This is your brain on ChatGPT.
I know, with a title like that you may be expecting me to show you a horrifying picture intended to scare you away from using LLMs, claiming that your brain will turn into this:
Photo by Andreas Haslinger on Unsplash
A recent M.I.T. study has found that using A.I. tools during essay-writing assignments leads to less creativity and less brain connectivity when compared to using traditional search engines or using no tools at all. A Swiss study has shown that more A.I. usage among young people seems to correlate with reduced critical thinking. In the M.I.T. study, researchers split participants into three groups– one group using LLM, one group using search engines, and one group using only their own brains– and instructed them to write essays on certain topics. Participants’ neural connectivity patterns were measured using EEG during the experiment. The brain-only group exhibited the strongest neural connections while the search engine group had the next strongest, and the LLM group had the weakest connections. Higher neural connectivity suggests better recall and more mastery of the topic. This is corroborated by the Swiss study which found that groups that reported greater incidence of A.I. usage also exhibited fewer critical thinking skills due to higher “cognitive offloading,” or the act of externalizing cognitive processes to reduce mental effort.
So, does A.I. usage lead to raisin brain? Not quite. Of course using A.I. requires less perceived effort than other research methods– there is a reason why it is so tempting to use and challenging to resist. However the M.I.T. researchers highlight another important finding as well. For the final round of the experiment, the brain-only group was then allowed to use LLMs to research one of the topics they had previously written about. That group exhibited much higher neural connectivity when using LLMs compared to the group that had used LLMs in every round of the experiment, suggesting that using LLMs to research a topic one is already familiar with can have beneficial effects for the brain. In other words, knowing the correct ways to prompt LLMs on a topic can result in a better essay and a better understanding of the topic overall.
Similarly, the Swiss researchers concluded that A.I. usage can lead to enhanced learning outcomes only if it is carefully balanced with critical thinking and cognitive engagement tasks. For example, students should engage with scholarly articles, take notes, and participate in class discussions on a topic before using LLMs to aid with their research on the topic. A.I. can be viewed as one tool in the toolbox instead of the whole solution.