Unveiling the sequential pattern of group cognitive engagement in GAI-enhanced learning environment with first-order Markov model

As we enter the era of “Hybrid Intelligence”, it becomes increasingly important for educators to explore how collaborative learning can become more efficient with the support of Generative AI (GAI). Collaborative learning is a scenario where complex cognitive engagement occurs among group members. Therefore, by analyzing the distribution characteristics and temporal patterns of group cognitive engagement in GAI-enhanced environments, we can gain a deeper understanding of how the intervention of GAI shapes the temporal patterns of cognitive engagement within groups. Furthermore, by comparing high-performing (HP) groups with low-performing (LP) groups, we can gain a clearer understanding of the temporal patterns of cognitive engagement that contribute to better collaboration performances in GAI-enhanced settings. However, current research has yet to address this issue. To bridge this gap, this study primarily employs the process mining method—the first-order Markov model (FOMM)—to examine differences in the distribution characteristics and temporal patterns of cognitive engagement between HP groups and LP groups in a self-designed GAI-enhanced collaborative learning environment. For this purpose, 39 undergraduate students from a key university in China were recruited to participate in an online collaborative writing (OCW) task. The results reveal significant differences between HP and LP groups. Specifically, HP groups engaged in discussions related to specific writing knowledge inspired by the content generated by the GAI. Moreover, HP groups were more actively involved in articulating, elucidating, and analyzing their ideas during collaboration, while LP groups were more inclined to engage in process monitoring intermittently throughout the task. This study not only provides valuable theoretical insights into how GAI’s intervention affects the temporal patterns of group cognitive engagement, but also offers guidance for designing effective scaffolding strategies to promote more efficient collaborative learning outcomes in GAI-enhanced settings.