Wharton paper: 'Cognitive Surrender' — 1,372-person experiments show AI overwrites human judgment
2C_ornot2C · x · 2026-11-22
A Wharton School paper by Steven Shaw and Gideon Nave, Thinking—Fast, Slow, and Artificial, extends dual-process theory with a "System 3": artificial cognition operating outside the brain. Its key finding is cognitive surrender — adopting AI outputs with minimal scrutiny.
Across three preregistered experiments (N=1,372; 9,593 reasoning trials on an adapted Cognitive Reflection Test), the authors secretly randomized AI accuracy:
- Participants consulted the AI on >50% of trials
- Accuracy rose +25 points when the AI was right, fell -15 points when it was wrong (Cohen's h=0.81)
- Participants didn't double-check outputs or catch errors — System 3 overrode intuition and critical thinking
- Using the AI measurably increased confidence even when it was completely wrong
Time pressure, incentives, and feedback shifted baselines but never eliminated the pattern. People with higher AI trust and lower need for cognition and fluid intelligence surrendered more. The authors argue this reframes human reasoning, autonomy, and accountability in the AI era.
More from AGI Musings
- Turing Award winner David Patterson declares AGI and ASI already achieved — davidpattersonx · 2026-11-22
- x402 micropayments framed as the blueprint for a million-agent app layer — kleffew94 · 2026-11-22
- Mathematicians' anti-AI arguments echo 1950s calculator backlash — Promptmethus · 2026-11-22
- From 400 applications to 1 hire: why the pick came from the reject pile — brandongalang · 2026-11-22
- "It's Not AI or SI": Blogger Coins 'Xeno Intelligence' as a Categorically Alien Intelligence — SydSteyerhart · 2026-11-22
- Top HN comment: you learned to code to make software — don't confuse tools with the job — tlakomy · 2026-11-22