A new study from the Massachusetts Institute of Technology has found that relying too heavily on chatbots can diminish critical thinking skills and potentially reduce a person’s ability to independently discern misinformation, adding to a growing body of research raising concerns about cognitive dependency on Artificial Intelligence tools.
The four-week study, released in April, tracked 67 participants who were quizzed on whether pairs of news-related headlines and images were real, with researchers comparing performance when participants used an Artificial Intelligence assistant running on GPT-4o integrated with Google Search against performance when working unaided. The findings revealed a clear trade-off: when assisted by the chatbot, participants showed a 21 percent higher chance of correctly identifying real versus fake content, yet their unassisted performance, when reviewing new images without the AI’s help, grew 15.3 percent worse by the fourth week of the study. The chatbot proved capable of guiding users toward accurate judgments, in one case advising a participant to examine a police badge more closely to identify a manipulated image, but the researchers found that this assistance often prioritised giving an accurate response over cultivating the user’s own reasoning ability.
Anku Rani, a PhD student at the Massachusetts Institute of Technology and co-lead author of the study, said that when people interact with Artificial Intelligence, they often feel they are becoming better at certain tasks despite research consistently showing otherwise. About one-quarter of participants in the study believed their detection skills were improving even as their actual performance declined, with researchers noting that participants using more prescriptive Artificial Intelligence systems tended to defer to the technology because it sounded knowledgeable, rather than engaging in independent reasoning. The study’s authors found that more nuanced, guided questioning from an AI system, rather than direct prescriptive answers, was more effective at preserving and building genuine critical thinking ability.
The findings sit alongside a broader pattern of research examining how dependence on technology can erode underlying human skills, comparable to how calculators and GPS devices have been shown to reduce capacity for mental arithmetic and independent navigation. A 2025 study published in The Lancet found that doctors using Artificial Intelligence classification tools to detect cancer eventually became worse at identifying cases without that assistance, while a neuroscientist at the Possibility Institute, a metascience research group, has separately warned that diverting too much cognitive effort to Artificial Intelligence systems may weaken the brain’s natural defences against dementia.
The Massachusetts Institute of Technology researchers acknowledged some limitations in their work, noting that participants were drawn predominantly from the United States and United Kingdom, meaning a more culturally and educationally diverse sample would be needed to determine whether similar skill degradation occurs across different populations. They also noted that studies tracking participants for longer than four weeks could help clarify whether the effects of AI over-reliance continue at the same rate over extended periods of use. The researchers said their findings carry particular relevance for educators increasingly incorporating Artificial Intelligence into learning environments, as well as for the wider public navigating an increasing volume of dubious online content spanning news, viral images, medical claims, and political rumours, stating that ensuring such tools build critical thinking rather than cognitive dependency is essential for maintaining public resilience to misinformation.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.
