Extended Abstract—Hybrid DRL-GenAI Architecture for Adaptive Tutoring of Vector Spaces in Virtual Reality
DOI:
https://doi.org/10.56198/z9byt635Keywords:
Immersive Learning, Adaptive Learning, Pedagogical Agents, Embodied Interaction, Serious Games, Mathematics EducationAbstract
Teaching vector spaces in linear algebra imposes a high cognitive load that hinders personalized tutoring in overcrowded classrooms. To address this challenge, this extended abstract presents the design and validation of an immersive architecture grounded in the Semantic Decoupling principle. This approach segregates strategic instructional decision-making via Deep Reinforcement Learning (DRL) from contextualized scaffolding generation via Generative AI (GenAI). The system was evaluated in a Virtual Reality environment through 156 tutorial interventions, a study involving 9 users, and an expert panel review. Results demonstrate high pedagogical resilience: although the DRL agent converged toward a conservative policy (33.3% agreement with the expert standard), the generative layer compensated for this limitation, maintaining perfect instructional coherence (Cohen’s Kappa of 1.00) and high student satisfaction (4.66/5.00). It is concluded that this modular design safeguards learner motivation against algorithmic suboptimality and offers educators a transparent, auditable, and scalable adaptive tutoring model.
Published
Conference Proceedings Volume
Section
Categories
License
The papers in this book comprise the proceedings of the meeting mentioned on the cover and title page. They reflect the authors' opinions and, in the interests of timely dissemination, are published as presented and without change. Their inclusion in this publication does not necessarily constitute endorsement by the editors or the Immersive Learning Research Network.
Contact: publications@immersivelrn.org
