[2A1-GS-10] AI application:

[2A1-GS-10-03] Game Scenario Generation Using State Transition-Based LLMs

〇Tomohiro Suzuki¹, Naoki Mori¹ (1. Osaka Metropolitan University)

Keywords:LLM, Game Scenario Generation, State Transition

Generative AI has advanced rapidly, significantly impacting various fields, including gaming. In particular, it has been applied to text and image generation, character design, and dialogue-based game mastering. However, one of the major challenges in AI-driven scenario generation is the inconsistency of generated narratives, which can lead to logical contradictions and incoherent storytelling. To address this issue, this study proposes a method that incorporates a state transition structure into large language models (LLMs) to improve narrative coherence. We define a structured data representation for state transitions and integrate it into prompts, allowing the model to better understand and maintain scenario logic. By leveraging this approach, we aim to facilitate the automated generation of structured and contextually consistent game scenarios across diverse objectives, scales, and genres. Our findings suggest that this framework can enhance the adaptability and reliability of AI-generated game narratives, paving the way for more robust applications in interactive storytelling and game design.