

Your Brain on ChatGPT with Nataliya Kosmyna
11 snips Sep 19, 2025
Nataliya Kosmyna, a research scientist at MIT Media Lab, explores how AI tools like ChatGPT influence our brains. She shares intriguing findings from an essay study that show differences in cognitive load and brain activity when using AI. The discussion touches on concerns about the impact of LLMs on learning, including perceived declines in diagnostic accuracy and creative limits of AI-generated content. Nataliya emphasizes the need for responsible AI policies and safeguards to protect vulnerable users as technology evolves.
AI Snips
Chapters
Transcript
Episode notes
LLMs Reduce Active Brain Connectivity
- Using an LLM to write essays reduces brain functional connectivity compared with writing unaided or using search.
- Lower connectivity correlated with more homogeneous, less memorable writing and reduced sense of ownership.
Generated Text Feels Less Memorable
- Essays produced with ChatGPT were more homogeneous in vocabulary and harder for authors to quote from memory.
- 83% of LLM users could not supply a quote they had just produced, and some reported no ownership of the text.
When You Use Tools Matters
- Timing of tool access may matter: prior brain effort then tool use increased connectivity compared with losing tool access.
- This suggests building foundational skills before relying on LLMs could be beneficial.