Multidimensional Modeling of Multimodal Learner Engagement and Emotion during Game-Based Learning

• In Event: Designing Engaging Learning Environments: Insights from AI, VR, and Games

Abstract

Game-based learning environments (GBLEs) promise to develop complex scientific reasoning, yet not all engagement behaviours equally support learning. Guided by the Integrative Model of Multidimensional SRL Engagement (IMMSE), this study examines how behavioural and cognitive engagement relate to task emotions while undergraduate learners interact with Crystal Island, a microbiology GBLE using multimodal data including eye-tracking, facial expressions, and logfiles. We fit a structural equation model (SEM) with a latent Engagement factor predicting emotions. The model showed good fit, and all dwell-time indicators loaded significantly on Engagement. Higher engagement was associated with reduced confusion and frustration, and with lower joy, suggesting emotional stabilization during sustained information processing. Findings demonstrate the value of multimodal measures for designing adaptive, affect-sensitive GBLEs.

Authors

• Megan Wiedbusch, University of Central Florida

• Matthew Moreno, University of Central Florida

• Cameron Marano, University of Central Florida

• Milouni Patel, University of Central Florida

• James C. Lester, North Carolina State University

• Roger Azevedo, University of Central Florida