This article is written by a student writer from the Her Campus at UPRM chapter and does not reflect the views of Her Campus.
Recently, in Spanish class we were assigned to read a short story titled Apocalipsis (1957) by the Argentinian writer, Marco Denevi. The short story revolved around the extinction of humans, caused by their overdependence on robots. According to the short story, the end of our species began when machines achieved perfection and humans, as a consequence, gradually handed over tasks like eating, sleeping, reading, writing, and even thinking; because by the click of a button, machines did everything for them. Spooky stuff, right? Let’s be honest, the short story does not sound far from the truth, does it? In fact, is there any technological innovation that comes to mind for you? For me, there certainly is. I instantly thought about our current fears with the usage of ChatGPT and the impact it has on our cognitive capacities.
MIT Study
On June 10th, 2025 the MIT Media Lab published the study: “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task.” The purpose of the study was to explore the consequences of using Large Language Models (LLM), like ChatGPT, on the brain while writing an essay.
Methodology:
54 participants from ages 18 to 39 were recruited. Participants were split into three different groups:
LLM Group: this group could solely use ChatGPT as a source for information.
Search Engine Group: participants were allowed to use any website to assist them, except AI.
Brain-only Group: participants were not allowed to use LLMs or make any online consultation.
Once assigned to their groups, participants were equipped with an electroencephalogram (EEG) headset, in order to collect their brain’s activity during their writing task. The task was first divided into 3 sessions, where the topic for the essays were taken from SAT tests and the participants only had 20 minutes to complete their task.
For the last session, participants who were assigned to the LLM group in sessions 1, 2, and 3 were then moved to the Brain-only group (now called LLM-to-Brain group). Likewise, participants in the Brain-only group were then assigned to the LLM group (now called Brain-to-LLM group).
The EEG analysis provided evidence that each group had significantly different neural connectivity patterns.
First three sessions:
LLM Group: showed the lowest brain activity especially by the 3rd session. The essays were mostly copy and paste and English teachers mentioned perceiving them as “soulless,” filled with “standard ideas, reocurring typical formulations, and statements.”
Search Engine Group: showed moderate cognitive engagement. Their work was consistent with the “google effect,” which implied that participants relied more on recognizing the information, rather than memorizing it.
Brain-only Group: displayed higher neural connectivity across frequencies related to memory, creativity, and processing of complex ideas like theta, alpha, and delta.
After session 4:
LLM-to-Brain group: participants could not rely on what they learned or recalled from the AI due to poor cognitive engagement of executive control and language production networks.
Brain-to-LLM group: participants showed stronger connectivity than the LLM group in sessions 1, 2, and 3. Writing an essay utilizing AI tools after lacking LLM intervention resulted in more extensive brain network interactions.
So, is ChatGPT Bad?
Although some of the findings in the study are concerning, the truth is more complex. In order to answer our question, it’s important to evaluate certain aspects and limitations that the study had, which were:
A limited number of participants, and after the 4th session the sample size reduced to 18 participants; this sample is not representative of a real-life environment.
It was released before completing the peer review process. Lead author Nataliya Kosmyna, stated in an interview with Time, “…I am afraid in 6-8 months, there will be some policymaker who decides, ‘let’s do GPT kindergarten.’ I think that would be absolutely bad and detrimental.”
During the first 3 sessions, participants chose one out of nine prompts for their essay. However, during the last session participants were offered three prompts made from the topics each participant wrote about previously. This could create a familiarisation effect, especially on the Brain-to-LLM group.
These limitations undoubtedly question the credibility of the study, nonetheless, its findings raise some questions. One of those questions being, should we be alarmed? It seems that every time that a revolutionary technology appears, we are biased to think the worst. We fear it might take away essential human capacities like writing, creating art, and even thinking. Even so, our fears with new technologies or inventions are definitely not new.
It goes way back to the IV century, where Plato in Phaedrus shared that Socrates was opposed to writing, as he thought that writing “will introduce forgetfulness into the soul of those who learn it.” We now know that writing did not kill the memory, instead it expanded it, allowing knowledge to become accumulative and nuanced. The same thing happened with inventions like the printing press, in a time where information was only accessible for the rich, the church or those in power, the novel invention allowed knowledge to travel rapidly and further, impulsing people to learn how to read. However, it is interesting to know that the printing press brought with it concerns about information overload, claiming it to be “confusing and harmful” for the mind. Sounds a bit familiar, huh?
So… should we be alarmed? Being alarmed is not the solution and neither is rejecting AI altogether, innovation is scary but it’s not the technology itself that we fear, our preoccupations are rooted in the fact that novelty brings change and those changes threaten the way we see our lives today.
“This version of ChatGPT is the weakest, most rudimentary artificial intelligence of its kind our students will ever use”
(Miller, 2022)
Let’s admit it, AI is not going anywhere. In fact, it will only get better as time progresses, just like everything around us. Sure, it’s good to know that if I have a question, a chatbot will quickly summarize papers, solve math problems, and also give a hand in wrapping up an article, but we shouldn’t aspire to immediate answers as a way to bypass effort or the process of gaining knowledge. That kind of dependence on technology does not strengthen our cognitive capacities, instead it weakens them. For that reason, we should keep in mind that AI or ChatGPT is meant to be a tool, meaning that we as users are responsible for how we interact with it.
We can definitely make use of tools like ChatGPT to help us with ideas, provide references for a topic of our interest and even ask for feedback, but we cannot allow it to replace our own process in learning, writing, and thinking since, these are the centers of how WE see the world and not the amalgamation of a bank of data. Lastly, when we use ChatGPT, let’s ask ourselves:
Is the information real and verifiable?
Are there any counterarguments to what ChatGPT provides me?
What are the sources? Are they credible?
And, most importantly, is the tool helping me or am I the tool that helps it?
Perhaps the real threat is not machines thinking for us, but rather that we do not notice when we gradually start handing over our tasks.