Characterizing diverse student behaviors, interactions, and perceptions with GAI-augmented learning analytics tool

Generative artificial intelligence (GAI)-enhanced learning analytics tools can support programming education by combining personalized assistance with access to individual and peer learning information. However, little is known about how students engage with self- and peer-oriented resources in these environments or how different patterns of engagement relate to the quality of their interactions with GAI and their perceptions of it. This study adopted a multi-method approach to examine students’ behavioral patterns while using a learner-facing learning analytics tool associated with a GAI-supported programming platform, along with the corresponding interaction qualities and perceptual evaluations. Using fine-grained behavioral trace data, stage-specific clustering identified three initial profiles (Deep, Self-Oriented, and Peer-Oriented Explorers), and an additional final-stage profile, Limited Explorers. Peer-generated information was examined as a social reference frame for comparison and reflection, and the clusters exhibited distinct patterns of engagement with self- and peer-generated content. Although perceptions of GAI were generally favorable, they did not differ significantly across the final-stage profiles. These findings underscore the importance of tailoring pedagogical support to learners’ stage-specific engagement profiles and current task status. The study concludes by proposing profile-informed, adaptive feedback and GAI-supported scaffolding, while calling for longitudinal research to trace individual transitions between engagement profiles in GAI-augmented programming education.