Modeling Skill Progression in Children Through Novel Multidimensional Probabilistic DDA

Dynamic Difficulty Adjustment (DDA) systems are increasingly used in serious games to personalize challenge, sustain engagement, and optimize cognitive learning. This study presents a multi-dimensional, probabilistic DDA framework implemented in Legends of Hoa’Manu, a modular cognitive training game for children that targets core executive functions through distinct training modules, each featuring an independent DDA engine. We focus on the Uka module, which adapts a running memory span task to train working memory. Using gameplay data from 148 children, we evaluate how the DDA system adjusts difficulty across multiple parameters in response to real-time learner performance. Results show that gameplay difficulty rapidly converged to individually appropriate levels, aligning with players’ Zone of Proximal Development (ZPD) within the first hour. The probabilistic adaptation strategy also maintained task variability after plateauing, preventing overfitting and sustaining learner engagement across diverse proficiency levels. These findings highlight the value of multi-dimensional, probabilistic adaptivity for game-based cognitive training.