Using AI bots like ChatGPT could be causing cognitive decline, new study shows

Published connected 21/06/2025 - 9:00 GMT+2•Updated 11:50

A caller pre-print study from nan US-based Massachusetts Institute of Technology (MIT) recovered that utilizing OpenAI’s ChatGPT could lead to cognitive decline. 

Researchers pinch nan MIT Media Lab broke participants into 3 groups and asked them to constitute essays only utilizing ChatGPT, a hunt engine, aliases nary tools. 

Brain scans were taken during nan effort penning pinch an electroencephalogram (EEG) during nan task. Then, nan essays were evaluated by some humans and artificial intelligence (AI) tools. 

The study showed that nan ChatGPT-only group had nan lowest neural activation successful parts of nan encephalon and had a difficult clip recalling aliases recognising their writing. The brain-only group that utilized nary exertion was nan astir engaged, showing some cognitive engagement and representation retention. 

The researchers past did a 2nd convention wherever nan ChatGPT group were asked to do nan task without assistance. In that session, those who utilized ChatGPT successful nan first group performed worse than their peers pinch penning that was “biased and superficial”.

A 'likely decrease' successful learning skills

The study recovered that repeated GPT usage tin travel pinch “cognitive debt” that reduces semipermanent learning capacity successful independent thinking. 

In nan agelong run, group pinch cognitive indebtedness could beryllium much susceptible to “diminished captious inquiry, accrued vulnerability to manipulation and decreased creativity,” arsenic good arsenic a “likely decrease” successful learning skills. 

“When participants reproduce suggestions without evaluating their accuracy aliases relevance, they not only forfeit ownership of nan ideas but besides consequence internalising shallow aliases biased perspectives,” nan study continued. 

The study besides recovered higher rates of restitution and encephalon connectivity successful nan participants who wrote each essays pinch conscionable their minds compared to nan different groups. 

Those from nan different groups felt little connected to their penning and were not capable to supply a quote from their essays erstwhile asked to by nan researchers. 

The authors urge that much studies beryllium done astir really immoderate AI instrumentality impacts nan encephalon “before LLMs are recognised arsenic thing that is nett affirmative for humans”,