Researchers Studied How Using ChatGPT Affects the Human Brain
The results weren't great
Hiya!
Artificial Intelligence (AI) has quickly gone from a purely science-fiction concept to an everyday science-nonfiction reality of our modern society. In the last few years, we’ve seen major leaps with AI, especially once systems like ChatGPT became available to the public.
Now, amongst the many other subjects vying for attention, we find ourselves increasingly debating AI’s influence over us and the roles it should play in our society.
On the one hand, it’s thought that AI will further elevate our knowledge and help us become better learners and workers, enabling us to achieve monumental feats that we couldn’t accomplish on our own. On the other hand, there are valid concerns that AI could be used for more nefarious purposes, such as invading our privacy and making corruption easier.
Meanwhile, some scientists are more interested in the physical influences AI has on us. For instance, researchers sought to gain a deeper understanding of how using AI affects our brains compared to not using it, and what they found was less than encouraging.
Debate About AI
ChatGPT was launched to the public on January 30, 2023, and its influence and controversy have grown ever since. One of the most significant debates centers on the potential impact of AI on learning.
For instance, does using AI for academic purposes progress our ability to personalize education and provide greater access to knowledge? Or does AI offer an easy path to cheating and academic dishonesty, along with an overall “dumbing down” of society as people increasingly rely on AI and thus cease developing critical thinking skills?
After all, there’s already a highly concerning rise in the rate of illiteracy, especially regarding reading comprehension, among Americans of all ages. So it seems reasonable to see a connection between students using AI too early in schools and not developing basic critical thinking and problem-solving skills.
Thankfully, scientists at the Massachusetts Institute of Technology (MIT) Media Lab are already devoting significant resources to investigating various effects generative AI tools can have on us.
For instance, research announced in March 2025, involving a collaboration between scientists at the MIT Media Lab and OpenAI, reported a link between the time spent interacting with ChatGPT and feelings of loneliness. That, generally speaking, the more time people spend engaging with ChatGPT, the lonelier they feel.
In a more recent paper, researchers at the MIT Media Lab took things further by documenting striking differences in brain activity between individuals who used ChatGPT to write and those who did not. The study is awaiting peer review, but, as I’ll discuss later, the researchers have a sensible reason for announcing their results now.
The Research
After noticing the increase in students using AI, Nataliya Kosmyna, a full-time research scientist at the MIT Media Lab with a background in artificial intelligence, neuroscience, and human-computer interaction, became curious about the impacts of using AI for schoolwork on students’ brains. So, Kosmyna and her colleagues decided to find out.
They recruited 54 participants, aged 18 to 39 years, who resided in the Boston area. Over four months, the participants were asked to write three essays (each with a 20-minute time limit) based on SAT prompts, including the downsides of having too many choices and the ethics of philanthropy.
For instance, one of the questions under the “Art” category was: “Do works of art have the power to change people’s lives?”
While all participants were given 20 minutes to write their essays, the researchers divided them into three groups.
One group, referred to as the Large Language Model Group, or LLM group, was instructed to write their essays using only OpenAI’s ChatGPT, without referring to any other website or tools, including search engines like Google.
However, the second group, called the Search Engine group, could use Google and any other website or tools to write their essays, except for those with Large Language Models (LLM), including ChatGPT.
Meanwhile, the last group, known as the Brain-Only group, wrote their essays using only their current knowledge, without any access to the internet or AI.
While writing their essays, all participants wore a wireless electrophysiology sensor system that recorded their brain’s electrical activity, or EEG signals, across 32 brain regions. This allowed the researchers to assess the participants’ cognitive load and engagement, thereby gaining a better understanding of their brain activity during the writing task.
The participants answered questionnaires in addition to being interviewed by the researchers after each session. The researchers then subjected the essays to a Natural Language Processing (NLP) analysis in addition to having them evaluated by human teachers and an AI judge.
Additionally, after the three sessions, the researchers called back a total of 18 participants from the LLM and Brain-only groups for a fourth round. Except this time, they swapped the groups to measure any possible enduring effects from the first three sessions, and instead of writing a new essay, the participants re-wrote one of their previous ones from scratch.
For the fourth round, participants who were initially assigned to the LLM group were switched to the Brain-Only group and instructed to write their essay using only their brain and current knowledge. Meanwhile, the initial Brain-Only group was asked to use only LLMs, including ChatGPT, to write their fourth essays.