Enhancing Narrative Design in Game Development Education through AI-Driven Generative Agents

This proposal seeks funding to develop an innovative AI-powered storytelling prototyping tool aimed at transforming and improving the learning efficacy of a very popular General Education (GE) course at CityUHK: GE4102 Video Game: History, Industry, Society, and Creativity. In this course, undergraduate students from various disciplines design and create a final game project, often with strong narrative and role-playing game (RPG) elements. Students engage in extensive creative writing for various aspects of the project. Traditional methods of story creation in this course, characterized by passive, linear writing processes and fragmented contributions, often result in incoherent narratives and limited creative exploration. To address these challenges, we propose leveraging Large Language Models (LLMs) and the framework presented in Generative Agents: Interactive Simulacra of Human Behavior (Park et al., 2023) to create a collaborative, dynamic, and iterative narrative design tool.Our tool will enable students to collaboratively design generative AI agents with customizable traits (e.g., backstories, relationships, goals) and simulate their interactions within a dynamic sandbox environment. By observing emergent narratives and social behaviors generated through agent interactions, students will iteratively refine their game's story, characters, and events. This dynamic, multi-directional process fosters creativity, ensures narrative coherence, and encourages experimentation with diverse story paths. It moves beyond the limitations of traditional, linear writing approaches.Crucially, the proposed tool will enhance team-based collaborative learning by bridging diverse educational backgrounds and skill levels. Students from disciplines such as English, Engineering, and Creative Media will engage actively in the narrative creation process. The AI-driven simulation transforms writing from a static, one-way task into an evolving, interactive experience. For instance, students with strong writing skills can refine character dialogues generated by the tool, while those from STEM backgrounds can focus on designing cause-and-effect systems that shape emergent story branches. By requiring real-time negotiation of ideas and iterative adjustments to AI outputs, the tool promotes equitable participation and leverages interdisciplinary strengths. This collaborative dynamic mirrors real-world creative industries, where teams with varied expertise co-develop complex narratives.The proposed system, built upon the open-source foundation of Park et al. (2023)'s work, will integrate a user-friendly web interface for real-time collaboration. Expected outcomes include enhanced student engagement, improved interdisciplinary skills in storytelling and the literacy of generative AI technology, and a scalable pedagogical model for integrating generative AI into creative education. This project aligns with CityUHK's commitment to innovation in general education and team-based learning, preparing students for the evolving intersection of AI and creative industries.