AAMLA Framework Predicts Student Collaboration in Game-Based Learning
A new framework called Affinity-Aligned Multimodal Learning Analytics (AAMLA) has been introduced by researchers to forecast student satisfaction in collaborative game-based learning settings. Central to this innovation is the Cross-modal Affinity-guided Modality Alignment (CAMA) module, which utilizes affinity matrices to model relationships between different modalities and applies contrastive learning to maintain consistency across them. This approach effectively minimizes the influence of less informative modalities, such as eye gaze, without eliminating them. Additionally, modality-specific projection layers are employed to translate diverse features. This research tackles the issue of modality degradation in educational contexts where the informativeness of individual modalities varies among student groups. The findings can be accessed on arXiv under ID 2605.16806.
Key facts
- AAMLA framework predicts student collaboration satisfaction in game-based learning
- CAMA module models inter-modal relationships via affinity matrices
- Contrastive learning enforces cross-modal consistency
- Adaptive suppression of uninformative modalities without discarding
- Modality-specific projection layers map heterogeneous features
- Addresses modality degradation in educational deployments
- Eye gaze exhibits inconsistent informativeness across student cohorts
- arXiv ID: 2605.16806
Entities
Institutions
- arXiv