
Date: Friday 6th November
Time: 12 noon to 1pm
Place: MS Teams meeting
Our first paper is the widely discussed MIT preprint, ‘Your Brain on ChatGPT’, which received considerable media attention. and, despite being in pre-press, has over 700 citations according to Litmaps. It argues that the use of LLMs leads to what the authors call ‘cognitive debt’, involving changes to neural connectivity patterns in the brain. These are associated with reduced recall of written work and diminished sense of ownership over the writing produced.
The research analysed the essay writing of 3 groups of participants using either no digital tools, simple search, or ChatGPT. Brain activity was monitored alongside the written work that was also analysed in multiple ways.
This is a very long paper, and it may be best to read the abstract/introduction, and parts of the discussion and conclusions, skipping most of the technical sections as appropriate. As most of us don’t have a background in cognitive science, that should give us enough context to discuss the paper’s main claims about cognitive debt, recall, ownership, and the possible educational implications of LLM use.
There are also some media articles about the research linked below.
Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. arXiv. https://doi.org/10.48550/arxiv.2506.08872
Potential discussion points:
In this research, the participants who produced essays with LLMs cited a reduced sense of ownership of the work, compared to those produced with minimal technology. Should this concern educators in relation to student confidence and well-being?
The idea of cognitive debt is introduced as taking shortcuts which weaken certain thinking skills. How persuasive is this argument? Is this distinctive to AI, or part of a long history of humankind gaining and transforming skills? For example, did the invention of writing change our capacity for memory recall, just as later technologies changed practices of painting, craft, or calculation?
The paper claims that the greater cognitive load experienced by participants in the ‘Brain-only’ group, namely the mental effort used in working memory, led to deeper learning outcomes and greater internalisation of knowledge. If we accept this claim, how do we balance the use of AI tools? How can we distinguish between reducing unnecessary effort and preserving the productive effort needed for learning?
Join this Reading Group:
Register for this Reading Group on Your Brain on ChatGPT taking place on 6th November at 12 noon.
You must join the the UAL AI Network to receive future Reading Group invitations.
Author: Mark Robinson, Learning Technologist, LCF
Email: m.robinson@fashion.arts.ac.uk

Leave a Reply