Sunday Idika, Saad Saihi
This paper introduces the design, creation, and evaluation of a new adaptive learning system featuring virtual intelligence, fully integrated with game-motivation elements and real-time difficulty-level control to improve learners' engagement and educational outcomes. The prototype includes a workaround combination of a Python backend based on Flask to dynamically fetch from the questions in the quiz data generator, and subsequently track the performance alongside a frontend of Unreal Engine 5.4 which provides an immersive 2D quiz interface. Content is derived from the OpenTDB API while the VaRest plugin allows easy data communication between backend and frontend. A rule-based threshold algorithm is used to personalize the difficulty of questions in real time based on learner performance. Evaluation results indicate that the system does a good job in matching the difficulty of the questions and proficiency of the user, with an average API response time of 190 ms well above standard benchmarks and increases user engagement in comparison to a conventional static quiz system. The work presented in this study contributes to the field because: (1) it introduces a new framework with innovative technical solutions that combines the technology of game engines with mechanisms of adaptive AI learning; (2) it provides proof of concept that personalization of learning in real time with interpretable rule-based algorithms is feasible; and (3) it lays down practical principles of design that can be used to create intuitive and gamified educational interfaces. While limited in terms of scale of user testing and the lack of more depth around machine learning integration, the research has left behind a foundation for adding more depth to adaptable learning platforms in the future. It features the value of low-latency architecture, the overlooked educational value of Unreal Engine, and the benefits of clear adaptation to improve the digital learning experience.