Cross-modal Affinity-aligned Multimodal Learning Analytics for Predicting Student Collaboration Satisfaction in Game-Based Learning

In three linesAAMLA, a multimodal learning analytics framework, predicts student collaboration satisfaction in game-based educational environments. The CAMA module aligns modalities (gaze, action units, pose) via affinity matrices and contrastive learning, adaptively suppressing uninformative modalities. Tests on 50 middle school students in EcoJourneys show improvement over unimodal baselines.

Summary generated by Claude — human-verified