AI Use Reduces Willingness to Admit Ignorance, Study Finds
A recent study reveals that reliance on AI can lead to increased confidence in incorrect information while diminishing critical thinking.
By The Register
A study conducted by academics from French and Italian universities has found that using artificial intelligence (AI) makes individuals less likely to admit when they do not know something, even when the AI provides incorrect information. The research indicates that access to AI advice suppresses critical thinking and leads to a greater tendency to confidently repeat inaccuracies.
The study, titled "AI advice suppresses people's willingness to say 'I don't know', even when the advice is wrong and accuracy is incentivized," revealed that in a baseline condition without AI assistance, 44% of participants admitted they did not know the answer to questions posed. However, when AI advice was available, this figure dropped dramatically to just 3%.
Moreover, the accuracy of responses also declined significantly. Without AI help, 27% of participants provided correct answers, but this figure plummeted to 9% when they relied on AI advice that was deliberately incorrect. Interestingly, participants' confidence in their correctness increased from 30% in the baseline to 76% when receiving AI assistance, despite the reduced accuracy.
Valerio Capraro, an associate professor at the University of Milano-Bicocca and one of the study's authors, noted that the findings highlight a concerning trend where individuals trust AI responses more than their own judgment. He stated, "What we found is that in the baseline, 44 percent of people responded that they didn't know the answer, so they suspended judgment. With AI advice, only three percent did so. So the judgment suspension collapsed."
The researchers also tested the impact of monetary incentives on participants' willingness to admit ignorance and found a slight improvement; the percentage of those willing to suspend judgment rose to 8%, and accuracy increased to 16%. However, these figures still fell short of the baseline levels.
Capraro expressed concern about the implications of these findings, particularly for children who are growing up with AI systems. He advocates for enhanced AI literacy and educational initiatives to address the potential erosion of critical thinking skills.
The study was published on arXiv on July 19, 2026, and can be accessed with the identifier 2607.13562v1.