Bridging Theory and Practice in Machine Learning Education Through Mixed Reality Gamification

teaching the abstract principles of machine learning (ml) remains a considerable challenge within computer science education. This study presents IMREG, an immersive mixed-reality (MR) educational game, coupled with a new framework that uses gamification to improve the understanding of core ML topics such as image recognition, reinforcement learning, and linear regression. The IMREG includes game-based learning in an MR setting, allowing students to interactively alter the inputs and parameters of the model. During the first stage, users modify the values of the image processing parameters. In the subsequent stage, they direct the behavior of the agent to influence the dynamics of reinforcement learning. The final stage involves systematically gathering data samples, visualizing outcomes through charts, and modifying regression coefficients to see prediction updates in real time. Engagement is increased through intuitive natural interactions, including gesture detection and eye-tracking. A quantitative study with 20 undergraduate students from both computer science and other disciplines reported positive responses regarding usability and engagement. A post-intervention assessment evaluated the knowledge of the three ML concepts. The participants answered three sets of multiple-choice questions, achieving a mean accuracy of 76.83%. In particular, students from non-computer science backgrounds obtained higher scores (mean score of 81.39%) than their computer science counterparts (mean score of 70%), suggesting the IMREG's capacity to widen access to ML education. These results support an effective MR game design framework, offering valuable guidance for creating inventive and captivating teaching approaches in computer science.