Your Brain (and Brief) on AI

Recently, MIT’s Medial Lab released a report on how AI impacts the brains of those who use it to write. In that study, researchers monitored brain activity in participants who were asked to write SAT essays over a period of four months. One group used AI, one group used Google search, and one group just used their own brain. The results were so dramatic that the lab released their study before full peer review, because they were concerned that students were being taught to use AI in a way that could be detrimental to their development.

The study showed that those who used AI for every essay started by using it primarily for organizational or proofing tasks. As time went by, they became more dependent on it for more and more functions, until they tended to just cut-and-paste directly from the result into the document. Their brain development showed a lack of attention and low executive control. And when asked later about their essays, they had a hard time recalling what they had written about.

The group that only used their brain, however, showed development in their brain through their work. They showed heightened activity in regions of the brain associated with creativity, memory and language processing. The group that “just” used their brains also expressed more satisfaction and ownership of their work.

Perhaps most interesting, from a legal writing perspective, was the fact that those who used AI became “disassociated” from their results. They felt less assurance about what was “their” work, and less interest in re-engaging with it. Those who had used “only” their brains to generate an essay, however, could later re-engage with it using AI without a decline in brain activity relating to creativity, language, and engagement.

This study comes at a time when lawyers are making headlines for not checking their work. They are using AI to generate briefs and memos with hallucinated cases, and submitting them to courts. Perhaps the study helps us understand how this is happening – they do not have ownership of their work. Not only have they not done the actual research, but they haven’t integrated the lessons from that research into their own brains in a way that helps them think critically about their writing. And since they have low confidence in their own knowledge, they are less likely to edit and check the work.

While this is an early study with a low sample, the results suggest that AI has a place in legal writing, just not at the start of the process. Traditional research tools require struggle and ownership of the results, which in turn sparks critical thinking and self-editing. Adding in AI tools after the first draft can help deepen engagement and improve the final product. But using AI as an initial tool can hinder the process.