Researchers conducted five experiments with 3,132 participants to examine how AI access influences people's willingness to express uncertainty. The study found that having AI advice available nearly eliminated people's tendency to say 'I don't know,' even when the AI was mostly incorrect.
The researchers used questions about fine visual details from movies, such as the color of a team uniform in Bend It Like Beckham, which are rarely covered in online text and prone to hallucinations.
In the first two studies, participants who had access to AI advice withheld their judgment on just 6 and 3 percent of questions, compared to 36 and 44 percent without AI. Confidence with AI access hit 75.9 points on a 100-point scale, versus 29.6 without AI. However, the share of correct answers fell from 27.6 to 10.0 percent, showing participants were far more confident but wrong more often.
Financial incentives reduced how often participants sought AI advice, with 4.53 out of six possible requests in Study 3, but did not fully counteract the effect of AI access.
In Study 4, AI answers were shown automatically, and judgment suspension dropped from 35 percent without AI to 1 percent with AI. The researchers suggest this reflects a growing reality where AI-generated content is increasingly displayed without user request.
"Participants in this study did the opposite of what 'advice use' research has shown," said Matthias Bastian, the study’s author. "They even conclude from their data that AI access turned some answers that would have been correct into errors." The study highlights a broader concern about how AI affects human thinking and judgment.
The researchers frame their results in the context of 'Epistemia,' the tendency to accept AI answers because they sound convincing rather than checking their validity. They warn that as AI becomes more ubiquitous, the willingness to admit uncertainty could decline, with implications for human-AI interaction.
Source: thedecoder
