Hyderabad: People who rely on artificial intelligence chatbots, such as ChatGPT, to help them with tasks like writing an essay could experience cognitive decline, suggests a new study. Researchers at the Media Lab of the Massachusetts Institute of Technology (MIT) found that reliance on large language models for essay writing reduced brain connectivity, memory recall, and sense of authorship.
The study—Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task—published at the research-sharing platform arXiv, also raised concerns about the long-term impact of AI on learning, as users who relied on LLMs showed weaker cognitive, linguistic, and behavioural performance over time.
Methodology of the study
To explore the neural and behavioural consequences of LLM-assisted essay writing, participants were divided into three groups: LLM, Search Engine, and Brain-only (no tools). Each group completed three sessions under the same conditions. In the fourth session, there was a change with LLM users being reassigned to the Brain-only group (LLM-to-Brain) and Brain-only users being reassigned to the LLM condition (Brain-to-LLM). Notably, a total of 54 participants took part in Sessions 1-3, with 18 completing Session 4.
Researchers used electroencephalography (EEG) to assess cognitive load during essay writing, analysed essays using NLP, and scored them with help from human teachers and an AI judge.
Study raises 'red flags' for LLM use
The study found significant differences in brain connectivity, with Brain-only participants exhibiting the strongest and most distributed networks, Search Engine users showing moderate engagement, and LLM users displaying the weakest connectivity. Notably, cognitive activity was observed to decrease with greater reliance on external tools.
In session 4, LLM-to-Brain participants were found to have reduced alpha and beta connectivity, suggesting under-engagement. Meanwhile, Brain-to-LLM users showed higher memory recall and greater activation in the occipito-parietal and prefrontal areas, resembling patterns seen in Search Engine users.
The study also noted that self-reported ownership of essays was lowest in the LLM group and highest in the Brain-only group. Additionally, LLM users struggled to accurately quote their own work. The research suggests that while LLMs offer immediate convenience, they come with potential cognitive costs.
"Over four months, LLM users consistently underperformed at neural, linguistic, and behavioural levels," MIT researchers said. "These results raise concerns about the long-term educational implications of LLM reliance and underscore the need for deeper inquiry into AI's role in learning."
