Standard ML models predicting student pass/fail from VLE data exhibit biased true-positive rates across demographic subgroups¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2025 · checked against its sources 2026-10-17 (automated, 1 of 1 entry) · not reviewed by a person · recheck by 2028: tool and model accuracy changes quickly

Subclaims¶

q2 i? A baseline logistic-regression classifier trained on OULAD showed unequal true-positive rates across gender, age, IMD, and disability subgroups, indicating disparate impact; the authors say the same holds for standard models such as XGBoost but show subgroup rates only for LR. → Raymond A. Opoku 2025

Evidence¶

Raymond A. Opoku 2025¶

Raymond A. Opoku, Bo Pei and Wanli Xing. (2025). Unveiling Accuracy-Fairness Trade-Offs: Investigating Machine Learning Models in Student Performance Prediction. Journal of Learning Analytics 12(2). https://doi.org/10.18608/jla.2025.8543

q2 · i? · design · r2

Baseline LR and XGBoost classifiers were trained on OULAD (25,690 students); group-specific TPRs are shown in Figure 6 for the LR classifiers only (baseline and bias-mitigated), with Tukey's range tests for pairwise TPR/FPR differences. The discussion says standard approaches "such as LR and XGBoost, can exhibit biased performance", but no subgroup TPRs are shown for XGBoost. The article reports "unequal true positive rates (TPRs)" across subgroups. No effect size is printed.

"The plots reveal biases in the baseline classifier performance across different subgroups. As evident from the varying prevalence rates and unequal true positive rates (TPRs). These biases suggest disparate impact based on factors such as sex, age, index of multiple deprivation and disability."

Discussion¶

Related Claims¶

• Preprocessing bias mitigation (DIR, RW, SUP) reduced subgroup disparities in TPR on OULAD, with effectiveness varying by demographic attribute — related
• Calibrated Equalized Odds Post-processing showed lower balanced accuracy and higher EOD than the preprocessing techniques DIR and RW in student performance prediction — related
• Baseline LR and XGBoost models achieve comparable predictive performance on OULAD, with XGBoost slightly higher AUC-ROC — related
• Post-hoc fairness interventions on a vendor-controlled EWS redistributed disparities across demographic groups without consistently reducing them — related
• Machine learning methods outperformed logistic regression in classification accuracy across reviewed studies — related
• Using demographic variables as predictors risks reinforcing biases embedded in training labels, including self-fulfilling prophecies — related
• Bias in predicting learning productivity, measured as accuracy differences across gender and parental educational background, was highly dependent on the ML method used — related