Duplicate from GBX1220Max/ai-fit-scan

Co-authored-by: Max Guo <GBX1220Max@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,60 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.avro filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mds filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# Video files - compressed
+*.mp4 filter=lfs diff=lfs merge=lfs -text
+*.webm filter=lfs diff=lfs merge=lfs -text

CITATION.cff

@@ -0,0 +1,19 @@
+cff-version: 1.2.0
+message: "If you use this dataset, please cite it as below."
+title: "AI-Fit-Scan: A Labeled Dataset of AI-Powered Fitness Applications on Google Play"
+authors:
+  - given-names: Baixin
+    family-names: Guo
+    email: gbx1220max@gmail.com
+    orcid: ""
+version: 1.0.0
+date-released: 2026-05-29
+repository-code: https://github.com/GBX-Max1220/AI-Fit-Scan
+license: CC-BY-4.0
+keywords:
+  - ai-fitness
+  - ai-washing
+  - mobile-apps
+  - google-play
+  - annotation
+  - health-fitness

LICENSE

@@ -0,0 +1,13 @@
+Creative Commons Attribution 4.0 International License
+
+Copyright (c) 2026 Baixin (Max) Guo
+
+This work is licensed under the Creative Commons Attribution 4.0 International License.
+To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/
+
+You are free to:
+- Share — copy and redistribute the material in any medium or format
+- Adapt — remix, transform, and build upon the material for any purpose, even commercially
+
+Under the following terms:
+- Attribution — You must give appropriate credit, provide a link to the license, and indicate if changes were made.

