First direct observation study: only 1% of 713K crowd submissions use AI, but 5% of workers drive 64%
manoelribeiro · x · 2026-10-14
Penn-led researchers directly observed AI use on an online research platform for the first time: surveying 2,500 Prolific workers and linking donated ChatGPT histories to submission records, covering 712,930 submissions across 127,000+ studies.
Key findings:
- 1 in 8 surveyed workers admitted ever using AI on a study, and 68% of observed weekly ChatGPT users had done so; yet AI assistance appeared in only 1% of submissions, with no detectable increase over 3+ years.
- Assistance was highly concentrated: 5% of workers accounted for 64% of assisted submissions, often localized within tasks.
- Workers were virtually always paid even when instructions prohibited AI, with approval rates similar to unassisted submissions — revealing gaps in platform safeguards.
More from Research
- Beyond automation: Stanford economists sketch a white-box, human-AI collaboration research agenda — AI Snake Oil · 2026-10-15
- AI-in-biology startup hiring MTS to build rigorous benchmarks for AI in biology — kenbwork · 2026-10-15
- SlipSense: multimodal tactile slip detection hits 96.7% F1 at 23.1ms with zero-shot robot transfer — chris_j_paxton · 2026-10-15
- 2021 slides reveal Garçon, Anthropic's internal interpretability infra for multi-GPU models — voooooogel · 2026-10-15
- Microduck robot runs onboard SLAM with just 64 pixels of depth, 4.8 cm median error — chris_j_paxton · 2026-10-15
- UIUC paper explains sparse LLM post-training updates through the lens of Fisher information — tokenbender · 2026-10-15