The xOrder fairness method proved inapplicable at institutional scale, with runtime scaling of approximately n^2.36¶

Claim · All claims
Evidence · 1 study · causal 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? One of seven initially selected methods, xOrder, was excluded because it is defined only for binary sensitive attributes and its dynamic program is computationally intractable at institutional scale; the article treats this inapplicability as a finding. → McConvey 2026

Evidence¶

McConvey 2026¶

McConvey, K., Zhai, A., Li, R., & Guha, S. (2026). Fairness Theatre: Evaluating Post-Hoc Fairness Interventions in Vendor-Controlled Early Warning Systems. arXiv. https://arxiv.org/abs/2609.38552 doi:10.48550/arxiv.2609.38552

q2 · i? · causal · r2

A benchmarking analysis of the xOrder method on the institution's data showed intractable scaling: "runtime scaling of approximately 𝑛2.36," extrapolating to months of computation and over one hundred gigabytes of memory for one domestic-international comparison. The method is also formally defined only for binary sensitive attributes, while two of the study's three attributes are not binary.

"benchmarking it on our data yields runtime scaling of approximately 𝑛2.36, which extrapolates to months of computation and over one hundred gigabytes of memory for a single domestic–international comparison at full cohort size."

Discussion¶

Related Claims¶

• Post-hoc fairness interventions on a vendor-controlled EWS redistributed disparities across demographic groups without consistently reducing them — related
• Two of six evaluated interventions directed corrections toward already-advantaged groups because disadvantage was operationalized using group size — related
• Runtime GenUI adaptation is too late, too costly, and accuracy-risky to be equitable at scale for learners needing non-default representations — related
• When demographic groups have different base rates, no method can satisfy calibration, equalized odds, and statistical parity simultaneously — related