README.md

@@ -0,0 +1,200 @@
+---
+license: cc-by-4.0
+task_categories:
+  - text-classification
+  - tabular-classification
+language:
+  - en
+tags:
+  - ai-fitness
+  - ai-washing
+  - mobile-apps
+  - google-play
+  - health-fitness
+  - annotation
+size_categories:
+  - n<1k
+---
+
+# AI-Fit-Scan: A Labeled Dataset of AI-Powered Fitness Applications on Google Play
+
+## Dataset Description
+
+- **Homepage:** https://huggingface.co/datasets/MaxGuo/ai-fit-scan
+- **Repository:** https://github.com/GBX-Max1220/AI-Fit-Scan
+- **Paper:** *Coming soon*
+- **Point of Contact:** Max Guo (gbx1220max@gmail.com)
+
+### Summary
+
+AI-Fit-Scan is a manually annotated dataset of **161 fitness-related mobile applications** scraped from Google Play, with **61 apps claiming AI capabilities** receiving fine-grained human annotation for AI authenticity. We introduce a three-tier labeling framework (True AI / Quasi-AI / Fake AI) and achieve **inter-annotator agreement of κ = 0.856 (Cohen's Kappa)**, indicating almost perfect agreement.
+
+**Key finding:** Among 61 apps claiming to be "AI-powered fitness" applications, **36.1% (22) are fake AI** — their core functionality does not use AI/ML, and **only 57.4% (35) are confirmed true AI**.
+
+### Why This Dataset Matters
+
+1. **AI-Washing quantification:** First systematic measurement of AI claim inflation in the fitness app market
+2. **Reusable annotation framework:** Three-tier (T/Q/F) labeling methodology with proven reliability
+3. **AI function taxonomy:** Six-category classification of AI capabilities in fitness apps
+4. **Academic utility:** Benchmark for studying AI claim verification, consumer deception, and health app regulation
+
+---
+
+## Dataset Structure
+
+### Files
+
+| File | Description | Rows |
+|------|-------------|------|
+| `ai_fit_scan_full.csv` | Complete 161-app dataset with L1 metadata labels | 161 |
+| `ai_fit_scan_annotated.csv` | 61 AI-claiming apps with L2 human annotation | 61 |
+
+### Columns
+
+**Both files:**
+- `name` — App name
+- `category` — Google Play category
+- `installs` — Download count
+- `rating` — User rating (0-5)
+- `reviews` — Number of user reviews
+- `description` — App description from Google Play
+- `l1_label` — L1 automatic classification (AI_FITNESS / FITNESS_NO_AI / EXCLUDE_GENERIC / AI_NOT_FITNESS / UNRELATED)
+
+**Annotated file only:**
+- `l2_label` — L2 human annotation (T / Q / F)
+- `ai_function` — AI capability category (see taxonomy below)
+- `annotator_a` — Annotator A label
+- `annotator_b` — Annotator B label
+- `l2_evidence` — Evidence for L2 label (source URL or rationale)
+
+### Label Definitions
+
+#### L1: Metadata-Based Classification
+
+| Label | Definition | Count |
+|-------|-----------|-------|
+| AI_FITNESS | Health/Fitness category + AI keywords in description | 61 |
+| FITNESS_NO_AI | Health/Fitness category, no AI keywords | 50 |
+| EXCLUDE_GENERIC | Generic AI tools (chatbots, translators, etc.) | 25 |
+| AI_NOT_FITNESS | AI keywords but non-fitness category | 13 |
+| UNRELATED | Neither fitness nor AI | 12 |
+
+#### L2: Human Annotation of AI Authenticity
+
+| Label | Definition | Count | % |
+|-------|-----------|-------|---|
+| **T (True AI)** | AI/ML is the core driver of the app's primary functionality. Removing AI would fundamentally change the product. | 35 | 57.4% |
+| **Q (Quasi-AI)** | App claims AI but evidence is inconclusive. May use rule-based algorithms or simple heuristics marketed as AI. | 4 | 6.6% |
+| **F (Fake AI)** | AI label is marketing only. Core functionality works without AI, or "AI" refers to basic automation/features unrelated to ML. | 22 | 36.1% |
+
+**Annotation criteria:**
+- **T:** Description explicitly describes AI/ML-driven personalization, adaptive planning, computer vision, or LLM-based coaching as a core feature
+- **Q:** App name or description claims "AI" but lacks specific mechanism description; could be rule-based
+- **F:** "Smart"/"AI" used as marketing buzzword; core is workout logging, pre-set routines, or timer functionality; AI only used for peripheral features (e.g., food photo recognition in a calorie counter)
+
+### AI Function Taxonomy
+
+| Category | Description | Count |
+|----------|-------------|-------|
+| Plan Generation | AI generates/adapts personalized workout plans based on user data | 28 |
+| LLM Chat Coach | LLM-powered conversational coaching interface | 8 |
