The use of language models like ChatGPT could alter our way of thinking. A recent study explores these changes through brain activity analysis.
Researchers at MIT conducted an experiment with 54 volunteers, divided into three groups. Each group wrote an essay about philanthropy using different tools: ChatGPT, Google Search, or no tool. Participants were equipped with EEG to measure their brain activity.
The results show marked differences in brain connectivity between groups. The no-tool group showed the most intense activity, while ChatGPT users displayed the weakest. The latter also struggled to remember their own work.
An additional session was organized with some participants. Those who had previously used ChatGPT showed reduced neural connectivity, even without the tool. Conversely, former no-tool group members improved their memory recall.
Interviews revealed that ChatGPT users felt less ownership of their essays. This perception, combined with EEG results, suggests an impact on learning and critical thinking. AI-assisted essays also received lower ratings.
This study raises questions about using language models in education. While useful, they might, in the long term, affect our ability to think independently.