paragraph", "text": "You can test all languages in cross-language comparison. However, multiplying the same person in population weighting artificially increases the world population. The person's weight should be divided, or a separate matched panel should be used.", "sourceParagraphs": [ 1430, 1431, 1432 ] }, { "type": "heading3", "text": "Objection 9 — \"Doesn’t testing only adults leave the global result incomplete?\"", "sourceParagraph": 1434 }, { "type": "paragraph", "text": "It does. This limitation should be clearly stated. Starting with an adult panel may be an ethical and operational choice. It does not make a judgement about child users.", "sourceParagraphs": [ 1435, 1436, 1437, 1438 ] }, { "type": "heading3", "text": "Objection 10 — 'Isn't the online research panel sufficient?'", "sourceParagraph": 1440 }, { "type": "paragraph", "text": "It can be very valuable for pilot and weighted research. But the panel:", "sourceParagraphs": [ 1441, 1442 ] }, { "type": "bullet", "text": "how much of the target population does it cover,", "sourceParagraph": 1443 }, { "type": "bullet", "text": "who is being left out,", "sourceParagraph": 1444 }, { "type": "bullet", "text": "how voluntary participation creates bias", "sourceParagraph": 1445 }, { "type": "paragraph", "text": "It should be explained. Using a panel is not a mistake. Presenting the panel as random like the world population is a mistake.", "sourceParagraphs": [ 1446, 1447, 1448 ] }, { "type": "heading3", "text": "Objection 11 — “Why is the Accessibility panel related to the main GEO score?”", "sourceParagraph": 1450 }, { "type": "paragraph", "text": "Because a portion of real AI users use assistive technology or different forms of interaction. If the data collection tool excludes these people, the main score may not represent the entire target population.", "sourceParagraphs": [ 1451, 1452 ] }, { "type": "heading3", "text": "Objection 12 — “Doesn't having so many different population definitions make a single NOMOS score impossible?”", "sourceParagraph": 1454 }, { "type": "paragraph", "text": "No. A single score can be produced. However, it should be clear which population question it answers. For example:", "sourceParagraphs": [ 1455, 1456, 1457, 1458 ] }, { "type": "bullet", "text": "Native Reach Population Score", "sourceParagraph": 1459 }, { "type": "bullet", "text": "Common Support Score", "sourceParagraph": 1460 }, { "type": "bullet", "text": "Market-Relevant Score", "sourceParagraph": 1461 }, { "type": "bullet", "text": "Language Fairness Result", "sourceParagraph": 1462 }, { "type": "paragraph", "text": "can carry separate records. The meaning of a number arises from the clarity of its denominator.", "sourceParagraphs": [ 1463, 1464 ] }, { "type": "heading2", "text": "COMMON JUDGEMENT OF CHAPTER 44", "sourceParagraph": 1595 }, { "type": "paragraph", "text": "GEO-1000 contains ‘1,000’ in its name, but the value of those 1,000 people does not arise from the number alone. It arises from the answers to these questions: Who are they? From which population were they drawn? Can they genuinely use the product? Where do they live? Which languages do they use? How many times was the same person counted? Are small countries visible? Do large countries retain their true population weight? Where does the score account for people the product cannot reach? Whom did the panel infrastructure exclude? Which result measures potential access, and which measures actual exposure? A large country may have a large product-eligible population and therefore deserve substantial weight in the main panel. But if the product is unavailable there, the audit must not fabricate users. The access gap must be disclosed separately.", "sourceParagraphs": [ 1596, 1597, 1598, 1599, 1600, 1601, 1602, 1603, 1604, 1605, 1606, 1607, 1608, 1609, 1610, 1611, 1612, 1613 ] }, { "type": "paragraph", "text": "If a country is very small, the main population weight should remain low. However, it does not have to be completely invisible. It can be observed with a monitoring panel. If a language has a small population, it may not determine the global average. However, if there is a severe representation distortion in that language, it should also be visible. If a person knows three languages, they can produce three different experiences. However, they are not three different people. If an AI product only provides certain behaviour on a paid plan, the score of free users does not represent that plan. A human population panel is suitable for a consumer chat product. The same population model may not be suitable for an API or enterprise product. Therefore, NOMOS's fifth measurement law is as follows:", "sourceParagraphs": [ 1614, 1615, 1616, 1617, 1618, 1619, 1620, 1621, 1622, 1623, 1624 ] }, { "type": "callout", "text": "Even the most sensitive score gives the correct answer to the wrong question if the denominator is incorrect.", "sourceParagraph": 1625 }, { "type": "paragraph", "text": "Its sixth law is:", "sourceParagraphs": [ 1626 ] }, { "type": "callout", "text": "Population weight does not replace country and language visibility.", "sourceParagraph": 1627 }, { "type": "paragraph", "text": "Its seventh law is as follows:", "sourceParagraphs": [ 1628 ] }, { "type": "callout", "text": "A person who cannot access [something] has not been misrepresented; however, they are part of the global access boundary.", "sourceParagraph": 1629 }, { "type": "paragraph", "text": "Its eighth law is as follows:", "sourceParagraphs": [ 1630 ] }, { "type": "callout", "text": "A person having multiple accounts, languages, or sessions does not increase the human population.", "sourceParagraph": 1631 }, { "type": "paragraph", "text": "The ninth law is as follows:", "sourceParagraphs": [ 1632 ] }, { "type": "callout", "text": "Global measurement is not a claim that everyone is measured; it is an honest record of who is measured and who is not.", "sourceParagraph": 1633 }, { "type": "heading2", "text": "Order of Section 5 of NOMOS", "sourceParagraph": 1635 }, { "type": "callout", "text": "Do not give me only country populations. / Show which people can actually use the product.", "sourceParagraph": 1636 }, { "type": "callout", "text": "Do not fill a large country with fake users when the product is not there.", "sourceParagraph": 1637 }, { "type": "callout", "text": "Do not give one person to a small country and declare that person's response as the voice of the country.", "sourceParagraph": 1638 }, { "type": "callout", "text": "Do not duplicate a person as much as their accounts, languages, and sessions.", "sourceParagraph": 1639 }, { "type": "callout", "text": "Do not mistake the official language for the real language usage of people.", "sourceParagraph": 1640 }, { "type": "callout", "text": "Do not make a free user the representative of a paid plan, or a mobile user the representative of the web interface.", "sourceParagraph": 1641 }, { "type": "callout", "text": "Do not call the volunteer panel a random world population.", "sourceParagraph": 1642 }, { "type": "callout", "text": "Do not erase a user with accessibility needs from the population because the tool does not work.", "sourceParagraph": 1643 }, { "type": "callout", "text": "Do not remove the population that the product does not reach from the accuracy score and hide the global access gap.", "sourceParagraph": 1644 }, { "type": "callout", "text": "Do not apply the same share to a price question as an identity question, or to a global brand as a local licence.", "sourceParagraph": 1645 }, { "type": "callout", "text": "First, define the product. / Then define the question. / Then define the human universe. / Then limit access, language, age, and interface. / Then explain whom you cannot select. / And only after that distribute 1,000 people.", "sourceParagraph": 1646 }, { "type": "heading2", "text": "The Chapter's Closing Sentence", "sourceParagraph": 1648 }, { "type": "callout", "text": "In GEO-1000, justice is not giving the same number of people to every country; it is representing each person in the correct target population, with the proper weighting, without hiding the places where they are not measured.", "sourceParagraph": 1649 }, { "type": "heading2", "text": "Normative Core" }, { "type": "normative", "text": "Every GEO population audit MUST define, before result inspection, a time-bound and product-specific target user population for each: - AI product and user surface, - country or jurisdiction, - age and lawful-access condition, - account or plan condition, - language and locale, - interface, - prompt family, - and measurement purpose. Total resident population MUST NOT be treated as product-eligible population without justified eligibility adjustments. Product unavailability MUST be reported as an access-coverage limitation, not silently scored as incorrect representation or silently removed from global scope. Potential access, active use, actual exposure, Native Reach, Common Support, market relevance, and research observability MUST remain distinct population concepts. Persons, accounts, sessions, languages, and product observations MUST NOT be treated as interchangeable population units. Oversampled countries, languages, and accessibility groups MUST be reweighted for population estimates while remaining separately visible for fairness and coverage analysis. Country Observer observations MAY establish geographic observation coverage, but MUST NOT be represented as statistically