Results of single-time measurement.

Student engagement assessment in online learning faces critical limitations: traditional methods fail to capture engagement’s dynamic, multidimensional nature, particularly in second language (L2) contexts where emotional factors are paramount. This study introduces a novel multi-method framework that combines real-time facial expression recognition with dynamic self-reporting and observational measures to provide comprehensive, temporally-sensitive engagement assessment in synchronous online classrooms. Through analyzing a focused sample of Chinese L2 learners, we examined correlations between automatically detected happiness expressions and six established engagement measurements across both static and dynamic scales. Results revealed a significant and robust correlation between happiness expressions and self-reported emotional engagement, representing the study’s primary validated finding. Additional correlations were found with specific mood indicators (happy and loving items). However, no significant correlations were observed with behavioral engagement, cognitive engagement, or flow experience. The facial expression recognition successfully captured dynamic engagement fluctuations, showing modest but significant correlations with both retrospective self-reports and classroom observations. Our findings demonstrate that facial expressions serve as valuable indicators of specific engagement dimensions in online L2 learning, offering a targeted tool for emotional engagement assessment within comprehensive, multi-method evaluation frameworks in digital education.