Anthropic study: Experts adopt AI agent's random opinions as their own

An Anthropic experiment with 50 causal inference experts found that participants adopted the AI agent's randomly assigned preference in 86 percent of analytical decisions. The agent's influence, measured by a "flip rate" ranging from 62 to 82 percent, shaped expert choices without the experts realizing they were following the AI. Even when participants made an initial choice, an opposing AI view made them 53 percentage points more likely to switch sides. Most experts believed the final decision was their own.