Repeated subject-level split experiments confirm that validation-level fairness improvements from MAE-gap regularization do not correspond to consistent generalization to unseen subjects¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? Across multiple re-drawn subject-level train-validation-test splits with paired baseline and fairness-model comparisons, validation improvements achieved by fairness regularization did not correspond to consistent generalization to unseen subjects. → Christoforos Fragkiadakis 2026

Evidence¶

Christoforos Fragkiadakis 2026¶

Christoforos Fragkiadakis, Seyed Sahand Mohammadi Ziabari, and Ali Mohammed Mansoor Alsahag. (2026). Fairness-Aware Multimodal Transformer Modeling for Real-Time Student Attention Estimation. https://arxiv.org/abs/2609.02232 doi:10.48550/arxiv.2609.02232

q2 · i? · design · r2

Robustness experiment in which the dataset was partitioned multiple times at the subject level with stratified demographic sampling; the same subject partitions were shared across baseline and fairness models for paired comparisons within each split.

"The results were consistent with the main experiments and further supported that the validation improvements achieved by fairness regularization did not correspond to consistent generalization to unseen subjects (Appendix B)."

Discussion¶

Related Claims¶

• Demographic MAE-gap regularization reduces targeted subgroup gaps on validation data, but these fairness gains do not consistently transfer to held-out test subjects — possibly the same claim (merge candidate)
• The Residual Fusion Transformer shows the lowest worst-group MAE and smallest gender gap among baselines, while the Sensor GRU shows the smallest age gap alongside the weakest overall performance — related