The objective was to optimize the learning outcomes, by personalising the educational experience and enhancing engagement. The adaptive game version offered personalised task recommendations based on an AI model which considers player performance and task difficulty. Next task to play was suggested via a recommender that dynamically selected the SQL tasks the student plays. We also implemented a hard coded help function with additional SQL guidance related to the task, available for 42 SQL keywords.

The evaluation included 41 participants in four sessions: 13 IBM IT apprentices, 20 FHNW IT BSc students in Brugg, seven BIT BSc students from Basel, and two MSc students. Sessions lasted from one to two hours, with feedback collected via Likert scale questions and short text responses. The results highlighted the possible impact of the class environment and previous programming experience, group composition and size, and session length on engagement and efficiency. In our experiment, two master students were the fastest, and BIT students in Basel the slowest, with students needing 42 to 62 seconds per task on average in each group, which is a fast pace. High levels of interest and engagement were evident, with the majority of participants giving positive feedback.

Our results show that using AI for dynamic difficulty adjustment and personalized learning is feasible and leads to good playing outcomes, with students progressing fluently through the game. The experiment showed the need for further educational enhancements in the game, allowing students to learn independently, via better SQL help menus.