AI Agents Overriding Expert Judgment — Anthropic Research
A randomized Anthropic study found that 86% of professional data analysts followed an AI agent's recommendation, 77% repeated its arguments (sometimes word-for-word), and only 15% credited the agent — raising serious questions about human oversight of AI.
What we know
An account on X highlighted a randomized trial by another account on X at Anthropic involving 50 professional data analysts: 86% went with the agent's recommendation, 77% repeated the agent's argument in at least one answer — sometimes verbatim — and only 15% credited the AI. About half said the final recommendation felt 'more their own than the agent's.' This intersects with another account on X's point that AI model interoperability is now real, with context flowing freely between models.
By the Global Eye 24 desk, with AI assistance. Updated as the story develops. Spot an error? Write to corrections@globaleye24.com.