AIM: 테이블탑 롤플레잉 게임(TTRPG) 을 위한 LLM 게임 마스터

다국어 초록 (Multilingual Abstract)

A Tabletop Role-Playing Game (TTRPG) is a form of interactive storytelling in which players assume the roles of protagonists within a scenario, while a human facilitator, known as the Game Master (GM), manages the interactions between the players and the fictional game world. TTRPGs progress primarily through player dialogue and lack explicit environmental definitions or win–loss conditions. The state of the game world often cannot be fully anticipated due to players ’ spontaneous actions. This study proposes an LLM–driven Game Master agent, AIM (Artificial Intelligence game Master), designed to facilitate TTRPG sessions. AIM constructed a graph-based virtual game world and applied Retrieval- Augmented Generation (RAG) to a database derived from TTRPG rule documents, enabling the agent to dynamically generate and manage the game environment. The graph structure allows explicit representation of interactions among world objects and provides a mechanism to resolve the ambiguous world states that frequently arise during gameplay. By leveraging the scenario-generation capabilities of large language models (LLMs), AIM demonstrates the potential of LLMs as autonomous Game Master agents for TTRPGs.

목차 (Table of Contents)

• 1. 서론 1
• 1.1 연구 배경 1
• 1.2 연구 방법 및 범위 3
• 2. 관련 연구 및 서비스 5
• 2.1 관련 연구 5
• 2.2 관련 서비스 7
• 3. TTRPG를 위한 LLM 게임 마스터(AIM) 9
• 3.1 AIM의 구조와 동작 9
• 3.2 규칙 편집기 11
• 3.3 게임 월드 및 플레이어 캐릭터 생성 19
• 3.4 게임 진행자 24
• 4. Dungeons & Dragons 적용 사례 29
• 4.1 게임 월드 생성 30
• 4.2 게임 서술과 규칙 인용 33
• 4.3 게임 월드 편집 36
• 5. 결론. 38