Closing the Loop: A Systematic Review of Experience-Driven Game Adaptation

Adaptive game systems aim to enrich player experiences by dynamically adjusting game content in response to user data. While extensive research has addressed content personalization and player experience modeling, the integration of these components into fully operational adaptive gameplay systems remains limited. This systematic review, conducted in accordance with PRISMA guidelines, analyzes 17 empirical studies published between January 2015 and May 2024, identifying and analyzing approaches that implement the complete experience-driven loop -- including player sensing, modeling, and content adaptation. Game telemetry remains the most prevalent sensing modality, although other non-invasive methods suitable for affective modeling -- such as facial expression analysis (FEA) and peripheral interaction data -- remain underutilized despite their potential for real-time emotional inference. Knowledge-based methods, such as rule-based systems and heuristics, dominate modeling and adaptation due to their interpretability and low resource demands, whereas machine learning approaches face challenges related to data availability and transparency. Despite their relevance to immersive and therapeutic experiences, affective states such as stress and anxiety remain largely ignored, as systems continue to favor performance over emotion-sensitive adaptation. These findings highlight a crucial research direction: advancing emotionally responsive game systems that move beyond performance optimization by incorporating underutilized sensing modalities -- such as FEA and peripheral interaction -- to enable real-time affect-driven personalization. Advancing in this direction holds strong potential to increase immersion, personalize gameplay, and support affect regulation across entertainment and therapeutic contexts.