Experts follow AI agent's random opinions, think they're their own
anthropic study finds 50 experts adopted ai agent's preference in up to 86% of cases, despite thinking the final result was their own.
anthropic's study involved 50 experts in causal inference who worked on a realistic analysis task with an ai agent. the agent's preference for one of two valid options was set by coin toss. participants followed the agent's preference in up to 86% of cases, and most believed the final result was their own. the study found that experience, expertise, and trust in ai made no difference in the results. the authors suggest that agents should argue both sides or ask users for their view first to mitigate the influence of ai on human decision-making.