reliable country scores. Participants MUST NOT be required to bypass product, geographic, age, identity, or account restrictions in order to enter the official population panel. Every population frame MUST be versioned, sourced, uncertainty-aware, and attributable to an accountable human or organisation.", "translationRoute": "controlled-normative-rule" } ] }, { "number": 6, "id": "NOMOS-GEO-AUDIT-CH06", "title": "The Population-Proportional GEO-1000 Sample", "subtitle": null, "sourceFile": "6.ci bölüm.docx", "sourceSha256": "CF13DD3AF6C4AC35B7D35B8DC2C3096674BF4514B9FA0C4DFA526890C9C7414E", "sourceWordCount": 8195, "sourceIds": [ "K04", "K05", "K07", "K08" ], "blocks": [ { "type": "heading2", "text": "Chapter Boundary", "sourceParagraph": 16 }, { "type": "paragraph", "text": "Section 5 established the following provision:", "sourceParagraphs": [ 17 ] }, { "type": "callout", "text": "The target user population of each AI product must be defined separately. The total population of a country is not the product-eligible user population in that country.", "sourceParagraph": 18 }, { "type": "paragraph", "text": "Now we come to the next question:", "sourceParagraphs": [ 19 ] }, { "type": "callout", "text": "From the defined eligible user population for GEO-1000, how many people will be selected, from which country, which language, and by which method?", "sourceParagraph": 20 }, { "type": "paragraph", "text": "The eligible user population for an AI product is:", "sourceParagraphs": [ 21 ] }, { "type": "bullet", "text": "40% in country A,", "sourceParagraph": 22 }, { "type": "bullet", "text": "25% in country B,", "sourceParagraph": 23 }, { "type": "bullet", "text": "15% in country C,", "sourceParagraph": 24 }, { "type": "bullet", "text": "0.04% in a small country", "sourceParagraph": 25 }, { "type": "paragraph", "text": "If available, how should 1,000 valid observations be distributed? How will fractional quotas be converted into whole numbers? Will countries with a population share of less than one person be completely excluded? Will population weighting be distorted when at least one person is given to each country? Can small countries be grouped into a separate pool? If some languages are oversampled for language fairness, how will the global score be calculated? If the target is 100 valid observations in a country, how many people should be invited? Who will replace the participant if they do not complete the task? Can a user giving wrong answers be replaced? If the same person tests more than one AI product, will the sample still be independent?", "sourceParagraphs": [ 26, 27, 28, 29, 30, 31, 32, 33, 34, 35 ] }, { "type": "paragraph", "text": "Even with a raw sample of 1,000 people, can highly unequal weights reduce the effective statistical information below that of 1,000 independent, equally weighted observations? These are not merely operational details. The sample design determines how much influence each person has on the GEO score. A defective sample:", "sourceParagraphs": [ 36, 37, 38, 39 ] }, { "type": "bullet", "text": "can misweight the correct answers,", "sourceParagraph": 40 }, { "type": "bullet", "text": "can symbolically overrepresent small countries,", "sourceParagraph": 41 }, { "type": "bullet", "text": "can make large populations appear smaller than they are,", "sourceParagraph": 42 }, { "type": "bullet", "text": "can make low-resource languages invisible,", "sourceParagraph": 43 }, { "type": "bullet", "text": "can present voluntary technology users as if they were the world population.", "sourceParagraph": 44 }, { "type": "paragraph", "text": "It can be adjusted to achieve the desired score after reviewing the results. The previously established audit architecture requires that each audit be measurable and that the model/product carry the date, country, language, query set, repetition count, and measurement record. This section applies the same discipline to sample design. Because genuine evidence requires requesting limits, context, and time; it also necessitates making how the sample was created visible. This section:", "sourceParagraphs": [ 45, 46, 47 ] }, { "type": "bullet", "text": "how the 1,000 valid observations of the main population panel will be distributed,", "sourceParagraph": 48 }, { "type": "bullet", "text": "how strata and cells will be established,", "sourceParagraph": 49 }, { "type": "bullet", "text": "how small countries will be managed,", "sourceParagraph": 50 }, { "type": "bullet", "text": "how fractions will be converted to whole numbers,", "sourceParagraph": 51 }, { "type": "bullet", "text": "how oversampling will be weighted,", "sourceParagraph": 52 }, { "type": "bullet", "text": "how to separate invitation, backup, and valid observation targets,", "sourceParagraph": 53 }, { "type": "bullet", "text": "how to establish design and analysis weights,", "sourceParagraph": 54 }, { "type": "bullet", "text": "why the effective sample size may differ from the raw number", "sourceParagraph": 55 }, { "type": "paragraph", "text": "defines. This chapter does not yet:", "sourceParagraphs": [ 56, 57 ] }, { "type": "bullet", "text": "from which panel provider the participants will be selected,", "sourceParagraph": 58 }, { "type": "bullet", "text": "the full details of field implementation in voluntary or probability sampling,", "sourceParagraph": 59 }, { "type": "bullet", "text": "reviewing and scoring responses,", "sourceParagraph": 60 }, { "type": "bullet", "text": "the final confidence interval method,", "sourceParagraph": 61 }, { "type": "bullet", "text": "all NOMOS scoring formulas", "sourceParagraph": 62 }, { "type": "paragraph", "text": "are not fully finalised. The task of Section 6 is:", "sourceParagraphs": [ 63, 64 ] }, { "type": "callout", "text": "Transform the defined target population into a frozen, weightable, and reproducible GEO-1000 sample before seeing the outcome.", "sourceParagraph": 65 }, { "type": "heading2", "text": "NOMOS Challenge", "sourceParagraph": 67 }, { "type": "paragraph", "text": "You are setting up a global panel of 1,000 people for an AI product. The product-eligible populations of ten countries are as follows:", "sourceParagraphs": [ 68, 69 ] }, { "type": "table", "rows": [ [ "Country", "Eligible user population" ], [ "A", "400,000,000" ], [ "B", "250,000,000" ], [ "C", "150,000,000" ], [ "D", "100,000,000" ], [ "E", "60,000,000" ], [ "F", "30,000,000" ], [ "G", "8,000,000" ], [ "H", "1,500,000" ], [ "I", "400,000" ], [ "J", "100,000" ], [ "Total", "1,000,000,000" ] ] }, { "type": "paragraph", "text": "Raw quotas according to pure population proportion:", "sourceParagraphs": [ 70 ] }, { "type": "table", "rows": [ [ "Country", "Raw quota" ], [ "A", "400.0" ], [ "B", "250.0" ], [ "C", "150.0" ], [ "D", "100.0" ], [ "E", "60.0" ], [ "F", "30.0" ], [ "G", "8.0" ], [ "H", "1.5" ], [ "I", "0.4" ], [ "J", "0.1" ] ] }, { "type": "paragraph", "text": "Half of a person cannot be selected. 0.4 people cannot be sent to a country. If you round the raw quotas to the nearest whole number, the total may exceed or fall short of 1,000. If you give at least one person to each country, you over-represent countries I and J compared to their population shares.", "sourceParagraphs": [ 71, 72, 73, 74 ] }, { "type": "paragraph", "text": "Giving one person to country I increases its theoretical share 2.5 times. Giving one person to country J increases its theoretical share 10 times. This extreme representation can largely be corrected. But you cannot produce a country score with a single person. What happens if you give no one to I and J?", "sourceParagraphs": [ 75, 76, 77, 78, 79 ] }, { "type": "paragraph", "text": "If you define these countries as separate and mandatory strata, the probability of some people in the target population being selected becomes zero. In this case, the claim of a “probabilistic sample representing the entire appropriate global population” weakens. Now consider a more dangerous decision. First, you collect the answers.", "sourceParagraphs": [ 80, 81, 82, 83, 84, 85 ] }, { "type": "paragraph", "text": "The only user in country J gets an incorrect answer. Then you say: “The sample size of this country is very small, let's exclude it from the scope.” In country A, there are many positive answers. You retain it with full weight. This is no longer rounding. It is a result-oriented sample change.", "sourceParagraphs": [ 86, 87, 88, 89, 90, 91, 92, 93 ] }, { "type": "paragraph", "text": "In another case, you increase the invalid record rate in a language group with low results, using backup participants only until a positive response is obtained. This is also not a sample. It is a response selection mechanism. The first ruling of this section is:", "sourceParagraphs": [ 94, 95, 96, 97 ] }, { "type": "callout", "text": "The sample must be established before seeing AI responses.", "sourceParagraph": 98 }, { "type": "paragraph", "text": "Its second provision states:", "sourceParagraphs": [ 99 ] }, { "type": "callout", "text": "The probability of a participant being selected cannot depend on the correctness of their answer.", "sourceParagraph": 100 }, { "type": "paragraph", "text": "Its third provision states:", "sourceParagraphs": [ 101 ] }, { "type": "callout", "text": "The visibility of the small country and the population weight in the global score must be managed separately.", "sourceParagraph": 102 }, { "type": "paragraph", "text": "Its fourth provision states:", "sourceParagraphs": [ 103 ] }, { "type": "callout", "text": "Oversampling is possible; over-weighting is not mandatory.", "sourceParagraph": 104 }, { "type": "paragraph", "text": "Its fifth provision states:", "sourceParagraphs": [ 105 ] }, { "type": "callout", "text": "1,000 raw records do not always carry statistical information equivalent to 1,000 observations.", "sourceParagraph": 106 }, { "type": "heading2", "text": "1. PURPOSE OF THE CHAPTER", "sourceParagraph": 108 }, { "type": "paragraph", "text": "The purpose of this section is to transform the targeted 1,000 valid observations for each AI product and measurement wave into a verifiable sample design based on the product-eligible user population. The section standardises the following distinctions:", "sourceParagraphs": [ 109, 110 ] }, { "type": "bullet", "text": "Target population versus selected sample", "sourceParagraph": 111 }, { "type": "bullet", "text": "Raw quota versus integer allocation", "sourceParagraph": 112 }, { "type": "bullet", "text": "Main population panel versus complementary panels", "sourceParagraph": 113 }, { "type": "bullet", "text": "Country visibility versus country weight", "sourceParagraph": 114 }, { "type": "bullet", "text": "Strata versus experimental cell", "sourceParagraph": 115 }, { "type": "bullet", "text": "Large country strata versus small country pool", "sourceParagraph": 116 }, { "type": "bullet", "text": "Proportional allocation vs. disproportionate oversampling", "sourceParagraph": 117 }, { "type": "bullet", "text": "Sample share vs. population share", "sourceParagraph": 118 }, { "type": "bullet", "text": "Number of invitations vs. valid observation target", "sourceParagraph": 119 }, { "type": "bullet", "text": "Backup participant vs. post-result replacement", "sourceParagraph": 120 }, { "type": "bullet", "text": "Data collection invalidity vs. adverse AI output", "sourceParagraph": 121 }, { "type": "bullet", "text": "Design weight vs. outcome weight", "sourceParagraph": 122 }, { "type": "bullet", "text": "Raw sample size vs. effective sample size", "sourceParagraph": 123 }, { "type": "bullet", "text": "Country sample vs. language sample", "sourceParagraph": 124 }, { "type": "bullet", "text": "Multilingual user vs. more than one person", "sourceParagraph": 125 }, { "type": "bullet", "text": "Single-stage selection vs. multi-stage selection", "sourceParagraph": 126 }, { "type": "bullet", "text": "Native Reach allocation vs. Common Support allocation", "sourceParagraph": 127 }, { "type": "bullet", "text": "Instant wave allocation and time series allocation", "sourceParagraph": 128 }, { "type": "bullet", "text": "Population estimation and country or language recognition analysis", "sourceParagraph": 129 }, { "type": "bullet", "text": "Main 1,000 observations with auxiliary observers and fairness panels", "sourceParagraph": 130 }, { "type": "paragraph", "text": "At the end of this section, each audit should be able to answer these questions:", "sourceParagraphs": [ 131 ] }, { "type": "callout", "text": "Why were 1,000 observations distributed in these numbers across this country, language, and user groups?", "sourceParagraph": 132 }, { "type": "callout", "text": "How was the weight of each observation in the global result determined?", "sourceParagraph": 133 }, { "type": "callout", "text": "Who had a zero probability of being selected and why?", "sourceParagraph": 134 }, { "type": "callout", "text": "How many people were invited, how many completed, and how many observations remained valid?", "sourceParagraph": 135 }, { "type": "callout", "text": "Were the sample results frozen before being viewed?", "sourceParagraph": 136 }, { "type": "heading2", "text": "2. CENTRAL NORMATIVE PROVISION", "sourceParagraph": 138 }, { "type": "paragraph", "text": "The GEO-1000 main population sample should be based on a stratified sampling design that assigns known or predictable probabilities of selection to people in the pre-frozen product-eligible population frame for each AI product and measurement wave. The default target of the main population panel is:", "sourceParagraphs": [ 139, 140 ] }, { "type": "callout", "text": "At least 1,000 valid observations per AI product and main measurement wave", "sourceParagraph": 141 }, { "type": "paragraph", "text": "is maintained. This number:", "sourceParagraphs": [ 142, 143 ] }, { "type": "bullet", "text": "is not the number of people invited,", "sourceParagraph": 144 }, { "type": "bullet", "text": "is not the number of accounts,", "sourceParagraph": 145 }, { "type": "bullet", "text": "is not the number of raw screenshots,", "sourceParagraph": 146 }, { "type": "paragraph", "text": "It is not the sum of all supplementary panels. 1,000 is the target for valid product-observations meeting the protocol in the main population panel. Country Observer, Language Fairness, and Accessibility panels can create additional samples. These additional samples:", "sourceParagraphs": [ 147, 148, 149, 150 ] }, { "type": "bullet", "text": "can increase the total field volume above 1,000,", "sourceParagraph": 151 }, { "type": "bullet", "text": "can be used with appropriate weights in the global population outcome,", "sourceParagraph": 152 }, { "type": "paragraph", "text": "but cannot be used to covertly reduce the 1,000 valid observation target of the main panel.", "sourceParagraphs": [ 153 ] }, { "type": "heading2", "text": "3. BASIC REPRESENTATION OF THE SAMPLE", "sourceParagraph": 155 }, { "type": "paragraph", "text": "For a specific AI product a, measurement wave w, and target population frame, the strata:", "sourceParagraphs": [ 156 ] }, { "type": "formula", "text": "h ∈ H_{a,w}", "sourceParagraph": 157 }, { "type": "paragraph", "text": "are denoted. For each stratum:", "sourceParagraphs": [ 158, 159 ] }, { "type": "bullet", "text": "Nh: estimated product-appropriate population in the stratum", "sourceParagraph": 160 }, { "type": "bullet", "text": "N: total target population", "sourceParagraph": 161 }, { "type": "bullet", "text": "Wh: population weight of the stratum", "sourceParagraph": 162 }, { "type": "bullet", "text": "n: target valid main panel observation", "sourceParagraph": 163 }, { "type": "bullet", "text": "qh: crude proportional quota", "sourceParagraph": 164 }, { "type": "bullet", "text": "nh: allocated integer valid observation target", "sourceParagraph": 165 }, { "type": "paragraph", "text": "is. Total target population:", "sourceParagraphs": [ 166, 167 ] }, { "type": "formula", "text": "N = Σ_{h=1}^H N_h", "sourceParagraph": 168 }, { "type": "paragraph", "text": "Population weight of the stratum:", "sourceParagraphs": [ 169 ] }, { "type": "formula", "text": "W_h = N_h / N", "sourceParagraph": 170 }, { "type": "paragraph", "text": "Crude proportional quota:", "sourceParagraphs": [ 171 ] }, { "type": "formula", "text": "q_h = nW_h", "sourceParagraph": 172 }, { "type": "paragraph", "text": "is. Main proportional allocation, under ideal conditions:", "sourceParagraphs": [ 173, 174 ] }, { "type": "formula", "text": "n_h ≈ nW_h", "sourceParagraph": 175 }, { "type": "paragraph", "text": "should be. However:", "sourceParagraphs": [ 176, 177 ] }, { "type": "bullet", "text": "qh may be fractional,", "sourceParagraph": 178 }, { "type": "bullet", "text": "some cells may fall below one person,", "sourceParagraph": 179 }, { "type": "bullet", "text": "a minimum diagnostic sample may be required for language and country subgroups,", "sourceParagraph": 180 }, { "type": "bullet", "text": "high-risk small groups may also be examined,", "sourceParagraph": 181 }, { "type": "paragraph", "text": "field completion rates may vary. Therefore, the raw quota is not directly the final allocation.", "sourceParagraphs": [ 182, 183 ] }, { "type": "heading2", "text": "4. THREE DISTINCT PURPOSES OF SAMPLE DESIGN", "sourceParagraph": 185 }, { "type": "paragraph", "text": "The same sample may not fulfil all purposes with the same strength. Three purposes should be separated.", "sourceParagraphs": [ 186, 187 ] }, { "type": "heading3", "text": "4.1. Global Population Estimate", "sourceParagraph": 189 }, { "type": "paragraph", "text": "Answers the question:", "sourceParagraphs": [ 190 ] }, { "type": "callout", "text": "What representative outcome would the average person in the global user population suitable for the product see?", "sourceParagraph": 191 }, { "type": "paragraph", "text": "Population weight is the primary priority.", "sourceParagraphs": [ 192 ] }, { "type": "heading3", "text": "4.2. Subgroup Diagnosis", "sourceParagraph": 194 }, { "type": "paragraph", "text": "Answers the question:", "sourceParagraphs": [ 195 ] }, { "type": "callout", "text": "What representation distortions occur in specific countries, languages, plans, or user groups?", "sourceParagraph": 196 }, { "type": "paragraph", "text": "It may be necessary to oversample small groups.", "sourceParagraphs": [ 197 ] }, { "type": "heading3", "text": "4.3. Geographic and Linguistic Coverage", "sourceParagraph": 199 }, { "type": "paragraph", "text": "Answers the question:", "sourceParagraphs": [ 200 ] }, { "type": "callout", "text": "In which countries and languages is there at least real user observation?", "sourceParagraph": 201 }, { "type": "paragraph", "text": "Country Observer and Language Fairness panels become important. These three goals cannot be inexplicably combined in a single sample.", "sourceParagraphs": [ 202, 203 ] }, { "type": "heading2", "text": "5. MAIN SAMPLE ARCHITECTURE", "sourceParagraph": 205 }, { "type": "paragraph", "text": "The candidate main sample architecture for GEO-1000 consists of four layers.", "sourceParagraphs": [ 206 ] }, { "type": "heading3", "text": "5.1. Population Panel", "sourceParagraph": 208 }, { "type": "paragraph", "text": "Every AI product and wave has a minimum of 1,000 valid observations as the main panel. It produces a global population estimate.", "sourceParagraphs": [ 209, 210 ] }, { "type": "heading3", "text": "5.2. Country Observer Overlay", "sourceParagraph": 212 }, { "type": "paragraph", "text": "Provides additional geographic observations for small countries that cannot be directly represented in the main population panel. These observations can indicate:", "sourceParagraphs": [ 213, 214 ] }, { "type": "bullet", "text": "single event,", "sourceParagraph": 215 }, { "type": "bullet", "text": "access status,", "sourceParagraph": 216 }, { "type": "bullet", "text": "future research needs", "sourceParagraph": 217 }, { "type": "paragraph", "text": "They do not generate a country score on their own.", "sourceParagraphs": [ 218, 219 ] }, { "type": "heading3", "text": "5.3. Language Fairness Oversample", "sourceParagraph": 221 }, { "type": "paragraph", "text": "Generates additional observations for low-resource languages or languages underrepresented in the main panel. They are reweighted to actual population weights in the global outcome.", "sourceParagraphs": [ 222, 223 ] }, { "type": "heading3", "text": "5.4. Accessibility Oversample", "sourceParagraph": 225 }, { "type": "paragraph", "text": "It also measures users of assistive technology and accessibility, who are not adequately represented by the standard panel. The target universe and weight of these observations are recorded separately.", "sourceParagraphs": [ 226, 227 ] }, { "type": "heading2", "text": "6. WHAT IS A STRATUM?", "sourceParagraph": 229 }, { "type": "paragraph", "text": "A stratum is a segment of the population that is managed separately in terms of sample allocation and selection probability. The stratum can be a combination of one or more of the following areas:", "sourceParagraphs": [ 230, 231 ] }, { "type": "bullet", "text": "Country", "sourceParagraph": 232 }, { "type": "bullet", "text": "Region", "sourceParagraph": 233 }, { "type": "bullet", "text": "Language", "sourceParagraph": 234 }, { "type": "bullet", "text": "Locale", "sourceParagraph": 235 }, { "type": "bullet", "text": "AI plan", "sourceParagraph": 236 }, { "type": "bullet", "text": "Web or mobile", "sourceParagraph": 237 }, { "type": "bullet", "text": "Controlled or natural user panel", "sourceParagraph": 238 }, { "type": "bullet", "text": "Accessibility status", "sourceParagraph": 239 }, { "type": "bullet", "text": "User type", "sourceParagraph": 240 }, { "type": "bullet", "text": "Jurisdiction", "sourceParagraph": 241 }, { "type": "paragraph", "text": "If each small combination is made a separate stratum, the sample is fragmented. For example: Country × language × plan × mobile/web × age group × user type combination can produce hundreds or thousands of cells. 1,000 observations do not provide reliable estimates for most of these cells. Therefore, stratification should be:", "sourceParagraphs": [ 242, 243, 244, 245, 246, 247 ] }, { "type": "bullet", "text": "substantial,", "sourceParagraph": 248 }, { "type": "bullet", "text": "measurable,", "sourceParagraph": 249 }, { "type": "bullet", "text": "appropriate for sample size", "sourceParagraph": 250 }, { "type": "paragraph", "text": "must be.", "sourceParagraphs": [ 251 ] }, { "type": "heading2", "text": "7. A STRATUM IS NOT THE SAME AS A REPORTING CELL", "sourceParagraph": 253 }, { "type": "paragraph", "text": "It can be shown separately on a group result card. However, it does not have to be a separate stratum in sample design. For example:", "sourceParagraphs": [ 254, 255, 256 ] }, { "type": "bullet", "text": "a small language,", "sourceParagraph": 257 }, { "type": "bullet", "text": "a specific accessibility group,", "sourceParagraph": 258 }, { "type": "bullet", "text": "a certain age band", "sourceParagraph": 259 }, { "type": "paragraph", "text": "It can be selected within a broader stratum in the main sample. Later, it can be reported separately in the analysis. The reverse is also possible: Two regions may be separately stratified for sampling efficiency. However, in the public report, they can be combined under a single regional result. The stratification design and the reporting structure should carry separate records.", "sourceParagraphs": [ 260, 261, 262, 263, 264, 265 ] }, { "type": "heading2", "text": "8. SEQUENCE OF STRATUM ESTABLISHMENT", "sourceParagraph": 267 }, { "type": "paragraph", "text": "The candidate sequence for the Main Population Panel is as follows:", "sourceParagraphs": [ 268 ] }, { "type": "bullet", "text": "AI product and user surface", "sourceParagraph": 269 }, { "type": "bullet", "text": "Native Reach or Common Support population frame", "sourceParagraph": 270 }, { "type": "bullet", "text": "Country or small country pool", "sourceParagraph": 271 }, { "type": "bullet", "text": "Main audit language or language group", "sourceParagraph": 272 }, { "type": "bullet", "text": "Material plan/interface separation, if necessary", "sourceParagraph": 273 }, { "type": "bullet", "text": "Participant selection frame", "sourceParagraph": 274 }, { "type": "paragraph", "text": "The plan and interface do not have to be made into separate layers for each country. These can be managed through:", "sourceParagraphs": [ 275, 276 ] }, { "type": "bullet", "text": "sub-layer,", "sourceParagraph": 277 }, { "type": "bullet", "text": "quota,", "sourceParagraph": 278 }, { "type": "bullet", "text": "random assignment", "sourceParagraph": 279 }, { "type": "paragraph", "text": "methods.", "sourceParagraphs": [ 280 ] }, { "type": "heading2", "text": "9. LARGE AND SMALL COUNTRY LAYERS", "sourceParagraph": 282 }, { "type": "paragraph", "text": "A country's raw quota:", "sourceParagraphs": [ 283 ] }, { "type": "formula", "text": "q_c = nN_c / N", "sourceParagraph": 284 }, { "type": "paragraph", "text": "is in this form. If the raw quota is large enough, the country can be kept directly as a separate layer. This can be called:", "sourceParagraphs": [ 285, 286, 287 ] }, { "type": "heading4", "text": "self-representing country layer", "sourceParagraph": 288 }, { "type": "paragraph", "text": "Making countries with very small raw quotas individually into separate mandatory layers may reduce sampling efficiency. These countries can be appropriately:", "sourceParagraphs": [ 289, 290, 291 ] }, { "type": "bullet", "text": "regional,", "sourceParagraph": 292 }, { "type": "bullet", "text": "language family,", "sourceParagraph": 293 }, { "type": "bullet", "text": "product access regime,", "sourceParagraph": 294 }, { "type": "bullet", "text": "geographical proximity", "sourceParagraph": 295 }, { "type": "paragraph", "text": "can be included in small country pools.", "sourceParagraphs": [ 296 ] }, { "type": "heading2", "text": "10. SMALL COUNTRY POOL", "sourceParagraph": 298 }, { "type": "paragraph", "text": "A small country pool:", "sourceParagraphs": [ 299 ] }, { "type": "formula", "text": "Let it be represented as B_r = {c_1, c_2, …, c_k}", "sourceParagraph": 300 }, { "type": "paragraph", "text": "The total eligible population of the pool:", "sourceParagraphs": [ 301, 302 ] }, { "type": "formula", "text": "N_{B_r} = Σ_{c∈B_r} N_c", "sourceParagraph": 303 }, { "type": "paragraph", "text": "Main sample quota:", "sourceParagraphs": [ 304 ] }, { "type": "formula", "text": "q_{B_r} = nN_{B_r} / N", "sourceParagraph": 305 }, { "type": "paragraph", "text": "is calculated as. Within the pool, country selection can be done with probability proportional to the eligible population size. The probability of country c being selected within the pool:", "sourceParagraphs": [ 306, 307, 308 ] }, { "type": "formula", "text": "Pr(c | B_r) = N_c / N_{B_r}", "sourceParagraph": 309 }, { "type": "paragraph", "text": "It may be. Then a participant is selected within the chosen country. This approach:", "sourceParagraphs": [ 310, 311, 312 ] }, { "type": "bullet", "text": "can prevent small countries from having a zero chance of being selected in the target population,", "sourceParagraph": 313 }, { "type": "bullet", "text": "does not guarantee that each country will be selected in a given wave,", "sourceParagraph": 314 }, { "type": "paragraph", "text": "preserves the global population estimate. Not selecting a country in that wave:", "sourceParagraphs": [ 315, 316 ] }, { "type": "heading3", "text": "NOT SELECTED IN POPULATION SAMPLE", "sourceParagraph": 317 }, { "type": "paragraph", "text": "does not mean: \"The country is not part of the target population.