Your Brain on ChatGPT: A Warning About AI-Induced "Cognitive Deficiency"
Image: AI-generated in Freepik.com based on a pikisuperstar illustration.
THE MEDIA
The recently-released pre-print paper, “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task” is capturing the headlines = ChatGPT & Cognitive Laziness: Are We Losing Mental Edge?, Is ChatGPT Dumbing Us Down? MIT Study Says Yes, MIT Study Says ChatGPT Can Rot Your Brain. The Truth Is A Bit More Complicated, and MIT Study Warns of Cognitive Decline with LLM Use.
While the MIT study doesn’t argue ChatGPT is “rotting our brains,” it does illustrate that LLMs reduce cognitive effort and shift mental processing - raising key questions for educators. This is especially important as business leaders are now calling for mandatory K-12 AI instruction (Over 250 CEOs sign open letter supporting K-12 AI and computer science education; Tech CEOs Call for Mandatory AI Education in U.S. Schools to Secure Future Workforce).
So, how do these finding shape our understanding AI-assisted learning?
THE MIT STUDY
MIT researchers wanted to uncover “the cognitive cost of using an LLM in the educational context of writing an essay.” Over the course of a few months, study participants took part in 3 evaluative sessions - the 4th session was optional.
In assessing the cognitive impact of independent work versus varied levels of technology-assisted work, the analysis hoped to answer (Page 11):
Do participants write significantly different essays when using LLMs, search engine and their brain-only?
How do participants' brain activity differ when using LLMs, search or their brain-only?
How does using LLM impact participants' memory?
Does LLM usage impact ownership of the essays?
In the first three sessions, participants:
Were randomly assigned to one of the following groups:
a. LLM Group (Group 1): Participants in this group were restricted to using OpenAI's GPT-4o as their sole resource of information for the essay writing task. No other browsers or other apps were allowed (Page 23)
b. Search Engine Group (Group 2): Participants in this group could use any website to help them with their essay writing task, but ChatGPT or any other LLM was explicitly prohibited; all participants used Google as a browser of choice. Google search and other search engines had "-ai" added on any queries, so no AI enhanced answers were used by the Search Engine group (Page 23)
c. Brain-only Group (Group 3): Participants in this group were forbidden from using both LLM and any online websites for consultation (Page 23)
Drafted three essays (within a 20-minute time limit) based upon a prompt they chose from a list of 3 unique options
Had their brain activity monitored during task completion
Took part in post-writing interviews (Page 28)
In the final, voluntary session, participants:
Were reassigned from the LLM Group to the Brain-only Group to complete their task, while the Brain-only group was reassigned to the LLM Group; session 4 excluded search engine use
Drafted an essay based on “a set of personalized prompts made out of the topics EACH participant already wrote about in sessions 1, 2, 3” (Page 28)
Had their brain activity monitored during task completion
Took part in post-writing interviews (Page 29)
In a discussion (Is Using ChatGPT to Write Your Essay Bad for Your Brain? New MIT Study Explained) with Time reporter Andrew R. Chow, the work’s main author, Nataliya Kosmyna, commented “when we looked at neural connectivity we also saw that there is actually a change when you use tools like search engine or LLM versus when you don't use any tools at all . . . what we saw that there is definitely some scaling down happening when . . . you are moving from no tools, meaning - like - you're only using your brain, up to all the way to the LLM. So, you know, you have pretty high neural connectivity happening in brain only, intermediate for search, and then definitely lower for the LLM.”
She continued, “we actually, in the end being again very careful with any speculations, do say the timing of introduction [of AI tools and platforms] might be extremely interesting and potentially important to explore in any future studies.”
HIGHLIGHT: SELECTED RESULTS
Having the subjects provide a single accurate quote from the essay they produced was one aspect of the post-task interview; there was noticeable variation in their ability to do this.
Figure 1: Tables highlighting a participant’s ability to provide both a quote and an accurate quote, based upon their degree of exposure to technology while drafting an essay.
HIGHLIGHT: EEG RESEARCH CONCLUSIONS (Pages 86-87)
Kosmyna clarified in the Time interview that the paper’s findings did not directly support the critique that “critical thinking is gone.” However, in the context of previous research, she believes we are getting a better understanding that a reliance on technology is having an impact on cognitive loads.
Electroencephalography (EEG) recorded “participants’ brain activity in order to assess their cognitive engagement and cognitive load, and to gain a deeper understanding of neural activations during the essay writing task.” (Page 2)
Overall, EEG results revealed a clear pattern: writing without AI engaged more brain regions associated with memory, planning, and creativity. In contrast, LLM use shifted the brain’s focus to more passive, evaluative modes.
