Leveraging Self-Adaptive Systems and Generative AI for Personalizing Educational Serious Games: Architecture and Future Challenges

The integration of self-adaptive systems (SAS) with Generative Artificial Intelligence (GAI) opens new possibilities to improve serious games (SG) by enabling dynamic adjustments in gameplay, customized content, and personalized learning experiences. This vision paper explores the potential of combining GAI with the MAPE-K framework to enable the generation and adaptation of real-time content in SGs. We propose a framework that addresses key challenges, including generalizing to different SGs, validating AI-generated content in real time, and tailoring content to individual players, while discussing future challenges such as managing conflicting goals, handling uncertainty, and defining robust evaluation metrics. Practical examples and an adaptable system architecture illustrate the framework envisioned. Finally, we discuss the challenges of evaluating SAS in this domain and outline future research directions.