Bridging the Affective Gap: A Conceptual Framework for Multimodal Emotion-Aware Adaptive Learning in Online Education

Abstract

Online learning has scaled rapidly, yet it strips away the nonverbal channel that lets teachers read confusion, frustration, or disengagement and adjust on the spot. This loss is the affective gap. Advances in facial, vocal, and textual emotion recognition make it technically feasible to sense learner affect at a distance, but the field still lacks an integrative design that connects sensing to pedagogy and to responsible governance. Most prior work either reports recognition accuracy on benchmark corpora or reviews that literature descriptively, leaving a gap between detection and classroom action. This paper contributes a conceptual framework, not an empirical system. We propose a four-layer architecture for emotion-aware adaptive learning that spans multimodal sensing, fusion and learning-centred emotion inference, an affective-pedagogical decision layer, and an analytics layer with a teacher in the loop, all wrapped in a cross-cutting governance boundary. We ground the design in a working taxonomy of learning-centred emotions, compare fusion strategies, formalise an uncertainty-gated mapping from inferred states to pedagogical interventions, and surface the privacy, consent, and fairness obligations the design must honour. The framework is offered as a blueprint and a research agenda; we close by outlining how each layer can be validated empirically.