“[W]riting an essay without assistance (Brain-only group) led to stronger neural connectivity across all frequency bands measured, with particularly large increases in the theta and high-alpha bands. This indicates that participants in the Brain-only group had to heavily engage their own cognitive resources . . . to meet the high working memory and planning demands of formulating their essays from scratch”
“LLM-assisted writing (LLM group) elicited a generally lower connectivity profile. While the LLM group certainly engaged brain networks to write, the presence of a LLM appears to have attenuated the intensity and scope of neural communication. The significantly lower frontal theta connectivity in the LLM group possibly indicates that their working memory and executive demands were lighter, presumably because the bot provided external cognitive support (e.g. suggesting text, providing information, structure). Essentially, some of the “human thinking” and planning was offloaded, and the brain did not need to synchronize as extensively at theta frequencies to maintain the writing plan. LLM group's reduced beta connectivity possibly indicated a somewhat lesser degree of sustained concentration and arousal, aligning with a potentially lower effort during writing”
“Brain-only group showed evidence of greater bottom-up flows (e.g. from temporal/parietal regions to frontal cortex) during essay writing. This bottom-up influence can be interpreted as the brain's semantic and sensory regions "feeding' novel ideas and linguistic content into the frontal executive system, essentially the brain generating content internally and the frontal lobe integrating and making decisions to express it”
“LLM group, with external input from the bot, likely experienced more top-down directed connectivity (frontal → posterior in high-beta). Their frontal cortex was often in the role of integrating and filtering the tool's contributions (an external source), then imposing it onto their overall narrative. This might be to an extent analogous to a “preparation” phase in creative tasks where external stimuli are interpreted by frontal regions sending information to posterior areas”
“Regarding executive function, the results show Brain-only group's prefrontal cortex was highly involved as a central hub (driving strong theta and beta connectivity to other regions), indicating substantial executive control over the writing process. LLM group's prefrontal engagement was comparatively lower, implying that some executive functions (like maintaining context, planning sentences) were most likely partially taken over by the LLM's automation”
“[W] while the quantity of executive involvement was less for LLM users, the nature of executive tasks may have shifted, from generating content to supervising the AI-generated content”
The “Brain-only group's brain networks were more activated in the manner of creative cognition: their enhanced fronto-parietal alpha connectivity suggest rich internal ideation, associative thinking, and possibly engagement of the default-mode network to draw upon personal ideas and memory”
“LLM group's reduced alpha connectivity and increased external focus might indicate a more convergent thinking style, they might lean on the LLM's suggestions (which could constrain the range of ideas) and then apply their judgment, rather than internally diverging to a wide space of ideas”
“[T]he directed connectivity analysis reveals a clear pattern: writing without assistance increased brain network interactions across multiple frequency bands, engaging higher cognitive load, stronger executive control, and deeper creative processing. Writing with AI assistance, in contrast, reduces overall neural connectivity, and shifts the dynamics of information flow. In practical terms, a LLM might free up mental resources and make the task feel easier, yet the brain of the user of the LLM might not go as deeply into the rich associative processes that unassisted creative writing entails.”
IMPLICATIONS FOR EDUCATION?
A call for caution is the message underlying both Your Brain on ChatGPT and Is Using ChatGPT to Write Your Essay Bad for Your Brain? New MIT Study Explained. Is the rush to infuse AI into all levels of our educational system necessary, appropriate, or productive in the long term? Will students benefit from using this technology during their initial stages of their learning journey - or is this technology better-suited for academic work later on?
As the writers summarized, “our analysis indicates that repeated essay writing without AI leads to strengthening of brain connectivity in multiple bands, reflecting an increased involvement of memory, language, and executive control networks. Prior use of AI tools, however, appears to modulate this trajectory [research authors’ bolding].” (Page 116)
“These findings resonate with current concerns about AI in education: while AI can be used for support during a task, there may be a trade-off between immediate convenience and long-term skill development [research authors’ bolding].” (Page 116)
The study shows an over-reliance on LLMs alters the process of how we think. As AI becomes even more embedded in education, the key may not be whether we use it, but when and how we introduce it to learners. Kosmyna argues that mental practice = working on a task for a while before the introduction of AI assistance = could represent the best approach for maximizing AI’s ability to augment a user’s output. A user develops their cognitive skills first, then adeptly applies AI as a tool to further aid in their learning, avoiding the need to excessively depend on AI to do the work.
She warns of embedding AI too quickly, asking “why do we need to move very fast and break things so fast in our own backyard?” - especially since these actions are not (yet) data-driven. “[D]oes [AI] make a difference? Does it make a change? And more importantly, include all all opinions that would matter in this case - teachers . . . educators, caretakers. If you're talking about children - where I think it would be so, so critical - because we all know those brains are developing. So why do we need to rush these deployments, right? Let's, maybe, pause evaluate and there might be excellent use cases. This technology is not going anywhere . . .”