<![CDATA[The increasing adoption of Learning Management Systems (LMS) in higher education generates learning traces that can support early detection of at-risk students. However, many academic early warning systems still present risk scores without explaining risk drivers or translating them into feasible instructor actions. This study designs an explainable and actionable artificial intelligence framework using a Design Science Research approach. The artifact integrates LMS data, learning feature engineering, risk prediction, SHAP-based explanation, constrained Large Language Model-based recommendation, and human-in-the-loop validation. Its novelty lies in connecting prediction, explanation, pedagogical recommendation, instructor validation, and post-intervention feedback within one ethically constrained Learning Analytics workflow. The design outputs include requirements, input-process-output architecture, risk-threshold logic, SHAP-LLM workflow, intervention taxonomy, recommendation template, and evaluation roadmap. The framework remains conceptual; future studies should instantiate it with real LMS data and evaluate predictive performance, explanation quality, recommendation usefulness, fairness, usability, and intervention impact.]]>
Copyrights © 2026
