A new pre-print study from the US-based Massachusetts Institute of Technology (MIT) found that using OpenAI’s ChatGPT could lead to cognitive decline.
Researchers with the MIT Media Lab broke participants into three groups and asked them to write essays only using ChatGPT, a search engine, or no tools.
Brain scans were taken during the essay writing with an electroencephalogram (EEG) during the task. Then, the essays were evaluated by both humans and artificial intelligence (AI) tools.
The study showed that the ChatGPT-only group had the lowest neural activation in parts of the brain and had a hard time recalling or recognising their writing. The brain-only group that used no technology was the most engaged, showing both cognitive engagement and memory retention.
The researchers then did a second session where the ChatGPT group were asked to do the task without assistance. In that session, those who used ChatGPT in the first group performed worse than their peers with writing that was “biased and superficial”.
A 'likely decrease' in learning skills
The study found that repeated GPT use can come with “cognitive debt” that reduces long-term learning performance in independent thinking.
In the long run, people with cognitive debt could be more susceptible to “diminished critical inquiry, increased vulnerability to manipulation and decreased creativity,” as well as a “likely decrease” in learning skills.
“When participants reproduce suggestions without evaluating their accuracy or relevance, they not only forfeit ownership of the ideas but also risk internalising shallow or biased perspectives,” the study continued.
The study also found higher rates of satisfaction and brain connectivity in the participants who wrote all essays with just their minds compared to the other groups.
Those from the other groups felt less connected to their writing and were not able to provide a quote from their essays when asked to by the researchers.
The authors recommend that more studies be done about how any AI tool impacts the brain “before LLMs are recognised as something that is net positive for humans”,
