Towards an AI-Augmented Textbook
Textbooks are a cornerstone of education, but they have a fundamental limitation: they are a one-size-fits-all medium. Any new material or alternative...
Research Paper
Summary
The study examines neural and behavioral effects of using large language models for essay writing, comparing LLM assistance with search‑engine use and unaided cognition. EEG results show reduced brain connectivity for LLM users, and behavioral analysis finds lower ownership and citation accuracy in that group.
Highlights
- EEG shows weakest functional connectivity for LLM‑assisted writers versus strongest networks for Brain‑only participants.
- LLM users report lowest essay ownership and have difficulty accurately quoting their own work.
- Condition switches (LLM‑to‑Brain, Brain‑to‑LLM) demonstrate that external tool use lowers neural engagement over multiple sessions.
- Both AI judge scores and human grading indicate consistent underperformance of LLM‑assisted essays.
auto-generated
Context
Audience
Cognitive neuroscientists, educational psychologists, artificial intelligence researchers, and educators interested in the impact of AI writing assistants
DomainCognitive Neuroscience of AI‑Assisted Writing
FormatPDF preprint on arXiv
Accessfree online
Topics
Related
Tool‑mediated learning researchCognitive load theory with generative AINeuroimaging methods in educationAI ethics and academic integrity literature