+| Nutrition AI | AI-driven diet and nutrition recommendations | 6 |
+| Pose/Motion Detection | Computer vision for real-time form checking and rep counting | 5 |
+| Wearable Integration | AI adjusts plans based on biometric data (HRV, sleep, etc.) | 4 |
+| Voice Coach | Real-time AI voice guidance during workouts | 3 |
+
+---
+
+## Data Collection
+
+### Methodology
+
+1. **Search strategy:** 10 keyword queries on Google Play (`"AI fitness"`, `"AI workout"`, `"AI personal trainer"`, `"AI gym"`, `"AI exercise"`, `"AI coach"`, `"AI training plan"`, `"AI running"`, `"smart fitness"`, `"AI health"`)
+2. **Deduplication:** Removed duplicate entries by package name
+3. **Collection date:** May 2026
+4. **Data extracted:** App name, category, installs, rating, reviews, description, developer, price, version history
+
+### Annotation Process
+
+1. **L1 automatic classification:** Keyword-based filtering using AI-related terms and category matching
+2. **L2 human annotation:** Two annotators independently labeled all 61 AI_FITNESS apps
+   - Annotator A: AI research assistant (automated labeling + L2 web search verification)
+   - Annotator B: Domain expert (CSCS-certified, HCI researcher)
+3. **L2 evidence collection:** For ambiguous cases (initially labeled Q), web search was conducted to verify AI technology claims against official websites, technical documentation, and third-party reviews
+4. **Disagreement resolution:** 5 disagreements (all Q vs T/F) resolved by adopting Annotator B's judgment after reviewing app screenshots and user reviews
+
+### Inter-Annotator Agreement
+
+| Metric | Value |
+|--------|-------|
+| Raw agreement | 56/61 = 91.8% |
+| Cohen's Kappa | **0.856** |
+| Interpretation | Almost Perfect (Landis & Koch, 1977) |
+
+---
+
+## Key Findings
+
+### 1. AI-Washing is Rampant
+36.1% of apps claiming "AI" in the fitness space have no meaningful AI in their core functionality. The most common fake-AI patterns:
+- "Smart" = basic algorithm/timer (e.g., Smart Workout Counter = interval timer)
+- "AI-Powered" added to existing traditional apps (e.g., MyFitnessPal adding AI food recognition)
+- "Personal Trainer" marketing without any ML (e.g., Home Workout - No Equipment)
+
+### 2. Fake AI Apps Have Higher Ratings
+| Group | Mean Rating | Median Rating |
+|-------|-------------|---------------|
+| True AI (T) | 4.28 | 4.50 |
+| Fake AI (F) | 4.41 | 4.54 |
+
+**High-rating apps sell outcome promises; low-rating apps sell AI technology.**
+
+### 3. Long Tail of Irrelevance
+- 28 of 61 "AI fitness" apps have <50K downloads
+- 15 apps have zero or missing ratings — likely inactive or abandoned
+- The market has not consolidated; room for genuine AI entrants
+
+### 4. AI Function Distribution
+Plan generation dominates (28/35 true-AI apps), while pose detection (5) and voice coaching (3) remain underdeveloped despite high user value potential.
+
+---
+
+## Limitations
+
+1. **Single platform:** Google Play only; Apple App Store data not included
+2. **Single time point:** Data collected May 2026; AI claims may change with updates
+3. **Description-dependent:** L2 annotation relies primarily on app descriptions and public documentation, not source code inspection
+4. **Binary framing:** The T/Q/F taxonomy simplifies a spectrum; some "Quasi-AI" apps may use sophisticated rule engines that approach ML-level personalization
+5. **Keyword bias:** Search results favor apps with "AI" in their name/description; genuine AI apps using different marketing language may be missed
+
+---
+
+## Ethical Considerations
+
+- App names and descriptions are publicly available Google Play metadata
+- Developer contact and revenue data are **not** included
+- We do not accuse any app of fraud; "Fake AI" refers to absence of ML in core functionality, not malicious intent
+- Apps labeled F may use AI in peripheral features (e.g., food recognition, recommendation systems)
+
+---
+
+## Citation
+
+```bibtex
+@dataset{guo2026aifitscan,
+  title={AI-Fit-Scan: A Labeled Dataset of AI-Powered Fitness Applications on Google Play},
+  author={Guo, Baixin (Max) and MaxCoze},
+  year={2026},
+  publisher={HuggingFace},
+  url={https://huggingface.co/datasets/MaxGuo/ai-fit-scan}
+}
+```
+
+## License
+
+Creative Commons Attribution 4.0 International (CC BY 4.0)
+
+## Acknowledgments
+
+- Cal Dietz Triphasic Training system — conceptual influence on training periodization analysis
+- Google Play Store — public data source

