Intelligent Immersive Accessibility: A Systematic Review of AI-Enhanced Virtual Reality Interaction Design for Inclusive and Equitable Education
Abstract
This study systematically synthesises findings about artificial intelligence (AI)-enhanced virtual reality (VR) and extended reality (XR) systems developed for accessibility and inclusive education. It examines the architectural, implementation, and evaluation principles behind AI-infused immersive learning environments as adaptive systems for educational contexts. We conducted a systematic literature review following the PRISMA 2020 guidelines. We searched Scopus, Web of Science, IEEE Xplore, ACM Digital Library, and ERIC databases for peer-reviewed articles published between 2021 and 2025. Only nine of those studies ultimately met the screening and eligibility criteria. The included studies were assessed through a structured narrative synthesis, which examined AI integration patterns, immersive interaction design, accessibility operationalisation, evaluation maturity, and architectural contributions. The study revealed that AI integration occurs mainly at the perceptual layer, including recognition, classification, and translation, but the system lacks a complete closed-loop adaptive architecture that connects learner modelling with real-time personalisation and longitudinal learning analytics. Interaction design is largely disability-dependent, while evaluation methods often choose between usability and system performance testing, and between system performance testing and educational outcome assessment. Architectural contributions exist as separate modules that fail to integrate accessibility, adaptive intelligence, pedagogy, and governance into a cohesive system. The Intelligent Immersive Accessibility Framework (IIAF), a conceptual, evidence-traceable but not yet empirically validated layered architecture, addresses this fragmentation. The emerging nature of AI-based immersive accessibility research helps explain why only a few studies were included, and why the field needs research that spans multiple contexts and extends over longer periods.