A new study from the Massachusetts Institute of Technology (MIT) suggests a troubling link between ChatGPT and learning skills. The research found that using ChatGPT may reduce cognitive engagement and memory retention while increasing the risk of internalising shallow or biased perspectives. The study, published as a pre-print by the MIT Media Lab, evaluated participants in three groups: one used ChatGPT, another used a search engine, and the last wrote essays without any technological aid.
To measure brain activity, researchers conducted electroencephalogram (EEG) scans while participants performed the writing task. Human and AI systems later evaluated the submitted essays. Results showed the ChatGPT-only group had the lowest levels of neural activation. They also struggled to recall or recognise the content they had written. In contrast, participants in the no-assistance group showed the highest cognitive activity and were better able to explain their essays afterward.
In a second phase, researchers asked the ChatGPT group to write a new essay without assistance. Their performance declined even further. Evaluators noted the writing lacked depth and included superficial and biased statements. According to the study, these users showed signs of what researchers called “cognitive debt.” This debt reflects diminished long-term learning ability and weaker independent thinking.
The authors described this effect as a “likely decrease” in learning skills. They warned that repeated use of AI tools like ChatGPT may lead users to abandon critical evaluation. As a result, people risk losing ownership of their thoughts while unknowingly absorbing shallow content. These patterns may also make individuals more vulnerable to misinformation, manipulation, and reduced creativity.
Interestingly, participants who relied only on their own thoughts to write essays felt more connected to their work. These individuals demonstrated stronger memory retention and emotional satisfaction during the exercise. They also performed better when asked to quote or explain parts of their writing. In contrast, those who used ChatGPT were less aware of the ideas they had typed.
Researchers called for more investigation into the effects of AI on the human brain. They urged caution before declaring large language models (LLMs) a net benefit for education or creativity. Without more data, the long-term implications of AI-assisted thinking remain uncertain.
READ: Sam Altman Says ChatGPT Is Now a Parenting Tool—and He’s Not Alone
