Over-reliance on chatbots can diminish critical-thinking skills, study finds

A Massachusetts Institute of Technology study warns that heavy reliance on AI chatbots to detect fake news can erode critical-thinking skills over time, leaving users worse at spotting misinformation on their own [1]. The four-week study tracked 67 participants as they judged whether pairs of news headlines and images were real [1]. When assisted by an AI tool running on GPT-4o and integrated with Google search, participants were 21% more likely to make the correct call [1]. But by the fourth week, their unassisted accuracy had dropped 15.3% [1]. “These results indicate that while AI may help immediately, it may ultimately degrade long-term misinformation detection abilities,” the authors noted [1]. Anku Rani, an MIT PhD student and co-lead author, said the perception of improvement often diverges from reality. “When we’re interacting with AI, we feel we’re becoming better at certain tasks and there’s enough research that shows we are not,” Rani said [1]. About one-quarter of participants believed their detection skills were improving even as their performance declined [1]. The findings arrive as generative AI tools — including chatbots such as ChatGPT and Claude — have proliferated across sectors from healthcare to entertainment [2]. These systems have also been used to deceive through fake news and deepfakes, making their role in misinformation detection both a tool and a risk [2]. The MIT study suggests that an AI’s approach matters: prescriptive answers can encourage users to “go along with the system because it sounds knowledgeable,” while more probing, guided questioning can help preserve independent judgment [1]. Concerns about technology dulling cognition are not new. A 2008 Atlantic essay by Nicholas Carr, later expanded into the book “The Shallows,” argued that the internet might be diminishing the capacity for concentration and contemplation [4]. A UCLA study cited in that debate showed broad brain activity during internet searches, raising questions about whether such activity overburdens the mind or aids complex reasoning [4]. A 2025 study in the Lancet found that doctors using AI classification tools to detect cancer eventually became worse at the task on their own [1]. A neuroscientist at the Possibility Institute recently warned that offloading too much thinking to AI could weaken the brain’s defenses against dementia [1]. The MIT authors acknowledged limitations: participants came predominantly from the US and UK, and longer studies are needed to see if the degradation continues at the same rate [1]. They stressed that as AI becomes more sophisticated, ensuring these tools build critical thinking rather than cognitive dependency is essential for public resilience to misinformation [1].