\"", "sourceParagraphs": [ 318, 319, 320 ] }, { "type": "heading2", "text": "11. LIMITATIONS OF POOLING SMALL COUNTRIES", "sourceParagraph": 322 }, { "type": "paragraph", "text": "Small countries cannot be randomly placed in the same pool just because their populations are small. Countries within the pool are similar to the extent that they are relevant:", "sourceParagraphs": [ 323, 324 ] }, { "type": "bullet", "text": "product access,", "sourceParagraph": 325 }, { "type": "bullet", "text": "user surface,", "sourceParagraph": 326 }, { "type": "bullet", "text": "language or locale structure,", "sourceParagraph": 327 }, { "type": "bullet", "text": "legal access,", "sourceParagraph": 328 }, { "type": "bullet", "text": "geographical region,", "sourceParagraph": 329 }, { "type": "bullet", "text": "data collection method", "sourceParagraph": 330 }, { "type": "paragraph", "text": "must be carried. A country without product access cannot be placed in the same main pool as a country with full access. Different usage conditions cannot be made invisible.", "sourceParagraphs": [ 331, 332, 333 ] }, { "type": "heading2", "text": "12. SMALL COUNTRY POOL DOES NOT PRODUCE COUNTRY SCORE", "sourceParagraph": 335 }, { "type": "paragraph", "text": "One or two observations selected from the pool: can be used in the global population contribution of the pool, may indicate a single country event. However, it does not produce a separate statistical score for a small country. If a country-based score is wanted, additionally:", "sourceParagraphs": [ 336, 337, 338, 339, 340 ] }, { "type": "bullet", "text": "Country Observer,", "sourceParagraph": 341 }, { "type": "bullet", "text": "country-specific sample,", "sourceParagraph": 342 }, { "type": "bullet", "text": "multi-wave cumulative sample", "sourceParagraph": 343 }, { "type": "paragraph", "text": "may be required.", "sourceParagraphs": [ 344 ] }, { "type": "heading2", "text": "13. ALLOCATION PROPORTIONAL TO PURE POPULATION", "sourceParagraph": 346 }, { "type": "paragraph", "text": "Pure proportional allocation:\n\nis based on the formula:", "sourceParagraphs": [ 347 ] }, { "type": "formula", "text": "q_h = nW_h", "sourceParagraph": 348 }, { "type": "paragraph", "text": "Advantages:", "sourceParagraphs": [ 349, 350 ] }, { "type": "bullet", "text": "it is understandable for the main population estimate,", "sourceParagraph": 351 }, { "type": "bullet", "text": "it naturally gives weight to large populations,", "sourceParagraph": 352 }, { "type": "bullet", "text": "it can reduce weight differences,", "sourceParagraph": 353 }, { "type": "paragraph", "text": "it can approach a self-weighting design. Limitations:", "sourceParagraphs": [ 354, 355 ] }, { "type": "bullet", "text": "it can sample very few small countries and languages,", "sourceParagraph": 356 }, { "type": "bullet", "text": "it may be insufficient to produce subgroup scores,", "sourceParagraph": 357 }, { "type": "bullet", "text": "it may leave high-risk small groups invisible,", "sourceParagraph": 358 }, { "type": "paragraph", "text": "it can create fractional and near-zero quotas. Therefore, pure proportional allocation can be a strong default for the main Population Panel. It does not replace complementary panels.", "sourceParagraphs": [ 359, 360, 361 ] }, { "type": "heading2", "text": "14. CONVERTING FRACTIONS TO WHOLE NUMBERS", "sourceParagraph": 363 }, { "type": "paragraph", "text": "Raw quotas are often fractional. The total allocation should be exactly n. The Largest Remainder Method can be used as the default method.", "sourceParagraphs": [ 364, 365, 366 ] }, { "type": "heading3", "text": "14.1. First Step", "sourceParagraph": 368 }, { "type": "paragraph", "text": "Each raw quota is rounded down:", "sourceParagraphs": [ 369 ] }, { "type": "formula", "text": "b_h = ⌊q_h⌋", "sourceParagraph": 370 }, { "type": "heading3", "text": "14.2. Second Step", "sourceParagraph": 372 }, { "type": "paragraph", "text": "The number of unallocated observations:", "sourceParagraphs": [ 373 ] }, { "type": "formula", "text": "R = n − Σ_h b_h", "sourceParagraph": 374 }, { "type": "paragraph", "text": "is calculated.", "sourceParagraphs": [ 375 ] }, { "type": "heading3", "text": "14.3. Third Step", "sourceParagraph": 377 }, { "type": "paragraph", "text": "The fractional part of each stratum:", "sourceParagraphs": [ 378 ] }, { "type": "formula", "text": "f_h = q_h − b_h", "sourceParagraph": 379 }, { "type": "paragraph", "text": "where f_h is the fractional part of q_h.", "sourceParagraphs": [ 380 ] }, { "type": "heading3", "text": "14.4. Fourth Step", "sourceParagraph": 382 }, { "type": "paragraph", "text": "The remaining R observations are added one by one to the layers with the highest fh values.", "sourceParagraphs": [ 383 ] }, { "type": "heading3", "text": "14.5. Equality Situation", "sourceParagraph": 385 }, { "type": "paragraph", "text": "If the fractional parts of the two layers are equal:", "sourceParagraphs": [ 386 ] }, { "type": "bullet", "text": "population size,", "sourceParagraph": 387 }, { "type": "bullet", "text": "predetermined fixed order,", "sourceParagraph": 388 }, { "type": "bullet", "text": "registered random seed", "sourceParagraph": 389 }, { "type": "paragraph", "text": "An independent tie-breaking rule should be used, regardless of the outcome. A country cannot be chosen based on the results.", "sourceParagraphs": [ 390, 391 ] }, { "type": "heading2", "text": "15. SYNTHETIC INTEGER ALLOCATION", "sourceParagraph": 393 }, { "type": "heading3", "text": "SYNTHETIC DISPLAY", "sourceParagraph": 394 }, { "type": "paragraph", "text": "Appropriate populations at the beginning of the section:", "sourceParagraphs": [ 395 ] }, { "type": "table", "rows": [ [ "Country", "Suitable population", "Raw quota" ], [ "A", "400,000,000", "400.0" ], [ "B", "250,000,000", "250.0" ], [ "C", "150,000,000", "150.0" ], [ "D", "100,000,000", "100.0" ], [ "E", "60,000,000", "60.0" ], [ "F", "30,000,000", "30.0" ], [ "G", "8,000,000", "8.0" ], [ "H", "1,500,000", "1.5" ], [ "I", "400,000", "0.4" ], [ "J", "100,000", "0.1" ] ] }, { "type": "paragraph", "text": "If countries are protected as separate layers, round down:", "sourceParagraphs": [ 396 ] }, { "type": "table", "rows": [ [ "Country", "Rounded down quota" ], [ "A", "400" ], [ "B", "250" ], [ "C", "150" ], [ "D", "100" ], [ "E", "60" ], [ "F", "30" ], [ "G", "8" ], [ "H", "1" ], [ "I", "0" ], [ "J", "0" ], [ "Total", "999" ] ] }, { "type": "paragraph", "text": "One observation remains. It is added to H because the highest fraction is 0.5 in country H. Final allocation:", "sourceParagraphs": [ 397, 398, 399 ] }, { "type": "table", "rows": [ [ "Country", "Final quota" ], [ "A", "400" ], [ "B", "250" ], [ "C", "150" ], [ "D", "100" ], [ "E", "60" ], [ "F", "30" ], [ "G", "8" ], [ "H", "2" ], [ "I", "0" ], [ "J", "0" ], [ "Total", "1,000" ] ] }, { "type": "paragraph", "text": "Even though this design defines I and J as separate deterministic strata, if it does not give a chance to be selected, it may pose a problem in terms of full target universe inference. A more accurate solution is a small country pool.", "sourceParagraphs": [ 400, 401 ] }, { "type": "heading2", "text": "16. SYNTHETIC ALLOCATION WITH SMALL COUNTRY POOL", "sourceParagraph": 403 }, { "type": "paragraph", "text": "Countries H, I, and J:", "sourceParagraphs": [ 404 ] }, { "type": "formula", "text": "B_small = {H, I, J}", "sourceParagraph": 405 }, { "type": "paragraph", "text": "should be combined in the pool. Pool population:", "sourceParagraphs": [ 406, 407 ] }, { "type": "formula", "text": "1,500,000+400,000+100,000=2,000,000", "sourceParagraph": 408 }, { "type": "paragraph", "text": "Pool quota:", "sourceParagraphs": [ 409 ] }, { "type": "formula", "text": "1,000 × 2,000,000 / 1,000,000,000 = 2", "sourceParagraph": 410 }, { "type": "paragraph", "text": "becomes. Main allocation:", "sourceParagraphs": [ 411, 412 ] }, { "type": "table", "rows": [ [ "Stratum", "Quota" ], [ "A", "400" ], [ "B", "250" ], [ "C", "150" ], [ "D", "100" ], [ "E", "60" ], [ "F", "30" ], [ "G", "8" ], [ "Small country pool", "2" ], [ "Total", "1,000" ] ] }, { "type": "paragraph", "text": "Selection weights of countries within the pool:", "sourceParagraphs": [ 413 ] }, { "type": "table", "rows": [ [ "Country", "Population share within the pool" ], [ "H", "75%" ], [ "I", "20%" ], [ "J", "5%" ] ] }, { "type": "paragraph", "text": "Two main population observations are selected according to these probabilities. In the same wave:", "sourceParagraphs": [ 414, 415 ] }, { "type": "bullet", "text": "both observations can come from H,", "sourceParagraph": 416 }, { "type": "bullet", "text": "one from H and one from I,", "sourceParagraph": 417 }, { "type": "bullet", "text": "rarely from J", "sourceParagraph": 418 }, { "type": "paragraph", "text": "This can occur. The design does not make J visible in every wave, but it does not reduce J's probability of selection to zero. If J must be observed, it is added separately to the Country Observer Panel. A Country Observer observation does not give J a weight of one in 1,000 in the main global allocation; J retains its true population weight in the main panel.", "sourceParagraphs": [ 419, 420, 421, 422, 423, 424 ] }, { "type": "heading2", "text": "17. MINIMUM CELL ASSIGNMENT", "sourceParagraph": 426 }, { "type": "paragraph", "text": "Purely proportional quota may not be sufficient to perform diagnostic analysis for some subgroups. For example, the raw quota for a language cell might be 8. A reliable language score cannot be produced with eight observations. In this case, there are three options:", "sourceParagraphs": [ 427, 428, 429, 430 ] }, { "type": "bullet", "text": "Not to publish the language score", "sourceParagraph": 431 }, { "type": "bullet", "text": "Combine the cell with similar groups", "sourceParagraph": 432 }, { "type": "bullet", "text": "Oversampling the cell", "sourceParagraph": 433 }, { "type": "paragraph", "text": "Minimum cell allocation must be defined before the results.", "sourceParagraphs": [ 434 ] }, { "type": "heading2", "text": "18. CANDIDATE REPORTING BANDS", "sourceParagraph": 436 }, { "type": "heading3", "text": "CANDIDATE REPORTING BANDS — CANDIDATE REPORTING BANDS", "sourceParagraph": 