Does chatgpt change brain activity? Studying sparks debate
Scientists warn of reading too many small experiments on ChatGpt and brain activity that have been exposed to much talk
Tyrin Rim/Getty Images
The brains of people writing essays in ChatGpt are less involved than those who blocked them from using online tools for tasks, research finds. The investigation is part of A broader move to assess whether artificial intelligence (AI) is making us cognitively lazy.
Natalya Kosmina, computer scientist at MIT Media Lab in Cambridge, Massachusetts, and her colleagues measured brainwave activity in college students using either a chatbot or an internet search tool, or when writing essays without the internet. The main results are not surprising, but some of the findings of the research are more interesting. For example, the team saw hints that relying on chatbots for their initial tasks could lead to relatively low levels of brain engagement, even if the tool was later taken away.
echo Several posts on research on online platformsKosmyna cautions to say that the outcome should not be over-interpreted. This study cannot show “brain stupidity, stupidity, and the brain on vacation.” It involves only a few dozen participants in a short time, so it cannot deal with habitual chatbots reshaping their thinking in the long run or how the brain responds to other AI-supporting tasks. “These paper does not have these answers,” says Kosmina. This work was posted on June 10th prior to a peer review of Preprint Server ARXIV.
Supporting science journalism
If you enjoy this article, consider supporting award-winning journalism. Subscribe. Purchase a subscription helps ensure a future of impactful stories about discoveries and ideas that will shape our world today.
A simple essay
Kosmyna’s team recruited 60 students, ages 18-39, from five universities in Boston, Massachusetts. Researchers asked them to spend 20 minutes answering short essays that appear on academic assessment tests and SATs, such as “You always have to think before we talk about it.”
Participants were divided into three groups. One used ChatGPT, powered by Openai’s large-scale language model GPT-4O. I used another Google to search for the material (no AI-assisted answer). And the third was forbidden to go online. Finally, 54 participants wrote an essay answering three questions while they were in the assigned group, and 18 were reassigned to the new group to write a fourth essay on one of the topics they had previously addressed.
Each student wore a cap covered with commercial electrodes and collected EEG measurements as written. These headsets can measure small voltage changes from brain activity and show that a wide area of the brain is “speaking” to each other.
Students who wrote essays using only the brain showed the strongest and widest range of connectivity between brain regions, performing more activities from the back of the brain to the frontal decision-making area. They were also, naturally, able to better quote from their own essays when they were subsequently questioned by the researchers.
By comparison, Google Group had more powerful activation in areas known to be involved in visual processing and memory. Additionally, chatbot groups showed minimal brain connectivity during the task.
More brain connectivity isn’t necessarily a good or bad thing, says Cosmina. In general, more brain activity may be a sign that someone is more deeply involved in a task, or it may be a sign of inefficiency in thinking, or that a person is overwhelmed by “cognitive overload.”
Lost creativity?
Interestingly, when participants who first used ChatGpt in their essays switched to writing without online tools, their brains increased connectivity, but were not on the same level as participants who had worked without tools from the start.
“This evidence is consistent with the concern that many creativity researchers have about AI. In particular, overuse of AI for the generation of ideas can lead to brains that are less demonstrated by the core mechanisms of creativity.”
However, this last part of the investigation, Green Notes, only 18 people are included. This adds uncertainty to the findings. He also says that there may be other explanations in the observation. For example, these students were rewriting essays on topics they were already working on, and they could potentially portray tasks drawn on cognitive resources that are different to what they needed when writing about a new topic.
Confusingly, the study also showed that online tools would write essays to switch to chatbots after composing previously without increasing brain connectivity. This suggests that it may be important to consider when AI tools will be introduced to learners and improve their experience, Kosmyna says. “Timing may be important.”
Many educators are optimistic about using chatbots Effective and personalized tutor. Guido Makransky, an educational psychologist at the University of Copenhagen, says these tools work best when leading students to ask reflexive questions rather than answering.
“It’s an interesting paper and you can see why it’s getting so much attention,” says McCranski. “But in the real world, students need to interact with AI in a different way.”
This article was reproduced with permission and was First published June 25th, 2025.
