A 3,132-person study finds AI advice makes people far less likely to say “I don’t know”
量子位 · wechat · 2026-07-23
AI answers made people far less willing to say “I don’t know” in a 3,132-person study
Researchers from Paris-Saclay, Sapienza University of Rome, and the University of Milano-Bicocca ran five experiments with 3,132 participants to test how AI advice changes human judgment on deliberately difficult movie-detail questions.
- Without AI, many participants chose to pause or admit uncertainty.
- Once AI was available, the share of “I don’t know” responses collapsed — in one experiment from 44% to 3%.
- Accuracy also fell sharply, from 27% to 9% in the same setup, while confidence jumped from 30% to 76%.
- Even when wrong answers carried a penalty, AI advice still pulled people toward confident but incorrect responses.
The paper argues that the deeper risk is not just hallucinations, but that AI lowers the threshold for people to decide they know enough to answer. The authors warn this could weaken critical thinking, especially for children growing up with AI as a default layer in learning and work.
Related event: Study: AI Hallucinations Undermine Human Critical Thinking(2 posts)→
More from AGI Musings
- San Francisco’s Alamo Square rents jump 42% amid the AI boom — Polymarket · 2026-07-23
- To stay above AI, humans need rest, sleep, and solitude — dosco · 2026-07-23
- Podcast asks whether a single cell can learn, remember, or act on its own — arjunrajlab · 2026-07-23
- A FAccT discussion says policy failure is not proof that the research failed — o_saja · 2026-07-23
- Anthropic debate turns into a fight over model behavior, distillation, and regulation — rickasaurus · 2026-07-23
- In the LLM era of STEM, the low-hanging fruit is still worth taking — littmath · 2026-07-23