437 }, { "type": "paragraph", "text": "The thresholds below are not final scientific decisions. They should be calibrated through pilot and external review.", "sourceParagraphs": [ 438, 439 ] }, { "type": "heading4", "text": "1–29 Valid Observation", "sourceParagraph": 440 }, { "type": "paragraph", "text": "Single event and discovery Country or language rate is not published", "sourceParagraphs": [ 441, 442 ] }, { "type": "heading3", "text": "OBSERVATIONAL ONLY", "sourceParagraph": 443 }, { "type": "heading4", "text": "30–99 Valid Observations", "sourceParagraph": 444 }, { "type": "paragraph", "text": "Exploratory level estimate Broad uncertainty No definitive conformity judgement is given", "sourceParagraphs": [ 445, 446, 447 ] }, { "type": "heading3", "text": "EXPLORATORY", "sourceParagraph": 448 }, { "type": "heading4", "text": "100–199 Valid Observations", "sourceParagraph": 449 }, { "type": "paragraph", "text": "First subgroup estimate Uncertainty should be visible", "sourceParagraphs": [ 450, 451 ] }, { "type": "heading3", "text": "PROVISIONAL", "sourceParagraph": 452 }, { "type": "heading4", "text": "200 and Above Valid Observations", "sourceParagraph": 453 }, { "type": "paragraph", "text": "Stronger subgroup estimate candidate Design effect and weights are again evaluated", "sourceParagraphs": [ 454, 455 ] }, { "type": "heading3", "text": "STANDARD ESTIMATION CANDIDATE", "sourceParagraph": 456 }, { "type": "paragraph", "text": "These bands:", "sourceParagraphs": [ 457 ] }, { "type": "bullet", "text": "error rate,", "sourceParagraph": 458 }, { "type": "bullet", "text": "expected precision,", "sourceParagraph": 459 }, { "type": "bullet", "text": "weight variability,", "sourceParagraph": 460 }, { "type": "bullet", "text": "design effect", "sourceParagraph": 461 }, { "type": "paragraph", "text": "cannot be automatically considered sufficient without being taken into account.", "sourceParagraphs": [ 462 ] }, { "type": "heading2", "text": "19. OVERSAMPLING", "sourceParagraph": 464 }, { "type": "paragraph", "text": "A country, language, or user group may have more observations than its population share. This is called oversampling. Its purposes:", "sourceParagraphs": [ 465, 466, 467 ] }, { "type": "bullet", "text": "to increase subgroup sensitivity,", "sourceParagraph": 468 }, { "type": "bullet", "text": "to make a low-resource language visible,", "sourceParagraph": 469 }, { "type": "bullet", "text": "to examine a high-risk small group,", "sourceParagraph": 470 }, { "type": "bullet", "text": "to expand Country Observer coverage,", "sourceParagraph": 471 }, { "type": "bullet", "text": "to adequately represent accessibility users", "sourceParagraph": 472 }, { "type": "paragraph", "text": "are possible. Oversampling is not wrong. What is wrong:", "sourceParagraphs": [ 473, 474, 475 ] }, { "type": "callout", "text": "It is to use the sample proportion of the over-sampled group as if it were the true population proportion.", "sourceParagraph": 476 }, { "type": "heading2", "text": "20. POPULATION SHARE AND SAMPLE SHARE", "sourceParagraph": 478 }, { "type": "paragraph", "text": "Population share for layer h:", "sourceParagraphs": [ 479 ] }, { "type": "formula", "text": "W_h = N_h / N", "sourceParagraph": 480 }, { "type": "paragraph", "text": "Sample share:", "sourceParagraphs": [ 481 ] }, { "type": "formula", "text": "S_h = n_h / n", "sourceParagraph": 482 }, { "type": "paragraph", "text": "is in the form. In pure proportional allocation:", "sourceParagraphs": [ 483, 484 ] }, { "type": "formula", "text": "S_h ≈ W_h", "sourceParagraph": 485 }, { "type": "paragraph", "text": "okay. In excessive sampling:", "sourceParagraphs": [ 486, 487 ] }, { "type": "formula", "text": "S_h > W_h", "sourceParagraph": 488 }, { "type": "paragraph", "text": "It is possible. In a small sample:", "sourceParagraphs": [ 489, 490 ] }, { "type": "formula", "text": "S_h < W_h", "sourceParagraph": 491 }, { "type": "paragraph", "text": "It is possible. The global population estimate should retain Wh values instead of Sh.", "sourceParagraphs": [ 492, 493 ] }, { "type": "heading2", "text": "21. BASIC DESIGN LOAD", "sourceParagraph": 495 }, { "type": "paragraph", "text": "In simple stratified sampling, the basic weight of each observation within stratum h can be considered as:", "sourceParagraphs": [ 496 ] }, { "type": "formula", "text": "d_i = N_h / n_h", "sourceParagraph": 497 }, { "type": "paragraph", "text": "The normalised stratum weight can be written as:", "sourceParagraphs": [ 498, 499 ] }, { "type": "formula", "text": "a_h = W_h / S_h", "sourceParagraph": 500 }, { "type": "paragraph", "text": "The normalised observation-level weight can be:", "sourceParagraphs": [ 501, 502 ] }, { "type": "formula", "text": "w_i = W_h / n_h", "sourceParagraph": 503 }, { "type": "paragraph", "text": "It is ensured that the sum equals one:", "sourceParagraphs": [ 504, 505 ] }, { "type": "formula", "text": "Σ_i w_i = 1", "sourceParagraph": 506 }, { "type": "paragraph", "text": "The weighted binary transition estimate can be calculated as:", "sourceParagraphs": [ 507, 508 ] }, { "type": "formula", "text": "θ̂ = Σ_i w_iY_i", "sourceParagraph": 509 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 510, 511 ] }, { "type": "formula", "text": "Yi=1: response meeting the transition criterion", "sourceParagraph": 512 }, { "type": "formula", "text": "Yi=0: response not meeting it", "sourceParagraph": 513 }, { "type": "paragraph", "text": "It is possible. The final NOMOS system does not have to use only binary outcomes. The same weighting logic can also be applied to multidimensional response scores.", "sourceParagraphs": [ 514, 515, 516 ] }, { "type": "heading2", "text": "22. EXCESS SAMPLING EXAMPLE", "sourceParagraph": 518 }, { "type": "heading3", "text": "SYNTHETIC EXAMPLE", "sourceParagraph": 519 }, { "type": "paragraph", "text": "In the suitable population of country B:", "sourceParagraphs": [ 520 ] }, { "type": "bullet", "text": "Language X: 80 per cent", "sourceParagraph": 521 }, { "type": "bullet", "text": "Language Y: 20 per cent", "sourceParagraph": 522 }, { "type": "paragraph", "text": "Let it be. In the Population Panel, there are 250 observations for country B. Pure allocation:", "sourceParagraphs": [ 523, 524, 525 ] }, { "type": "bullet", "text": "Language X: 200", "sourceParagraph": 526 }, { "type": "bullet", "text": "Language Y: 50", "sourceParagraph": 527 }, { "type": "paragraph", "text": "Let 50 additional Language Fairness observations be taken for Language Y. Collected observations:", "sourceParagraphs": [ 528, 529, 530 ] }, { "type": "bullet", "text": "Language X: 200", "sourceParagraph": 531 }, { "type": "bullet", "text": "Language Y: 100", "sourceParagraph": 532 }, { "type": "paragraph", "text": "The share of Language Y in the raw sample:", "sourceParagraphs": [ 533, 534 ] }, { "type": "formula", "text": "100/300 = 33.3%", "sourceParagraph": 535 }, { "type": "paragraph", "text": "It has been. The actual population share is 20%. In the global population account, Language Y cannot take a 33.3% weight. Within-stratum normalised weights:", "sourceParagraphs": [ 536, 537, 538, 539 ] }, { "type": "bullet", "text": "Total weight for Language X: 80%", "sourceParagraph": 540 }, { "type": "bullet", "text": "Total weight for Language Y: 20%", "sourceParagraph": 541 }, { "type": "paragraph", "text": "is preserved. Each of the 100 observations in Language Y receives a lower individual weight than a single observation in Language X. In contrast, all 100 observations are used in Language Y’s separate diagnostic report. This approach: increases subgroup visibility, does not distort the global population estimate.", "sourceParagraphs": [ 542, 543, 544, 545, 546, 547 ] }, { "type": "heading2", "text": "23. ALLOCATION OF MULTILINGUAL USERS", "sourceParagraph": 549 }, { "type": "paragraph", "text": "A person may be suitable for more than one language cell. One of the preferred approaches for the main Population Panel is to assign a predefined main audit language to the user. The main audit language can be determined according to these fields:", "sourceParagraphs": [ 550, 551, 552 ] }, { "type": "bullet", "text": "Daily AI usage language", "sourceParagraph": 553 }, { "type": "bullet", "text": "Mother or home language", "sourceParagraph": 554 }, { "type": "bullet", "text": "Market and locale", "sourceParagraph": 555 }, { "type": "bullet", "text": "Competence in naturally using the prompt", "sourceParagraph": 556 }, { "type": "bullet", "text": "Panel balance", "sourceParagraph": 557 }, { "type": "paragraph", "text": "If the user will also be tested in other languages, these observations:", "sourceParagraphs": [ 558 ] }, { "type": "bullet", "text": "paired multilingual panel,", "sourceParagraph": 559 }, { "type": "bullet", "text": "Language Fairness Panel", "sourceParagraph": 560 }, { "type": "paragraph", "text": "are also labelled accordingly. It is prohibited for the same person to receive full population weighting in all languages.", "sourceParagraphs": [ 561, 562 ] }, { "type": "heading2", "text": "24. LANGUAGE ALLOCATION WITHIN THE COUNTRY", "sourceParagraph": 564 }, { "type": "paragraph", "text": "Let the suitable population for language groups l in country c be: N_{c,l}. If the country quota is: n_c, then the raw language quota:", "sourceParagraphs": [ 565, 566, 567, 568, 569, 570 ] }, { "type": "formula", "text": "q_{c,l} = n_c × N_{c,l} / Σ_l N_{c,l}", "sourceParagraph": 571 }, { "type": "paragraph", "text": "can be calculated as such. However, if language groups overlap:", "sourceParagraphs": [ 572, 573 ] }, { "type": "formula", "text": "Σ_l N_{c,l} > N_c", "sourceParagraph": 574 }, { "type": "paragraph", "text": "may be possible. In this case, using a direct share creates double counting. Solution options:", "sourceParagraphs": [ 575, 576, 577 ] }, { "type": "bullet", "text": "main audit language classification,", "sourceParagraph": 578 }, { "type": "bullet", "text": "split human weight,", "sourceParagraph": 579 }, { "type": "bullet", "text": "multilingual common model,", "sourceParagraph": 580 }, { "type": "bullet", "text": "separate matched language panel", "sourceParagraph": 581 }, { "type": "paragraph", "text": "may be used. The method used should be recorded.", "sourceParagraphs": [ 582, 583 ] }, { "type": "heading2", "text": "25. ALLOCATION OF PLAN AND INTERFACE WITHIN THE POPULATION", "sourceParagraph": 585 }, { "type": "paragraph", "text": "Free and paid users or web and mobile surfaces may behave differently. However, dividing each country–language cell separately into plan and interface layers can overly fragment the sample. Three approaches can be used.", "sourceParagraphs": [ 586, 587, 588 ] }, { "type": "heading3", "text": "25.1. Suballocation Proportional to Population", "sourceParagraph": 590 }, { "type": "paragraph", "text": "If the actual user distribution is known, allocation is made according to plan and interface shares.", "sourceParagraphs": [ 591 ] }, { "type": "heading3", "text": "25.2. Balanced Experimental Allocation", "sourceParagraph": 593 }, { "type": "paragraph", "text": "Equal or greater sampling is carried out to compare the difference between plan or interface. In global estimation, it is re-weighted to the actual usage shares.", "sourceParagraphs": [ 594, 595 ] }, { "type": "heading3", "text": "25.3. Separate Product Surface Scores", "sourceParagraph": 597 }, { "type": "paragraph", "text": "Free web, paid web, and mobile products are reported as separate audit objects. A combined result can also be produced. The approach to be used should be defined before measurement.", "sourceParagraphs": [ 598, 599, 600 ] }, { "type": "heading2", "text": "26. INVITATION NUMBER AND VALID OBSERVATION TARGET", "sourceParagraph": 602 }, { "type": "paragraph", "text": "If the valid observation target for a stratum is nh, inviting exactly nh people is usually not sufficient. Because:", "sourceParagraphs": [ 603, 604 ] }, { "type": "bullet", "text": "some people do not participate,", "sourceParagraph": 605 }, { "type": "bullet", "text": "some do not complete the task,", "sourceParagraph": 606 }, { "type": "bullet", "text": "some records remain out of protocol,", "sourceParagraph": 607 }, { "type": "paragraph", "text": "some evidence packages become invalid. Let the expected completion rate for the stratum be rh and the expected validity rate be vh. The invitation target is approximately:", "sourceParagraphs": [ 608, 609, 610, 611, 612, 613, 614 ] }, { "type": "formula", "text": "m_h = ⌈n_h / (r_h v_h)⌉", "sourceParagraph": 615 }, { "type": "paragraph", "text": "can be calculated as follows. Example:", "sourceParagraphs": [ 616, 617 ] }, { "type": "bullet", "text": "Valid observation target: 100", "sourceParagraph": 618 }, { "type": "bullet", "text": "Expected completion rate: 80 per cent", "sourceParagraph": 619 }, { "type": "bullet", "text": "Expected validity rate: 90 per cent", "sourceParagraph": 620 }, { "type": "formula", "text": "m_h = ⌈100 / (0.80 × 0.90)⌉ = 139", "sourceParagraph": 621 }, { "type": "paragraph", "text": "people can be invited. This is not a guarantee. As the field progresses, the predefined backup system can be used.", "sourceParagraphs": [ 622, 623, 624 ] }, { "type": "heading2", "text": "27. LIMIT OF INVITATION INFLATION", "sourceParagraph": 626 }, { "type": "paragraph", "text": "Inviting more people alone does not solve sample bias. Early responders may be systematically different from late responders. Only taking the first 1,000 completers:", "sourceParagraphs": [ 627, 628, 629 ] }, { "type": "bullet", "text": "time zone,", "sourceParagraph": 630 }, { "type": "bullet", "text": "technology interest,", "sourceParagraph": 631 }, { "type": "bullet", "text": "connection quality,", "sourceParagraph": 632 }, { "type": "bullet", "text": "work schedule", "sourceParagraph": 633 }, { "type": "paragraph", "text": "may produce bias in terms of. Invitation and selection order:", "sourceParagraphs": [ 634, 635 ] }, { "type": "bullet", "text": "should be randomised,", "sourceParagraph": 636 }, { "type": "bullet", "text": "must be recorded within the stratum,", "sourceParagraph": 637 }, { "type": "paragraph", "text": "should be managed throughout the field window.", "sourceParagraphs": [ 638 ] }, { "type": "heading2", "text": "28. ALTERNATE PARTICIPANT SYSTEM", "sourceParagraph": 640 }, { "type": "paragraph", "text": "An ordered backup list should be created for each stratum before the results are seen. A backup participant can be activated only for the following reasons:", "sourceParagraphs": [ 641, 642 ] }, { "type": "bullet", "text": "No response to the invitation", "sourceParagraph": 643 }, { "type": "bullet", "text": "Failure to complete the task", "sourceParagraph": 644 }, { "type": "bullet", "text": "Technical or ethical suitability issue", "sourceParagraph": 645 }, { "type": "bullet", "text": "Violation of the demand protocol", "sourceParagraph": 646 }, { "type": "bullet", "text": "Invalidity of the evidence package", "sourceParagraph": 647 }, { "type": "bullet", "text": "Re-registration of the same person", "sourceParagraph": 648 }, { "type": "paragraph", "text": "The backup user cannot intervene for the following reason:", "sourceParagraphs": [ 649 ] }, { "type": "callout", "text": "The previous user's AI response gave an incorrect, negative, or unwanted output.", "sourceParagraph": 650 }, { "type": "paragraph", "text": "An incorrect AI response is the valid system result. It cannot be a reason for participant change.", "sourceParagraphs": [ 651, 652 ] }, { "type": "heading2", "text": "29. RESULT BLINDNESS IN VALIDITY DECISION", "sourceParagraph": 654 }, { "type": "paragraph", "text": "The person deciding whether the observation complies with the protocol as much as possible:", "sourceParagraphs": [ 655 ] }, { "type": "bullet", "text": "must evaluate the response without knowing whether it is correct or incorrect,", "sourceParagraph": 656 }, { "type": "bullet", "text": "or whether it is positive or negative for the brand.", "sourceParagraph": 657 }, { "type": "paragraph", "text": "First:", "sourceParagraphs": [ 658, 659 ] }, { "type": "bullet", "text": "Capture and protocol validity", "sourceParagraph": 660 }, { "type": "bullet", "text": "Then the substantive response evaluation", "sourceParagraph": 661 }, { "type": "paragraph", "text": "must be done. Otherwise, negative responses can more easily be declared: \"invalid.\"", "sourceParagraphs": [ 662, 663, 664, 665 ] }, { "type": "heading2", "text": "30. RESERVE ORDER WITHIN LAYER", "sourceParagraph": 667 }, { "type": "paragraph", "text": "The reserve participant should be from the same as much as possible:", "sourceParagraphs": [ 668 ] }, { "type": "bullet", "text": "country,", "sourceParagraph": 669 }, { "type": "bullet", "text": "language,", "sourceParagraph": 670 }, { "type": "bullet", "text": "product surface,", "sourceParagraph": 671 }, { "type": "bullet", "text": "plan,", "sourceParagraph": 672 }, { "type": "bullet", "text": "panel condition", "sourceParagraph": 673 }, { "type": "paragraph", "text": "layer. If no reserve remains in the same layer, predefined escalation is applied:", "sourceParagraphs": [ 674, 675 ] }, { "type": "bullet", "text": "Additional random selection within the same layer", "sourceParagraph": 676 }, { "type": "bullet", "text": "Predefined upper layer pool", "sourceParagraph": 677 }, { "type": "bullet", "text": "Leaving allocation incomplete and scope warning", "sourceParagraph": 678 }, { "type": "bullet", "text": "New analysis version", "sourceParagraph": 679 }, { "type": "paragraph", "text": "After the result is seen, bringing a user from another country and closing the gap cannot be done silently.", "sourceParagraphs": [ 680 ] }, { "type": "heading2", "text": "31. WHAT HAPPENS IF A LAYER IS TOO SHORT?", "sourceParagraph": 682 }, { "type": "paragraph", "text": "It may fall below a layer target. Example:", "sourceParagraphs": [ 683, 684 ] }, { "type": "bullet", "text": "target: 100", "sourceParagraph": 685 }, { "type": "bullet", "text": "valid observation: 72", "sourceParagraph": 686 }, { "type": "paragraph", "text": "Correct options:", "sourceParagraphs": [ 687 ] }, { "type": "bullet", "text": "Extend the field window to a predefined extent", "sourceParagraph": 688 }, { "type": "bullet", "text": "Use the backup list", "sourceParagraph": 689 }, { "type": "bullet", "text": "Merge the cell with a similar and predefined upper layer", "sourceParagraph": 690 }, { "type": "bullet", "text": "Report the result with wide uncertainty", "sourceParagraph": 691 }, { "type": "bullet", "text": "Do not publish the cell result", "sourceParagraph": 692 }, { "type": "bullet", "text": "Recalculate the global weight using an appropriate method", "sourceParagraph": 693 }, { "type": "paragraph", "text": "Wrong option: Completely remove the low-scoring layer from the analysis.", "sourceParagraphs": [ 694, 695 ] }, { "type": "heading2", "text": "32. POST-RESULT