Algorithmic Justice in Education Hybrid XGBoost–DNN for Real-Time Detection and Fairness Optimization in HR Decision-Making

Administrative justice plays a vital role in maintaining equity and trust within public education systems. However, implicit biases and discriminatory patterns often distort fairness in human resource management. This study introduces a hybrid XGBoost–Deep Neural Network (DNN) model for detecting administrative discrimination in educational institutions. Using a dataset of ۱۲,۴۸۰ administrative decisions from regional education departments, the proposed model achieved ۹۴.۶% accuracy, outperforming baseline algorithms by ۸.۷%. Statistical validation (p < ۰.۰۱, ۹۵% CI) confirms the model’s robustness and generalizability across diverse demographic and organizational variables. Feature importance analysis identified seniority (۳۴%), gender (۲۷%), and institutional hierarchy (۱۹%) as dominant fairness factors. The model’s interpretability, ensured through SHAP and LIME frameworks, enables transparent auditing of administrative decisions. This research provides a practical foundation for deploying AI-driven fairness analytics in public sector management and policy-making, fostering equity, accountability, and trust in educational administration.

Keywords:

Administrative justice , Machine learning fairness , XGBoost , Deep Neural Network , Bias detection , Educational management , SHAP , Algorithmic transparency