Spaces:
Running
Running
| /** | |
| * @szl-holdings/a11oy-knowledge — Proposed Axioms A10–A14 | |
| * Author: Lutar, Stephen P. · ORCID 0009-0001-0110-4173 · Apache-2.0 | |
| * Source: publications_harvest/niche_mind/INNOVATIONS.md §3 | |
| */ | |
| import type { ProposedAxiom } from './schema.js'; | |
| export const PROPOSED_AXIOMS: ProposedAxiom[] = [ | |
| { | |
| id: 'A10', | |
| name: 'temporalConsistency', | |
| statement: 'Gate verdict is invariant under time-shift within registered clock-drift bound ε_clock.', | |
| statement_latex: '\\forall \\text{input}, \\forall \\Delta t \\leq \\varepsilon_{\\text{clock}}: \\text{verdict}(\\Lambda, \\text{input}, t) = \\text{verdict}(\\Lambda, \\text{input}, t + \\Delta t)', | |
| source_file: 'INNOVATIONS.md', | |
| source_section: '§3', | |
| maturity: 'defined', | |
| citation: 'https://doi.org/10.5281/zenodo.20119582', | |
| floor: 0.90, | |
| why_not_subsumed: 'A5 (deterministicReplay) is spatial determinism. A10 is temporal determinism — different evaluation time, same verdict. These are independent properties.', | |
| function_signature: 'temporalConsistency(receipt: Receipt, deltaT_ms: number, clockDriftBound_ms: number): boolean', | |
| falsifiability_test: 'Evaluate same receipt at t and t+100ms; assert verdict identical.', | |
| }, | |
| { | |
| id: 'A11', | |
| name: 'causalSeparability', | |
| statement: 'Receipts from disjoint actor sets carry statistically independent entropy (no shared randomness source).', | |
| statement_latex: '\\text{actor}(r) \\cap \\text{actor}(r\') = \\emptyset \\implies \\text{receipt\\_hash}(r) \\perp \\text{receipt\\_hash}(r\')', | |
| source_file: 'INNOVATIONS.md', | |
| source_section: '§3', | |
| maturity: 'defined', | |
| citation: 'https://doi.org/10.5281/zenodo.20119582', | |
| why_not_subsumed: 'A4 ensures two witnesses are distinct per receipt. A11 ensures receipts from different actors carry independent entropy — enabling cross-actor attribution without information leakage.', | |
| function_signature: 'causalSeparability(actorSetA: Set<ActorId>, actorSetB: Set<ActorId>): boolean', | |
| falsifiability_test: 'Inspect canonical-chain.seed.json; assert no shared PRNG seed between disjoint actor slots.', | |
| }, | |
| { | |
| id: 'A12', | |
| name: 'constructiveTransparency', | |
| statement: 'Every Λ score is re-derivable from public inputs; no hidden weights or parameters.', | |
| statement_latex: '\\forall r: \\Lambda(r) = f(\\text{public\\_inputs}(r)) \\text{ where } f \\text{ is the published Lean-formalized scorer}', | |
| source_file: 'INNOVATIONS.md', | |
| source_section: '§3', | |
| maturity: 'defined', | |
| citation: 'https://doi.org/10.5281/zenodo.20053148', | |
| why_not_subsumed: 'TH_L1 proves uniqueness of the scorer given axioms. A12 adds auditability: the scorer must be a pure function of declared inputs. A "hidden trust boost" based on user history would satisfy TH_L1 but violate A12.', | |
| function_signature: 'constructiveTransparency(scorer: LambdaScorer): boolean // assert scorer is pure function', | |
| falsifiability_test: 'Run scorer on identical axis vectors from two different actors; assert identical output.', | |
| }, | |
| { | |
| id: 'A13', | |
| name: 'adversarialRobustness', | |
| statement: 'Gate verdict is stable under bounded perturbation of any single axis by ε ≤ 0.05, except when original score is within ε of a floor.', | |
| statement_latex: '\\forall x \\in [0,1]^9, \\|\\delta\\|_\\infty \\leq 0.05: \\text{gate\\_pass}(x) = \\text{gate\\_pass}(x + \\delta)', | |
| source_file: 'INNOVATIONS.md', | |
| source_section: '§3', | |
| maturity: 'defined', | |
| citation: 'https://doi.org/10.5281/zenodo.20119582', | |
| floor: 0.05, | |
| why_not_subsumed: 'A5 (soundnessAxiom) is binary — pass or fail. A13 adds a stability radius: defines how close to a floor boundary a score must be before perturbation can flip the verdict.', | |
| function_signature: 'adversarialRobustness(receipt: Receipt, epsilon: number): { is_robust: boolean; vulnerable_axis?: string }', | |
| falsifiability_test: '30 adversarial tests from runtime_memo §6.5: axis-flip attacks with ε=0.05 on each of 9 axes.', | |
| }, | |
| { | |
| id: 'A14', | |
| name: 'economicGrounding', | |
| statement: 'Gate passes only if action declared cost ≤ registered actor budget at evaluation time.', | |
| statement_latex: '\\text{gate\\_pass}(r) \\implies \\text{cost}(r) \\leq B_{\\text{actor}}(t)', | |
| source_file: 'INNOVATIONS.md', | |
| source_section: '§3', | |
| maturity: 'defined', | |
| citation: 'https://doi.org/10.5281/zenodo.20162352', | |
| why_not_subsumed: 'No existing axiom bounds economic cost. Required by financial services (SR 11-7 position limits), insurance (underwriting ceilings), capital markets (SEC Rule 17a-4 order size). A14 adds a financial constraint as a first-class gate condition.', | |
| function_signature: 'economicGrounding(action_cost: number, actor_id: ActorId, timestamp_ms: number): { pass: boolean; budget_remaining: number }', | |
| falsifiability_test: 'Register actor with budget=100; submit action with cost=101; assert gate fails.', | |
| }, | |
| ]; | |
| export const getProposedAxiom = (id: string): ProposedAxiom | undefined => | |
| PROPOSED_AXIOMS.find(a => a.id === id); | |