AI advice made people three times less accurate but twice as confident, researchers found
What happened
A study from researchers at three French and Italian universities found that access to AI advice sharply reduced people’s tendency to admit uncertainty. When given AI recommendations, willingness to say “I don’t know” dropped from 44% to just 3%. Accuracy also plummeted from 27% correct answers to 9%. Despite this fall in accuracy, confidence soared from 30% to 76% after consulting AI.
Why it matters
This research exposes a critical risk for anyone relying on AI input to make decisions. Rather than acting as a reality check, AI advice suppressed the natural caution that helps people think critically. The result was a tripling of incorrect answers paired with double the confidence in those wrong answers. For businesses, founders, and operators, this means AI can unintentionally encourage overconfidence and weaken judgment rather than improve it. Trusting AI without skepticism may increase costly errors in hiring, customer support, or strategic choices.
What to watch next
Watch for shifts in AI system design that explicitly encourage users to maintain skepticism and recognize uncertainty. Training and user interfaces will need to evolve so operators do not blindly accept AI outputs. Investors and executives should demand more rigorous testing of AI’s impact on decision quality before deploying it broadly in high-stakes environments. Regulators may also focus on how AI advice influences consumer confidence and error rates, weighing the cost of misplaced trust.
AI Quick Briefs Editorial Desk
