Over-reliance on chatbots can diminish critical-thinking skills, study finds
A new MIT study shows that relying too heavily on AI assistants like ChatGPT can weaken people's ability to independently detect misinformation.
A study from the Massachusetts Institute of Technology has found that excessive dependence on chatbots can impair critical thinking and the ability to discern truth from falsehood. The four-week study, published in April, involved 67 participants who were asked to determine whether news headlines and images were real.
Participants completed tasks both with and without the help of an AI assistant based on GPT-4o integrated with Google Search. Researchers found that the AI assistant increased the likelihood of correctly identifying misinformation by 21%, but participants' independent performance without AI support worsened by 15.3% by the fourth week.
Co-lead author Anku Rani noted that people often feel more skilled when interacting with AI, even though they are not. About a quarter of participants believed their skills were improving, even as their performance declined.
The study highlights that the approach of an AI system—whether prescriptive or probing—affects users' ability to maintain good judgment. When AI simply tells users what to do, they tend to follow it blindly because it sounds knowledgeable.
The study has limitations: participants were mostly from the US and UK, and four weeks may not be enough to assess long-term effects. However, the findings are important for education, where AI tools are increasingly used, and for the general public facing an abundance of dubious online information.
