Supervised Contrastive Learning for Ordinal Engagement Measurement

TL;DR

This work addresses automated engagement measurement in virtual learning by tackling class imbalance and the ordinal nature of engagement levels. It proposes a novel framework that uses supervised contrastive learning with time-series augmentation and an ordinal classification scheme, incorporating a fusion of affective and behavioral features via a sequential encoder. Evaluated on the DAiSEE dataset, the approach shows improved class-specific performance and overall accuracy compared with prior methods, demonstrating robustness to imbalanced data and class confusion. The findings support deploying ordinal-contrastive models for scalable, video-based engagement monitoring and point to future multimodal extensions and applications in education and rehabilitation.

Abstract

Student engagement plays a crucial role in the successful delivery of educational programs. Automated engagement measurement helps instructors monitor student participation, identify disengagement, and adapt their teaching strategies to enhance learning outcomes effectively. This paper identifies two key challenges in this problem: class imbalance and incorporating order into engagement levels rather than treating it as mere categories. Then, a novel approach to video-based student engagement measurement in virtual learning environments is proposed that utilizes supervised contrastive learning for ordinal classification of engagement. Various affective and behavioral features are extracted from video samples and utilized to train ordinal classifiers within a supervised contrastive learning framework (with a sequential classifier as the encoder). A key step involves the application of diverse time-series data augmentation techniques to these feature vectors, enhancing model training. The effectiveness of the proposed method was evaluated using a publicly available dataset for engagement measurement, DAiSEE, containing videos of students who participated in virtual learning programs. The results demonstrate the robust ability of the proposed method for the classification of the engagement level. This approach promises a significant contribution to understanding and enhancing student engagement in virtual learning environments.

Figures (1)

• Figure 1: Supervised contrastive learning for a binary classification problem.