Immersive Learning Analytics for VR-Based Education Systems

Abstract

Immersive learning analytics (ILA) captures the rich multidimensional behavioural stream generated during virtual reality learning -- head orientation, gaze patterns, hand movement trajectories, controller interaction logs, physiological signals, voice activity, and spatial navigation paths -- to infer learner cognitive states, engagement levels, and learning progress that conventional clickstream analytics cannot access. VR's sensor-rich environment enables unprecedented visibility into the learning process: where the learner looks reveals attention; hesitation in movement reveals uncertainty; interaction frequency reveals engagement; and spatial navigation patterns reveal spatial understanding development. This paper proposes the Immersive Learning Analytics Framework (ILAF), a systematic evaluation of six ILA modalities -- gaze analytics, spatial navigation analytics, hand interaction analytics, physiological signal analytics, voice activity analytics, and multimodal fusion -- applied to four VR educational applications (anatomy education, physics simulation, surgical skill training, and language immersion) across 1,920 learner sessions. ILAF introduces the ILA Insight Score (ILAIS) measuring predictive utility, interpretability, and privacy impact. Key results: multimodal fusion achieves the highest ILAIS (0.912), predicting learning outcomes with AUC = 0.924; gaze analytics achieves the highest single-modality performance (AUC = 0.884); spatial navigation uniquely identifies spatial reasoning development not captured by any other modality. The framework provides ILA modality selection guidance and privacy design principles for VR learning systems.