Abstract:

The increasing use of learning analytics in higher education has created new opportunities for understanding student learning processes and supporting data-informed decisions. At the same time, concerns regarding transparency and interpretability have emerged, particularly when analytics rely on complex models. This paper examines the potential of explainable artificial intelligence (XAI) to enhance transparency and trust in learning analytics systems.

Explainable artificial intelligence is defined as a set of approaches that make algorithmic outputs understandable to users by clarifying how specific indicators contribute to predictions or recommendations. In educational contexts, such explanations may support both instructors and students in interpreting learning data and adjusting practices accordingly.

The study adopts a mixed-methods design implemented during the 2023–2024 academic year in three undergraduate online courses at a Romanian university, involving 132 students and 5 instructors. An explainable learning analytics dashboard was developed and integrated into the learning management system. The dashboard displayed key indicators (e.g., platform activity, assignment submission patterns, and engagement levels) alongside explanatory elements showing how these variables contributed to predicted performance categories.

Quantitative data included platform usage logs and assessment results collected before and after the introduction of the dashboard (two course iterations). Descriptive and comparative analyses were used to examine changes in engagement (e.g., frequency of access) and performance indicators. Qualitative data were collected through 10 semi-structured interviews with students and 5 with instructors, focusing on system usability and perceived impact. Data were analyzed using thematic analysis.

Results suggest moderate positive effects associated with the use of explainable analytics. Platform data showed a 12% increase in regular access among students identified as previously low-engagement, while timely assignment submission rates increased from 72% to 80%. Interview data indicated that students used explanatory features to better understand how their behaviors influenced performance, with several reporting adjustments in study routines (e.g., more consistent access to materials). Instructors reported improved confidence in interpreting analytics outputs and used the dashboard to support discussions about learning expectations.

At the same time, some participants reported difficulty interpreting multiple indicators simultaneously, indicating potential cognitive overload. Differences in data literacy also influenced how effectively the dashboard was used.

The study suggests that explainable analytics can enhance transparency and support reflective engagement with learning data. However, their effectiveness depends on careful interface design and user support. XAI should therefore be approached not only as a technical solution, but as a pedagogical tool integrated with data literacy and instructional practices.

Keywords:

Explainable artificial intelligence, learning analytics transparency, higher education ethics, data literacy, educational technology.