Making AI Headlines · Page 7
Experts Follow a Coin's Advice and Call It Their Own
· New York time
In an Anthropic study, 50 experts used an AI agent whose preferences were set by a coin toss. On judgment calls, the experts followed the agent's view up to 86 percent of the time, and most believed the conclusions were their own, according to the study. The result focuses on a subtle form of influence. The agent's preference was assigned randomly, but it was presented through an AI system. The study reports that experts often adopted its view even when the preference itself had been determined by a coin. The figure is an upper reported rate for judgment calls in the study, not a claim that every expert followed the agent 86 percent of the time or that all professional decisions work this way. The group included 50 experts, so its results are evidence from that study rather than a census of everyone who uses AI. Anthropic is itself an AI company, giving it a direct interest in how people understand and use agents. The finding raises a question about a tool that can present a position fluently: an answer can influence a judgment even when the preference behind it was assigned at random. The experts brought professional knowledge to the exercise. The coin brought two sides. In the study's most striking result, the coin still managed to become a remarkably successful colleague.
Read this story in the full newspaper →Satire. Not factual reporting. Read the original reporting from the linked publishers.