STUDENTS who rely on large language models (LLMs) for essay writing show weaker brain activity and lower engagement compared to those who work without assistance, a new study by MIT researchers has shown.
This comes at a time when people are using AI to do most daily tasks, including writing and coding, among others, creating cognitive debt.
This refers to the mental cost or deficit that builds up when we rely on shortcuts, tools, or simplified thinking instead of fully engaging our cognitive abilities, which, unlike financial debt, cannot be paid back.
The study involved 54 participants divided into three groups: LLM users, search engine users, and a Brain-only group that wrote essays without external tools.
One group wrote essays using ChatGPT, one Google, while the last one used nothing but their own brain.
Each group completed four sessions over four months, wearing electroencephalography (EEG) headsets, which tracks real-time brain activity.
The essays were evaluated with natural language processing, while scoring was done by human teachers and an AI judge.
Findings revealed that Brain-only participants had the strongest and most distributed neural networks, search engine users were moderately engaged, and LLM users showed the weakest connectivity, indicating reduced cognitive effort.
Participants who switched from LLM to Brain-only tasks in the final session showed under-engagement, while those moving from Brain-only to LLM had higher memory recall, similar to search engine users.
Therefore, when ChatGPT users were told to write their essays without AI in the final session, their brains were still measurably weaker than those who never used AI at all.
LLM users also reported the lowest sense of ownership over their work and struggled to quote their own essays, whereas Brain-only participants felt the most ownership.
Dangers of Cognitive Debt
Having a high cognitive debt can make people overly dependent on external tools, reducing their memory, critical thinking, and problem-solving skills.
It can also prevent deep understanding and make it harder to retain or apply knowledge, which may, over time, weaken creativity, decision-making, and adaptability, especially in new or complex situations.
In learning or work environments, high cognitive debt can also lower engagement, confidence, and a sense of ownership, ultimately limiting long-term growth and performance.
How to Repay Cognitive Debt
While cognitive debt, unlike financial debt, can’t be paid, it can be reduced by actively engaging the brain instead of relying solely on shortcuts or tools.
This includes practices such as deliberate practice, reflection, and problem-solving. Rewriting notes, summarising information, and explaining concepts to others reinforce learning.
The study concludes that while LLMs make writing easier, over time, their use may diminish neural, linguistic, and behavioural performance, raising questions about their role in learning. – IOW Data.
