AI-Guided Adaptive Immersive Learning Experiences

Abstract

AI-guided adaptive immersive learning combines the presence and embodiment of virtual reality with real-time AI systems that continuously monitor learner state, adjust virtual environment parameters, and provide personalised guidance to maintain optimal learning conditions throughout immersive sessions. Unlike static VR environments that deliver identical experiences to all learners, AI-guided systems dynamically reshape the immersive experience -- adjusting narrative pace, difficulty of challenges, NPC behaviour, environmental complexity, and scaffolding intensity -- based on real-time inference of learner cognitive load, engagement, and knowledge state from multimodal sensor streams. This paper proposes the AI-Guided Adaptive Immersive Learning Framework (AGAILF), a systematic design and evaluation of five AI guidance architectures -- rule-based adaptation, Bayesian knowledge tracing adaptation, reinforcement learning environment agent, large language model narrative director, and multimodal cognitive state adaptation -- applied to three immersive educational scenarios: historical simulation (secondary education), molecular biology exploration (undergraduate), and clinical decision-making training (postgraduate medical). AGAILF evaluates 2,160 learner sessions across three scenarios, introducing the Adaptive Immersive Learning Score (AILS) integrating learning outcomes, engagement quality, and adaptation responsiveness. Key results: multimodal cognitive state adaptation achieves the highest AILS (0.924) with 44.2% learning gain improvement over static VR; LLM narrative director achieves highest learner engagement (4.68/5.0); RL environment agent reduces cognitive overload incidents by 56.4%. The framework provides AI guidance architecture selection principles for immersive education designers.