Alharbi
Artificial intelligence is increasingly shaping many sectors, and education is no exception. The increasing deployment of machine learning (ML) models in education has introduced unprecedented opportunities for personalized learning, predictive analytics, and data-informed decision-making. Machine learning has started to influence the way we design and deliver education, offering tools that adapt to students’ needs and help institutions make decisions. Despite these opportunities, there are growing concerns about how such models actually make their predictions. However, the black-box nature of widely adopted models, such as XGBoost and deep learning, poses significant challenges to trust, transparency, and accountability. Black-box-type models do not provide explanations useful in high-stakes educational environments where placing or grading students on algorithmic outputs becomes a tangible reality. Educators often see the predictions but not the reasoning behind them, which makes it difficult to build trust or act confidently on the results. Explainable AI (XAI) offers a promising solution by enhancing the interpretability of AI-driven decisions, ensuring that educators can understand and trust model predictions. In this paper, we propose a framework that brings XGBoost together with SHAP (SHapley Additive exPlanations) to make its inner workings more transparent. Our aim is not only to improve performance but also to show why certain outcomes appear and how they can be understood in practice. The framework is developed with teachers and decision-makers in mind, since accuracy alone is rarely enough in real classrooms. We suggest that this approach can open the door to more responsible and practical uses of AI in education, while also giving researchers a clearer direction for future studies. Finally, the findings will contribute to the development of transparent, ethical, and trustworthy AI systems in education, bridging the gap between advanced ML techniques and practical educational applications.