Application of Procedural Content Generation to Game NPC Dialogue Design: Research Progress and Challenges

Abstract

As open-world and role-playing games grow in scale, NPC dialogue increasingly supports characterization, narrative progression, and worldbuilding. Manually authored dialogue provides control but becomes costly across many characters and game states. This review examines procedural content generation (PCG) in NPC dialogue, tracing preset text, branching narratives, template and rule-based systems, and generative AI. It considers narrative design, human-computer interaction, and artificial intelligence. Evidence suggests that procedural techniques can reduce repetitive authoring, while some user studies associate open-ended dialogue with longer conversations and higher engagement in specific settings. These findings do not establish universal superiority. Large language models increase conversational flexibility but introduce narrative inconsistency, character drift, hallucinations, latency, and moderation challenges. The review proposes a risk-sensitive hybrid approach that preserves human-authored critical scenes while constraining AI-generated optional dialogue through game-state checks, character memory, and evaluation.