Construction and Empirical Study of a Learning Engagement Representation Model in GAI-Supported Collaborative Learning

In recent years, collaborative learning environments have garnered widespread attention for their capacity to promote learners’ active engagement in learning activities. However, issues such as low participation and free-riding remain prevalent in collaborative settings. With the rapid advancement of Generative Artificial Intelligence (GAI), its integration into collaborative learning shows significant potential, particularly in supporting personalized learning, thereby offering new opportunities to enhance student engagement. Despite this potential, limited research has focused on the representation and assessment of learning engagement within GAI-supported collaborative learning contexts. Grounded in the theories of collaborative learning engagement, self-regulated learning, and distributed cognition, this study constructs a digital representation model of learning engagement in GAI-supported collaborative learning. A corresponding set of quantifiable indicators is also proposed. The model is applied within a collaborative writing scenario, where learners’ GAI interaction logs, screen and audio recordings, and academic performance data are analyzed to explore patterns of learning engagement. The findings reveal that behavioral engagement follows a U-shaped trajectory, cognitive engagement exhibits a fluctuating upward trend, and emotional engagement increases steadily. Furthermore, behavioral, cognitive, and emotional engagement are intrinsically interrelated and collectively contribute to task performance, albeit with varying degrees and mechanisms of influence. Learners categorized into low-interaction high-engagement, high-interaction low-engagement, and high-interaction high-engagement groups displayed imbalanced engagement across dimensions. The proposed model offers a comprehensive framework for representing student learning engagement in GAI-supported collaborative learning, providing novel insights into the dynamic evolution of engagement under GAI-enhanced conditions.