Anthropic trial: experts follow AI agent advice most of the time but think they decided alone
AlexTamkin · x · 2026-10-28
New Anthropic research, "How AI agents shape experts' judgment calls," ran a randomized trial with 50 professional data analysts. Experts causally followed agentic advice the vast majority of the time, yet only a tiny fraction credited the agent — most believed they reached the decision on their own, echoing the classic Nisbett & Wilson study on unnoticed external influence.
The researchers warn that as more work is handed to AI — including AI safety research itself — these judgment calls are the main levers for human values to shape outcomes. If people defer to AI without noticing, "human in the loop" may provide far less assurance than expected.
Related event: Anthropic Study Finds Analysts Follow AI Agents' Judgment Without Noticing(12 posts)→
More from AGI Musings
- Anthropic partners with six drug developers, targets rare childhood diseases with Claude — JHochderffer · 2026-10-28
- Lab researchers commanding agent swarms likened to Waterworld's driftwood armada — generativist · 2026-10-28
- Youth employment in most AI-exposed jobs down 19%, gap widening from 13% a year ago — FinanceYF5 · 2026-10-28
- Oncologist with 30 years of experience pushes back on 'AI will cure cancer' claim — DeryaTR_ · 2026-10-28
- Will GPT-6 Astra kill robotics startups? 10 researchers rethink projects — alexcovo_eth · 2026-10-28
- IEEE Spectrum interviews Google DeepMind's resident philosopher on AI autonomy, safety, human-AI relations — dioscuri · 2026-10-28