Evaluation of engagement in online learning: insights based on human factor analysis

Abstract

Although evaluating learner engagement in online learning is crucial for improving learning outcomes, the limited face-to-face interaction between educators and learners makes this evaluation difficult. Therefore, to address this issue, we analyze the relationship between human factors and learner engagement states. To this end, we design a human factor-based learner engagement evaluation framework to identify the factors that significantly impact learners' engagement states. Specifically, we first extract learners' eye-gaze angles, eye blinks, facial expressions, and facial landmarks as human factors. Next, we analyze the temporal features and evaluate learner engagement by combining these extracted human factors using state-of-the-art time series analysis models. In addition, we construct a learner engagement evaluation dataset, called recorded videos for student engagement (ReViSE). Extensive experimental analysis demonstrates that a combination of eye-gaze angles, eye blinks, and facial landmarks provides the best performance on the ReViSE dataset, whereas adding facial expressions yields the best performance on the DAiSEE dataset. Furthermore, we show that eye blinks and facial expressions are the most effective human factors for assessing online learner engagement.