TEMPO.CO, Jakarta – A study by researchers from the Massachusetts Institute of Technology (MIT) has revealed that the use of artificial intelligence (AI) models such as ChatGPT for essay writing could lead to cognnitive decline.
The recent study revealed that the Large Language Model (LLM) erodes critical thinking skills, creativity, and decreases memory capabilities.
These findings also reflect the impact of LLM use in education. “We demonstrate the pressing matter of a likely decrease in learning skills based on the
results of our study,” the researchers wrote, as quoted from the study’s abstract published by the MIT Media Lab on Tuesday, June 10, 2025.
The scientific article was published under the title ‘Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task.’ A total of eight researchers were involved in the study.
The research team referred to these negative effects as cognitive debt resulting from the efficiency of LLM use to create essays. Users also tended to reduce their tendency to critically evaluate the opinions generated by AI.
The study involved 54 participants aged 18-39 years, including 32 women, 19 men, two participants who identified as non-binary, and one participant who did not disclose their gender.
They were undergraduate, postgraduate, and doctoral students from MIT, Wellesley College, Harvard University, Tufts University, and Northeastern University. All these campuses are located in the city of Boston, Massachusetts, United States.
The first group was allowed to fully utilize the LLM GPT-4o model owned by OpenAI for essay writing. The second group was only allowed to use existing sites with a Google search engine but were instructed to exclude AI-based search results. The third group solely relied on their brain’s capabilities without AI assistance and web searches.
In the first stage, the participants were briefed about the research and then proceeded to fill out a questionnaire that also contained their experiences in using the LLM AI model. Subsequently, in the second stage, the participants were fitted with Neuroelectrics Enobio 32 headsets to collect EEG (electroencephalography) brain signal activity throughout the duration of the study and for each session.
In the third stage, the researchers conducted a calibration test by having the participants solve multiplication or addition math problems. Following this, they took a break and then proceeded to move their eyes vertically and horizontally and closed their eyes for two minutes to test the headset’s response. They then had 20 minutes to write an essay. Participants wrote an essay based on one of nine topic choices.
For the fourth session, the participants were asked to write another personalized essay. However, not all participants attended the fourth session. “Session 4 was not a mandatory session, and thus, due to participants’ availability and scheduling constraints, only 18 participants were able to attend,” wrote the research team.
The results showed that the LLM-only group entirely had the weakest overall neural connectivity compared to the group without LLM and only using web search engines. Researchers believe that the use of LLM diverts part of the brain’s cognitive load. Additionally, the LLM-only group also struggled to correctly use their own citations.
“Participants who were in the Brain-only group reported higher satisfaction and demonstrated higher brain connectivity, compared to other groups,” wrote the research team.
According to the research team, this study could serve as an initial guide to understanding the cognitive and practical impact of AI in the student environment. They also emphasized the importance of longitudinal studies to understand the long-term impact of LLMs like ChatGPT on the human brain before they are recognized as something truly valuable for humans.
Editor’s Choice: OpenAI Confirms ChatGPT Outage, Assures Imminent Recovery
Click here to get the latest news updates from Tempo on Google News
