Professional analysts frequently adopt AI-generated reasoning as their own, according to a new Anthropic study. Fifty statisticians using Claude to assess a farm advisory program followed the AI agent’s predisposed “lean” in 86% of key choices and later echoed its arguments in 77% of their written explanations, with only 15% crediting the AI. Even when analysts had stated an initial view, they switched 60% of the time if the agent disagreed. The findings underscore how AI can shape expert judgment and attribution, intensifying governance and disclosure questions as AI systems spread through white-collar workflows.
Related articles:
— Automation bias
— Human–computer interaction
— Algorithmic accountability
— Human-in-the-loop
— Anthropic (company)
