MIT Study: Using ChatGPT Won’t Make You Dumb (Unless You Do It Wrong)

A nuanced AI study, you’ve got to love it!

MIT has published a study on how using large language models (LLMs) like ChatGPT affects your brain and your cognitive capacity. It’s 200 pages — I’ve read it, so you don’t have to.

The findings are relevant and surprisingly nuanced. The authors’ criticism of LLMs — because this is a paper criticizing the abuse of LLMs in educational settings — is supported and qualified by the findings — there’s no black or white here.

The most salient aspect of the study is that it uses a brain monitoring method (electroencephalography, EEG) that directly measures brain activity. This is critical because it doesn’t depend on behavioral data or self-report (what people say they think or feel), which generally suffers from dubious validity.

This is your brain telling you what happens when you abuse ChatGPT.

I. Experimental design: essay writing

To give you a mental image of the study, let me start with the experimental design.

The authors gathered 54 people and put them into three groups: LLM group, search engine group (I will leave them aside), and brain group. Each participant was allowed to use…