Stability and discrimination trade-offs of fairness mitigation

Establish whether bias-mitigation steps, including reweighing and per-group threshold shifts, yield a stable passing fairness verdict without sacrificing model discrimination as measured by AUROC or AUPRC.

Background

Fairness audits convert continuous metrics into binary pass-or-fail decisions, but uncertainty in the metric does not automatically establish whether the resulting verdict will remain unchanged across repeated audits. Bias-mitigation procedures may improve a point-estimate fairness metric while producing verdicts that remain sensitive to cohort variation.

The unresolved issue is whether mitigation procedures can simultaneously secure a stable passing verdict and preserve predictive discrimination, rather than improving fairness only at the cost of AUROC or AUPRC. The paper evaluates reweighing and threshold shifting on the Texas-100X hospital length-of-stay dataset, but the broader trade-off remains an open issue beyond the reported cohort and experimental setting.

References

Existing uncertainty methods also leave open whether bias-mitigation steps, such as reweighing or per-group threshold shifts, yield a stable passing verdict at the cost of model discrimination measured as AUROC or AUPRC.

— VFR-Audit: Verdict-Level Reliability for Fairness Audits in Hospital Length-of-Stay Prediction  (2608.30846 - Joy et al., 31 Aug 2026) in Abstract

Several substantive questions consequently remain open. First, the present cases identify residual differences more clearly than the mechanisms producing them. Future work should manipulate interaction trajectories prospectively while holding AI capability as constant as possible, allowing the causal effects of timing, feedback, disagreement, explanation, or reconstruction to be isolated. Second, later human outcomes should be observed across multiple horizons so that immediate transfer, delayed retention, attenuation, and recovery are not conflated. Third, domain-specific measurement should separate latent mechanisms where aggregate performance is ambiguous, including the distinction between discrimination and decision criterion in diagnostic settings.

— What Human-AI Interaction Leaves Behind: Residual Human States -- A Trajectory-Based Framework and Multi-Study Analysis  (2610.18765 - Herrera-Poyatos et al., 13 Oct 2026) in Section 6, subsection “Limitations and future research”