Preprint: LLM-Driven Scenario Generation for Serious Game Moderation: A Process Model ApproachLLM-Driven Scenario Generation for Serious Game Moderation: A Process Model Approach

Abstract

Serious games (SG) have gained increasing attention in education due to their capacity to enhance learner engagement. Recent advances in Artificial Intelligence (AI), particularly Large Language Models (LLMs), offer new opportunities to support SG design and moderation through content generation. However, procedural guidance for educators on how to reliably integrate these models remains limited, and LLMs introduce challenges such as hallucinations, factual inaccuracies, and bias, highlighting the need for structured frameworks. This study investigates how LLMs can be systematically employed to generate playable scenarios that support SG moderation. Following a Design Science Research (DSR) approach, a business process model (BPM) artifact was developed and evaluated iteratively through controlled experiments and expert interviews. The findings suggest that LLMs can generate coherent and playable scenarios when guided by a structured process model. Expert feedback emphasized limitations related
to creativity control and the need for empirical validation in classroom settings. This study provides practical guidance for educators and contributes to a broader understanding of AI integration in SG.