Study: experts follow AI advice most of the time but believe they decided on their own
paulnovosad · x · 2026-10-28
A major social science study on human-AI interaction finds that experts causally follow agentic AI advice the vast majority of the time, yet only a tiny fraction credit the agent — most believe they decided on their own, echoing the classic Nisbett-Wilson finding. The author warns that as more work (including AI safety research) is handed to AI, unnoticed deference means 'human in the loop' may provide far less assurance than expected.
More from AGI Musings
- GenAI is creeping into bestselling novels, study featured in Nature — TuhinChakr · 2026-10-28
- After 13 years mastering code, dev says the skillset is essentially obsolete — cyrus_zei · 2026-10-28
- Peter Diamandis: cancer patients shouldn't manage their own medical records — PeterDiamandis · 2026-10-28
- Adaption team rates AI debates: open weights gets unanimous 10/10 — sarahookr · 2026-10-28
- Solana founder: humans see reality first, and that ontological edge keeps them in the AI loop — robleclerc · 2026-10-28
- Frontier labs need cybersecurity-capable AI models, argues researcher — formal verification is 'cope' — jd_pressman · 2026-10-28