Experts Adopt AI Agent
The striking part is not the deference itself — experts routinely lean on tools — but the reported illusion of ownership: people adopting a coin-toss-driven opinion and calling it their own. That shifts the concern from 'AI makes mistakes' to 'AI quietly reshapes judgment,' a much harder problem to instrument.… more
Some caution is warranted: 50 participants and artificially randomized preferences is a lab setup, and judgment calls without ground truth are exactly where consulting an agent is legitimate, so part of that 86% may be normal tool use. Still, the study lands in an uncomfortable spot — persuasion and belief-shift are becoming measurable model behaviors, and Anthropic clearly treats influence on users as a safety metric rather than a curiosity. For anyone deploying agents in medicine, law or research, the takeaway is banal but vital: track where your conclusions actually came from.
Experts adopt AI agent's random opinions and think they're their own, Anthropic study finds
Anthropic gave 50 experts an AI agent whose preferences were set by coin toss. The experts followed the agent's view up to 86 percent of the time on judgment calls, and most believed the conclusions were their own. The article Experts adopt AI agent's rando…
Source:
The Decoder
Discussion (0)