An AI-Guided Personalized Physics Tutor with Mixed Reality Support for Undergraduate Physics Education

The rapid advancement of artificial intelligence (AI) and immersive technologies has opened new possibilities for enhancing undergraduate teaching and learning, particularly in physics education, where students are often required to reason about abstract concepts and complex spatial relationships. In foundational courses such as PHY1101: Introductory Classical Mechanics, students frequently struggle to connect mathematically dense formulations and two-dimensional representations with the underlying three-dimensional physical phenomena. These difficulties commonly result in fragmented conceptual understanding, overreliance on procedural problem solving, and reduced confidence in applying physical principles across contexts.This project proposes PhysiVerse, an AI-guided, Mixed Reality (MR)-supported personalized physics tutoring framework designed to support student learning in PHY1101. PhysiVerse is conceived not merely as a visualization tool, but as a student-centred AI course tutor that responds adaptively to individual learning difficulties and provides structured instructional scaffolding aligned with the course syllabus and learning outcomes. The project directly addresses persistent learning bottlenecks in introductory mechanics, including challenges in vector reasoning, rotational dynamics, reference frame interpretation, and energy conservation.At the pedagogical core of PhysiVerse is a structured instructional logic termed DTST (Diagnosis–Trigger–Scaffold–Transfer). Under this framework, the AI tutor first diagnoses the underlying conceptual difficulty reflected in a student’s query, rather than simply evaluating answer correctness. Based on this diagnosis, the system selectively triggers an appropriate instructional response, determining whether the concept can be addressed through text-based explanation, web-based 3D visualization, or MR-supported interactive exploration. Immersive MR activities are activated only when spatial or dynamic intuition is pedagogically essential. Each interactive experience is paired with guided prompts that support prediction, manipulation, and reflection, before guiding students back to formal problem-solving tasks to reinforce conceptual transfer and coherence.Within the PhysiVerse learning environment, abstract physical quantities such as forces, motion, and rotation are transformed into manipulable three-dimensional representations that students can directly explore and interrogate. Students engage first with foundational MR learning scripts explicitly aligned with the PHY1101 syllabus and gain adaptive access to more advanced interactive models as their conceptual understanding develops. This structured scaffolding approach accommodates diverse student backgrounds and learning paces, while minimizing cognitive overload and maintaining alignment with course-level learning outcomes.The project builds upon the team’s established experience in developing MR-based physics learning modules, ensuring efficient development, technical feasibility, and pedagogical reliability. PhysiVerse will be embedded into the departmental AI physics teaching platform and integrated into regular PHY1101 teaching activities. Key deliverables include an AI-enabled MR resource library, reusable DTST tutoring policies, scaffold prompt templates, and course-agnostic instructional scripts that can be readily adapted to other undergraduate physics courses with minimal redesign.Project effectiveness will be evaluated through learning analytics, student surveys on perceived conceptual understanding and learning confidence, and targeted pre/post concept-check questions focused on core mechanics concepts. Overall, this project aims to establish a scalable, transferable, and pedagogically grounded model for personalized, technology-enhanced physics education that meaningfully improves student learning and supports sustainable teaching development in undergraduate physics.