Human–LLM Co-Storytelling
Studying how players interact with AI narrative systems through the lens of interactive narrative play, using AI Dungeon as a case study.
How do players actually experience and appropriate LLM-driven storytelling in practice? This in-progress project examines large-scale human–AI narrative play through an analysis of over 51,000 Reddit submissions and 255,000 comments from the AI Dungeon community.
Using topic modelling, clustering, and qualitative interpretation, I investigate how players negotiate agency with a generative system, reinterpret its breakdowns as creative affordances, and develop adaptive co-storytelling strategies over time. Drawing on game studies and play theory, I conceptualize LLM-based co-storytelling as a procedural system where narrative emerges through iterative human–AI interaction — rather than purely as a writing assistant or content generator.