A complex dynamical system approach to student engagement

Background

Multimodal data analysis has been approached through three main avenues: (1) joint effect approach, (2) triangulation approach, and (3) separate latent construct approach. While these approaches have advanced our understanding of the learning process, they fail to capture its dynamic and emergent nature.

Aim

This study examines multimodal data through the lens of complex dynamical system (CDS) approach. We investigated whether a CDS approach could provide unique insights into predicting and understanding cognitive engagement during learning.

Sample

The participants comprised 61 third-year medical students (47.5 % females).

Methods

From a CDS perspective, we analyzed eye gaze, head pose, and facial action units of participants engaged in an interactive learning environment.

Results

We found that specific parameters of eye gaze, head pose, and facial expressions significantly predicted cognitive engagement levels. Network density was also identified as a significant predictor of cognitive engagement. Notably, network density explained a greater proportion of the variation in cognitive engagement compared to any other individual variable considered. Additionally, we found that students in the low engagement group demonstrated consistently weak but stable interconnections among behavioral indicators, while the high engagement group displayed tightly clustered interaction patterns among variables.

Conclusions

These findings highlight the added value of a CDS approach for modeling the dynamic complexity of cognitive engagement. This study represents a significant step in rethinking the research agenda in multimodal learning analytics. Methodologically, this study demonstrates the potential of CDS-based analytical techniques for gaining insights into physiological and psychological processes underlying engagement and learning.