#488: Redistributing Cognition in AI-Supported Writing

Writing is a distributed cognitive activity that unfolds across people, tools, and representations. The emergence of large language models (LLMs) introduces a new node in this distributed system, transforming how writers plan, draft, review, and reflect. This study examines how students engage with an LLM-powered writing platform in an upper-division engineering writing course. Using qualitative coding of 8,878 student-AI chat messages, we analyzed how cognitive work in writing was redistributed between human and AI. We found that students most often used the AI to seek evaluative feedback, positioning it as a proxy instructor or peer. Patterns of interaction varied across instructors, indicating that instructional design shaped how cognitive tasks were distributed. Lastly, students also demonstrated moments of critical engagement, questioning or resisting AI suggestions. Overall, we found that LLM-supported writing reorganizes the cognitive system of writing, positioning AI as a partial collaborator.