ai_fit_scan_annotated.csv

@@ -0,0 +1,62 @@
+app_index,l2_label,ai_function,annotator_a,annotator_b,l2_evidence
+1,T,plan_generation,T,T,Official website confirms AI-powered algorithms + scientifically backed
+2,T,plan_generation,Q,T,"fitaiapp.com: 7.5M users, AI adaptive plans, 24/7 AI coach, 100M+ training data points"
+3,T,plan_generation,T,T,Description: combines structured training with AI mechanism
+4,T,plan_generation,T,T,"Industry veteran, hyper-personalized AI since early days"
+5,Q,plan_generation,Q,Q,"Description primarily disclaimers, no specific AI mechanism described"
+6,T,nutrition_ai,T,T,"40M users, AI diet + fitness, well-established"
+7,F,plan_generation,Q,F,"Domain 6 months old, WHOIS hidden, scam-suspect rating"
+8,T,plan_generation,T,T,AI Personal Trainer with personalized training
+9,T,plan_generation,T,T,"Adaptive workout, adjusts to user life"
+10,F,plan_generation,Q,F,Multiple同名apps on GP; lightweight implementation likely
+11,T,plan_generation,Q,T,User viewed app; confirmed AI features
+12,T,plan_generation,T,T,"Industry benchmark, ""uses AI to help you make progress"""
+13,F,plan_generation,F,F,"Traditional tracker, ""plan workouts, log training"""
+14,T,plan_generation,T,T,Science-backed AI personal trainer
+15,T,plan_generation,T,T,AI trainer with personalized plans
+16,T,plan_generation,T,T,Dual AI: personal trainer + nutritionist
+17,F,plan_generation,Q,F,Renpho exercise bike companion; AI is in hardware (auto-resistance)
+18,T,plan_generation,T,T,Creates personalized workouts based on goals/history/equipment
+19,T,plan_generation,T,T,"Explicitly ""not generic"", AI-Powered"
+20,T,llm_chat_coach,T,T,AI workout generator + calorie counter + macro tracker + meal planner
+21,F,plan_generation,Q,F,"User viewed app; ""algorithms"" ≠ ML, rule-based"
+22,T,pose_detection,T,T,"""Advanced AI and machine learning"" explicitly stated"
+23,F,plan_generation,F,F,"Smart tracker = log app, no AI core"
+24,F,plan_generation,F,F,"Claims ""AI-Driven"" but description is pure workout planner"
+25,F,plan_generation,F,F,Traditional gym tracker/planner
+26,T,plan_generation,T,T,"""Keeps adjusting as you go"", adaptive"
+27,Q,llm_chat_coach,Q,Q,"""Powered by AI"" but description generic"
+28,T,plan_generation,T,T,"""Powered by AI and guided by top experts"""
+29,T,llm_chat_coach,Q,T,"mytrainerapp.io: agentic AI, 24/7 LLM chat coach, adaptive plans"
+30,T,plan_generation,T,T,"Periodized, science-backed, adapts to fatigue"
+31,T,plan_generation,T,T,Adaptive training plans for endurance athletes
+32,F,plan_generation,Q,F,"Freediving CO2 tolerance calculator, not ML"
+33,F,plan_generation,F,F,"Pre-set templates, not AI"
+34,T,plan_generation,T,T,"AI-powered endurance training, academic backing"
+35,F,llm_chat_coach,F,F,"AI journal/diary, not fitness AI"
+36,T,plan_generation,T,T,AI-generated football training plans in 60 seconds
+37,F,llm_chat_coach,F,F,"AI therapy/coaching, mental health not fitness"
+38,T,nutrition_ai,T,T,AI-powered workout + nutrition plans
+39,T,plan_generation,Q,T,"Multi-source: predictive algo since 2019, MediaPipe pose detection, Coach+ LLM chat"
+40,T,nutrition_ai,T,T,"AI weight-loss coach, photo food recognition + voice"
+41,T,plan_generation,T,T,AI Coach creates plans from 5000+ exercises
+42,T,plan_generation,Q,T,User viewed app; running coach with Strava data-driven adaptation
+43,T,plan_generation,T,T,Creates fully personalized plans using AI
+44,T,plan_generation,T,T,AI running coach with deep analysis
+45,T,plan_generation,T,T,Adaptive coach powered by Strava data
+46,T,llm_chat_coach,T,T,"""Powered by ChatGPT-4 technology"" explicitly stated"
+47,T,plan_generation,T,T,AI-powered personalized plans + auto-adjustment
+48,F,plan_generation,Q,F,"Dr. Muscle review: AI only does progressive overload, poorly calibrated"
+49,T,plan_generation,T,T,"""Artificial intelligence, training plans"""
+50,F,plan_generation,F,F,"AI tracker but core is ranking/challenges, AI auxiliary"
+51,T,pose_detection,T,T,Auto rep counting + form analysis via Wear OS (computer vision)
+52,Q,plan_generation,Q,Q,"Description vague, ""AI Fitness Coach"" generic"
+53,F,plan_generation,F,F,"Traditional equipment companion, Coach is pre-set"
+54,F,plan_generation,F,F,"""Smart"" = basic tracker"
+55,F,plan_generation,F,F,"Simple tracker, no AI"
+56,F,nutrition_ai,F,F,"Calorie counter, AI only for food photo recognition (peripheral)"
+57,F,plan_generation,F,F,"Traditional 7-min workout, zero AI"
+58,F,plan_generation,F,F,"""Smart"" = Excel/PDF replacement"
+59,Q,plan_generation,Q,Q,"""Personal trainer alternative"", vague AI claim"
+60,F,plan_generation,F,F,Pure interval timer
+61,F,plan_generation,F,F,Simplest interval timer with counting sound