The Hidden Cognitive Costs of Overusing ChatGPT

The Double-Edged Sword of AI

Artificial intelligence is rapidly becoming a cornerstone of our daily lives, making processes more efficient and boosting productivity across countless domains. We are constantly amazed by its capabilities. In journalism, for example, newsrooms are increasingly deploying artificial intelligence for tasks like copy editing, content generation, and personalizing news delivery. Publications have noted that AI tools have significantly improved content quality and expanded their coverage.

While we should embrace Generative AI in education with thoughtful consideration, it's crucial to be clear-eyed about its potential downsides. Concerns are mounting about the negative cognitive effects for learners who use Large Language Model (LLM) tools like ChatGPT, Grok, and Gemini.

A Warning from a Recent Study

These concerns extend beyond the widely discussed fear that over-reliance on AI can "threaten truth." A recent study from the MIT Media Lab suggests a more personal risk: exclusive reliance on GenAI for writing could, over time, impair critical thinking, memory, and curiosity.

We must be careful not to repeat the mistakes made with social media. We now understand that the long-term cognitive effects of smartphones and social media have contributed to a mental health crisis among our youth, leading to a surge in cognitive issues, anxiety, and attention deficits. As we integrate GenAI into learning environments, we must consider these potential cognitive impacts on young learners.

How AI Affects the Brain: The MIT Experiment

The MIT study found that exclusively using ChatGPT for writing tasks has more downsides than upsides for learners. The experiment involved 54 participants, aged 18 to 39, who were divided into three groups. Each group was asked to write multiple essays based on college-admissions-test style prompts.

• Group 1: Used ChatGPT.
• Group 2: Used Google’s search engine.
• Group 3: The “brain only” group, which used no algorithmic tools.

As the participants worked, their brain activity was monitored using EEG scans to measure cognitive engagement.

The Alarming Results: Reduced Brain Activity and Memory

The findings were stark. The ChatGPT group showed the least amount of brain activity and, according to the study, "consistently underperformed at neural, linguistic, and behavioral levels."

Most revealing was the impact on thinking and memory. In the group that used ChatGPT, a mere 20% could recall what they had written, and an astonishing 16% could not even recognize the text they had produced. This group also relied progressively more on copy-and-pasting, resulting in essays that were wooden and repetitive.

In contrast, the “brain-only” group showed the highest neural activity across all measured bands, which indicates the formation of new ideas, creativity in language use, and better memory consolidation. The Google Search group fell in the middle but showed improvement when later allowed to use ChatGPT to rewrite their initial drafts, suggesting AI is better as a refinement tool than a starting point.

The Path Forward: Education Over Prohibition

In response to these challenges, some have suggested reactionary measures like bringing back blue book exams and in-class writing assignments. However, a better approach is to educate learners and parents on the negative cognitive effects of exclusive reliance on GenAI tools.

As educators, we have a clear task: to innovate our teaching methods and persuade students to use GenAI as a powerful tool for research and editing—but not as a substitute for the crucial work of writing a first draft. The goal is to harness the power of AI without sacrificing the cognitive skills that are essential for deep learning and creativity.

This post is based on an opinion piece by Anup Kumar, a professor of communication in the Levin College of Public Affairs and Education at Cleveland State University.