Identifying Student Profiles Within Online Judge Systems Using Explainable Artificial Intelligence

AI can provide meaningful educational insights beyond pass/fail grades: The study demonstrates that machine learning models can analyze student submission patterns in online programming systems to identify behavioral profiles and provide richer feedback to both students and instructors, moving beyond simple binary success/failure assessments.

Explainability is essential for educational AI applications: The researchers specifically incorporated explainable AI (XAI) techniques to ensure that the feedback generated by their model is "human-understandable," recognizing that opaque AI recommendations would be inappropriate in educational contexts where students and teachers need to understand the reasoning behind assessments.

Student behavior patterns are predictive of learning outcomes: The model successfully predicted whether students would pass or fail assignments based solely on their submission patterns (tested on ~2,500 submissions from 90 students), suggesting that how students approach problem-solving—not just their final code—contains valuable information about their learning trajectory.

Responsible AI requires considering multiple stakeholders: The research design explicitly addresses the needs of both students (who benefit from personalized feedback) and instructors (who gain insights into class-wide learning patterns), demonstrating that educational AI systems should be designed with multiple user perspectives in mind rather than optimizing for a single outcome.