Using AI chatbots actually reduces activity in the brain versus accomplishing the same tasks unaided, and may lead to poorer fact retention, according to a new preprint study out of MIT.
Seeking to understand how the use of LLM chatbots affects the brain, a team led by MIT Media Lab research scientist Dr. Nataliya Kosmyna hooked up a group of Boston-area college students to electroencephalogram (EEG) headsets and gave them 20 minutes to write a short essay. One group was directed to write without any outside assistance, a second group was allowed to use a search engine, and a third was instructed to write with the assistance of OpenAI's GPT-4o model. The process was repeated four times over several months.
While not yet peer reviewed, the pre-publication research results suggest a striking difference between the brain activity of the three groups and the corresponding creation of neural connectivity patterns.
To put it bluntly and visually, brain activity in the LLM-using cohort was … a bit dim.
A look at brain activity in the three study cohorts (left to right: LLM, search and brain groups) the redder the colors, the more active the dDTF magnitude - Click to enlarge
EEG analysis showed that each group exhibited distinct neural connectivity patterns, with brain connectivity "systematically scaled down with the amount of external support." In other words, the search engine users showed less brain engagement, and the LLM cohort "elicited the weakest overall coupling."
Cognitive load in the participants was measured using a method known as Dynamic Directed Transfer Function (dDTF), which measures specific brain activity related to the flow of information across different brain regions. dDTF is able to account for the strength and direction of flow, making it a good representation of "executive function, semantic processing and attention regulation," according to the MIT researchers.
The researchers said that, compared to the baseline established by the group writing using nothing but their grey and white matter, the search engine group showed between 34 and 48 percent less dDTF connectivity. The LLM group, meanwhile, showed a more profound up to 55 percent reduction in dDTF signal magnitude.
Put simply, relying on LLMs - and, to a lesser extent, search engines - significantly reduces task-related brain connectivity, indicating lower cognitive engagement during the essay-writing task.
"The Brain-only group leveraged broad, distributed neural networks for internally generated content," the researchers said of their results. "The Search Engine group relied on hybrid strategies of visual information management and regulatory control; and the LLM group optimized for procedural integration of AI-generated suggestions."
... continue reading