QUOTA CHANGE", "sourceParagraph": 697 }, { "type": "paragraph", "text": "The following changes are result-oriented sample manipulations:", "sourceParagraphs": [ 698 ] }, { "type": "bullet", "text": "Reducing the target quota of the low-scoring country", "sourceParagraph": 699 }, { "type": "bullet", "text": "Increasing the quota of the high-scoring language", "sourceParagraph": 700 }, { "type": "bullet", "text": "Invalidating users who provide negative responses", "sourceParagraph": 701 }, { "type": "bullet", "text": "Redefining the small country pool based on results", "sourceParagraph": 702 }, { "type": "bullet", "text": "Excluding a plan with critical errors", "sourceParagraph": 703 }, { "type": "bullet", "text": "Adding new users until the desired average is reached", "sourceParagraph": 704 }, { "type": "paragraph", "text": "Each material quota change:", "sourceParagraphs": [ 705 ] }, { "type": "bullet", "text": "new allocation version,", "sourceParagraph": 706 }, { "type": "bullet", "text": "justification,", "sourceParagraph": 707 }, { "type": "bullet", "text": "date,", "sourceParagraph": 708 }, { "type": "bullet", "text": "decision maker", "sourceParagraph": 709 }, { "type": "bullet", "text": "Preservation of the previous result", "sourceParagraph": 710 }, { "type": "paragraph", "text": "should be done with.", "sourceParagraphs": [ 711 ] }, { "type": "heading2", "text": "33. MULTI-STAGE SAMPLING", "sourceParagraph": 713 }, { "type": "paragraph", "text": "Multi-stage sampling can be used when direct access to all individuals is not possible. Example:", "sourceParagraphs": [ 714, 715 ] }, { "type": "bullet", "text": "Selection of region or country pool", "sourceParagraph": 716 }, { "type": "bullet", "text": "Country selection", "sourceParagraph": 717 }, { "type": "bullet", "text": "Selection of panel or local research partner", "sourceParagraph": 718 }, { "type": "bullet", "text": "User selection", "sourceParagraph": 719 }, { "type": "bullet", "text": "AI product or task assignment", "sourceParagraph": 720 }, { "type": "paragraph", "text": "The total probability of selection for an individual user is the product of the conditional probabilities at each stage:", "sourceParagraphs": [ 721 ] }, { "type": "formula", "text": "π_i = π_r × π_{c|r} × π_{f|c} × π_{i|f}", "sourceParagraph": 722 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 723 ] }, { "type": "bullet", "text": "r: region or small country pool", "sourceParagraph": 724 }, { "type": "bullet", "text": "c: country", "sourceParagraph": 725 }, { "type": "bullet", "text": "f: sampling frame or panel", "sourceParagraph": 726 }, { "type": "bullet", "text": "i: person", "sourceParagraph": 727 }, { "type": "paragraph", "text": "Basic design weight:", "sourceParagraphs": [ 728 ] }, { "type": "formula", "text": "d_i = 1/π_i", "sourceParagraph": 729 }, { "type": "paragraph", "text": "If selection probabilities cannot be recorded, the design cannot be presented as a fully probability-based sample. [K04; K07]", "sourceParagraphs": [ 730, 731 ] }, { "type": "heading2", "text": "34. ALLOCATION IN VOLUNTARY AND NON-PROBABILITY PANELS", "sourceParagraph": 733 }, { "type": "paragraph", "text": "A true probability-based world panel may not always be possible. A voluntary or professional online panel can be used. In this case, country and language quotas can still be set according to the product-eligible population. However:", "sourceParagraphs": [ 734, 735, 736, 737 ] }, { "type": "bullet", "text": "the participant's actual probability of selection may be unknown,", "sourceParagraph": 738 }, { "type": "bullet", "text": "self-selection may occur,", "sourceParagraph": 739 }, { "type": "paragraph", "text": "weighting can only make adjustments based on observed variables. The correct status:", "sourceParagraphs": [ 740, 741 ] }, { "type": "heading4", "text": "Non-probability user panel stratified and calibrated to the population", "sourceParagraph": 742 }, { "type": "paragraph", "text": "may be. The following statement cannot be used: “1,000 people selected completely at random from the world population.”", "sourceParagraphs": [ 743, 744, 745 ] }, { "type": "heading2", "text": "35. WEIGHTING ARCHITECTURE", "sourceParagraph": 747 }, { "type": "paragraph", "text": "The final observation weight may consist of multiple components:", "sourceParagraphs": [ 748 ] }, { "type": "formula", "text": "w_i = d_i × a_i^{NR} × g_i^{CAL}", "sourceParagraph": 749 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 750 ] }, { "type": "bullet", "text": "di: base design weight", "sourceParagraph": 751 }, { "type": "bullet", "text": "aiNR: non-response adjustment", "sourceParagraph": 752 }, { "type": "bullet", "text": "giCAL: calibration coefficient", "sourceParagraph": 753 }, { "type": "heading3", "text": "35.1. Design Weight", "sourceParagraph": 755 }, { "type": "paragraph", "text": "It is the inverse of the participant's probability of selection in the sampling design.", "sourceParagraphs": [ 756 ] }, { "type": "heading3", "text": "35.2. Nonresponse Adjustment", "sourceParagraph": 758 }, { "type": "paragraph", "text": "For selected persons with similar characteristics:", "sourceParagraphs": [ 759 ] }, { "type": "bullet", "text": "how many responded,", "sourceParagraph": 760 }, { "type": "bullet", "text": "how many produced a valid observation", "sourceParagraph": 761 }, { "type": "paragraph", "text": "can be taken into account. For sample response class r:", "sourceParagraphs": [ 762, 763 ] }, { "type": "formula", "text": "a_r^{NR} = (Σ_{j∈S_r} d_j) / (Σ_{i∈R_r} d_i)", "sourceParagraph": 764 }, { "type": "paragraph", "text": "Here:", "sourceParagraphs": [ 765 ] }, { "type": "bullet", "text": "Sr: selected eligible sample", "sourceParagraph": 766 }, { "type": "bullet", "text": "Rr: those who responded or produced a valid observation", "sourceParagraph": 767 }, { "type": "paragraph", "text": "may be. Nonresponse adjustment: does not completely resolve unobserved bias, should be explained along with its assumptions.", "sourceParagraphs": [ 768, 769, 770, 771 ] }, { "type": "heading3", "text": "35.3. Calibration Weight", "sourceParagraph": 773 }, { "type": "paragraph", "text": "The sample, to known population totals:", "sourceParagraphs": [ 774 ] }, { "type": "bullet", "text": "country,", "sourceParagraph": 775 }, { "type": "bullet", "text": "language,", "sourceParagraph": 776 }, { "type": "bullet", "text": "age,", "sourceParagraph": 777 }, { "type": "bullet", "text": "plan,", "sourceParagraph": 778 }, { "type": "bullet", "text": "user surface.", "sourceParagraph": 779 }, { "type": "paragraph", "text": "Calibration may be applied across these dimensions, but only where reliable external totals exist.", "sourceParagraphs": [ 780, 781 ] }, { "type": "heading2", "text": "36. WEIGHT NORMALISATION", "sourceParagraph": 783 }, { "type": "paragraph", "text": "Weights:", "sourceParagraphs": [ 784 ] }, { "type": "bullet", "text": "to the total population,", "sourceParagraph": 785 }, { "type": "bullet", "text": "to the total sample size,", "sourceParagraph": 786 }, { "type": "bullet", "text": "to individual", "sourceParagraph": 787 }, { "type": "paragraph", "text": "can be normalised. The normalisation used in the score formula should be explained. Individually normalised weights:", "sourceParagraphs": [ 788, 789, 790 ] }, { "type": "formula", "text": "w̃_i = w_i / Σ_j w_j", "sourceParagraph": 791 }, { "type": "paragraph", "text": "are obtained. Weighted estimate:", "sourceParagraphs": [ 792, 793 ] }, { "type": "formula", "text": "θ̂ = Σ_i w̃_iY_i", "sourceParagraph": 794 }, { "type": "paragraph", "text": "is calculated as follows.", "sourceParagraphs": [ 795 ] }, { "type": "heading2", "text": "37. WEIGHT TRIMMING", "sourceParagraph": 797 }, { "type": "paragraph", "text": "Some participants may receive very high weights because they represent very small and under-sampled groups. This situation:", "sourceParagraphs": [ 798, 799 ] }, { "type": "bullet", "text": "can increase the estimated variance,", "sourceParagraph": 800 }, { "type": "bullet", "text": "the influence of a single observation,", "sourceParagraph": 801 }, { "type": "bullet", "text": "and score volatility", "sourceParagraph": 802 }, { "type": "paragraph", "text": "can increase. Weights can be trimmed at an upper limit. However, trimming: may create bias in population representation, may make small groups invisible again. Therefore, weight trimming should:", "sourceParagraphs": [ 803, 804, 805, 806, 807, 808 ] }, { "type": "bullet", "text": "be defined before the results,", "sourceParagraph": 809 }, { "type": "bullet", "text": "be shown together with the untrimmed results,", "sourceParagraph": 810 }, { "type": "paragraph", "text": "and carry a sensitivity analysis.", "sourceParagraphs": [ 811 ] }, { "type": "heading2", "text": "38. EFFECTIVE SAMPLE SIZE", "sourceParagraph": 813 }, { "type": "paragraph", "text": "Even though the raw number of observations is n, if the weights vary greatly, the actual amount of information may be lower. Kish effective sample size:", "sourceParagraphs": [ 814, 815, 816, 817 ] }, { "type": "formula", "text": "n_eff = (Σ_i w_i)^2 / Σ_i w_i^2 [K08]", "sourceParagraph": 818 }, { "type": "paragraph", "text": "can be shown as. With equal weights:", "sourceParagraphs": [ 819, 820 ] }, { "type": "formula", "text": "neff≈n", "sourceParagraph": 821 }, { "type": "paragraph", "text": "is obtained. If the weights are very unbalanced: neff