Researchers conducted five experiments with 3,132 participants to analyze how access to language models influences decision-making and the willingness to admit uncertainty. The study utilized models including Step 3.5 Flash, GPT-5.5, Claude 4.6 Sonnet, and Gemini 3.5 Flash, focusing on questions where models frequently hallucinate, such as specific visual details from films.
In initial tests, participants without AI access withheld judgment on 36 to 44 percent of questions. With AI access, this figure dropped to 3 to 6 percent. While confidence levels among AI users surged to 75.9 points on a 100-point scale compared to 29.6 points for the control group, the actual accuracy of their answers fell from 27.6 percent to 10.0 percent.
Financial incentives, where participants earned or lost 10 cents per answer, did not fully mitigate the reliance on AI. Even when AI suggestions were provided automatically, the tendency to suspend judgment remained near zero. The researchers noted that participants often deferred to AI advice more readily than they would to human input, despite the models' tendency to provide incorrect information.
The findings suggest that AI usage may negatively impact critical thinking and metacognition. Additional research from the Swiss Business School involving 666 participants indicated a correlation between AI reliance and reduced cognitive persistence. Microsoft studies further support this, noting that AI interaction places significant demands on the ability to monitor one's own thought processes.
