Study maps how students use a GAI-augmented learning analytics tool in programming education

Generative artificial intelligence is reshaping how students learn to program, but researchers still know little about how learners actually use the dashboards and peer information that accompany these tools. A new study published in Humanities and Social Sciences Communications examines exactly that question, analyzing how students engaged with a learner-facing learning analytics tool connected to a GAI-supported programming platform.

The research team, led by Dan Sun and including Zhanshan Yang, Jiayue Shi, and Junfeng Yang, adopted a multi-method approach. They combined fine-grained behavioral trace data with measures of interaction quality and students’ perceptual evaluations of the GAI system. The goal was to understand how students engage with self-oriented resources, which show their own learning information, and peer-oriented resources, which expose how classmates are progressing.

Using stage-specific clustering of the behavioral data, the study identified three initial engagement profiles: Deep Explorers, Self-Oriented Explorers, and Peer-Oriented Explorers. At a later stage, an additional profile emerged, labeled Limited Explorers. This stage-specific structure suggests that students' engagement patterns differ depending on the stage of engagement with the learning environment, although the study did not track how individual students transition between profiles over time.

Peer-generated information played a distinctive role in the analysis. The researchers treated it as a social reference frame that students could use for comparison and reflection. The clusters differed in how they engaged with self-generated versus peer-generated content, indicating that learners draw on these two types of information in distinct ways depending on their profile.

Perceptions of the GAI tool were generally favorable across the sample. Notably, however, these perceptions did not differ significantly across the final-stage profiles. In other words, students who explored the tool deeply and those who engaged only in limited ways held broadly similar views of the generative AI system, even though their behaviors diverged.

The authors argue that these findings underscore the importance of tailoring pedagogical support to learners’ stage-specific engagement profiles and their current task status. Rather than offering uniform guidance, instructors and system designers could adapt feedback to whether a student is a Deep Explorer, a Self-Oriented or Peer-Oriented Explorer, or a Limited Explorer at a given point in time.

To that end, the study concludes by proposing profile-informed, adaptive feedback and GAI-supported scaffolding as design directions for GAI-augmented programming education.

The authors also call for longitudinal research to trace how individual students transition between engagement profiles over time. Because the present study captures patterns at specific stages rather than tracking each learner’s trajectory, understanding those transitions remains an open question for future work in this emerging area of AI-supported education.

Subject of Research: Student engagement patterns with a GAI-augmented learning analytics tool in programming education

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

Article References: Sun, D., Yang, Z., Shi, J., & Yang, J. (2026). Characterizing diverse student behaviors, interactions, and perceptions with GAI-augmented learning analytics tool. Humanities and Social Sciences Communications. https://doi.org/10.1057/s41599-026-09564-z

Image Credits: AI Generated

DOI: 10.1057/s41599-026-09564-z

Keywords: generative artificial intelligence, learning analytics, programming education, student engagement, clustering, peer learning, GAI, educational technology, adaptive feedback, scaffolding, Characterizing, diverse