A new study from the Massachusetts Institute of Technology’s Media Lab has sparked concern among educators and researchers, suggesting that frequent use of ChatGPT may be diminishing users’ critical thinking abilities.
The study, led by MIT research scientist Nataliya Kosmyna, analyzed the behavior and brain activity of 54 participants aged 18 to 39 from the Boston area. Divided into three groups, participants were asked to write SAT-style essays using either OpenAI’s ChatGPT, Google Search, or no digital assistance. Brain activity was monitored using EEG across 32 regions.
The results were stark: those who used ChatGPT demonstrated the lowest levels of brain engagement, and their performance lagged behind others at neural, linguistic, and behavioral levels. Over time, the AI-assisted group relied increasingly on copy-and-paste tactics, often outsourcing the majority of the writing task to the chatbot.
“The ChatGPT users showed a sharp drop in mental effort and originality,” Kosmyna explained. “By the third essay, many of them weren’t even trying to write—just refining what the AI gave them.”
Though the study has not yet undergone peer review and its sample size is limited, Kosmyna felt an urgent need to share the findings early. “What really motivated me to put it out now is fear that in six to eight months, a policymaker might say, ‘Let’s introduce GPT in kindergarten,’” she said. “Developing brains are at the highest risk.”
The paper warns that growing dependence on large language models (LLMs) like ChatGPT could hinder learning, especially among young users still developing core cognitive skills.
In addition to EEG data, researchers observed that the ChatGPT-written essays were notably uniform and lacked original thought. Two independent English teachers who reviewed the submissions described them as “soulless” and formulaic, relying heavily on repeated phrases and predictable structure.
This research builds on earlier MIT Media Lab studies into the psychological impacts of generative AI. Previous findings linked prolonged ChatGPT use with increased feelings of loneliness and reduced self-driven idea generation.
As AI tools become increasingly integrated into education, Kosmyna and her team urge caution. “We’re not saying never use AI,” she said. “But if students stop engaging with ideas and just ask the machine to do it for them, we risk trading short-term ease for long-term intellectual decline.”
