Responsible Learning Analytics for Identifying and Supporting Academically At-Risk Students in Secondary Schools

Abstract

Learning analytics may help secondary schools identify students who need additional support, but prediction alone does not improve educational outcomes. This integrative review synthesizes 289 publications on predictive modeling, educational measurement, student support, fairness, privacy, and implementation. The analysis examines predictive validity, data quality, fairness and bias, intervention design, privacy and consent, and implementation capacity. The evidence shows that early-warning systems fail when missing or delayed data distort risk estimates, subgroup performance is hidden by aggregate accuracy, alerts lack actionable explanations, or schools lack resources to respond. Ethical implementation also requires proportionate data use, professional review, student recourse, and monitoring for unintended effects. The paper contributes a prediction-to-support framework that separates risk estimation from intervention decisions and outcome evaluation. It also specifies indicators and propositions for testing whether learning analytics produces timely, fair, and effective academic support.