EXPLAINABLE AI FOR STUDENTS PERFORMANCE PREDICTION SYSTEM

Abstract

In the rapidly evolving landscape of educational technology (EdTech), data-driven systems hold immense promise for revolutionizing personalized learning and proactive student support. Specifically, Artificial Intelligence (AI) models designed to predict student performance—identifying those at risk of academic failure or course dropout—are becoming integral tools for institutional administrators and educators. However, the adoption of these powerful predictive systems often meets a significant barrier: the “black box” problem. Traditional, high-accuracy models, such as deep neural networks or complex ensemble methods, function opaquely, delivering a prediction score without any corresponding justification. When an AI labels a student as "at risk," the lack of insight into why that determination was made severely limits the efficacy and ethical acceptance of the system.

This abstract outlines the critical necessity and practical implementation of an Explainable AI (XAI) framework integrated directly into a student performance prediction system. We argue that in an environment as delicate and high- stakes as education, high accuracy is insufficient; transparency is paramount to building trust and driving positive human intervention. Without clear explanations, teachers are forced to rely blindly on an algorithm, which undermines their professional judgment, prevents the identification of systemic biases within the data, and, most importantly, fails to provide actionable insights needed for customized student support.