The Audit Decides the Verdict: Instrument Effects Rival Demographic Bias in LLM Decision Audits

• URL: http://arxiv.org/abs/2609.09048v1
• Date: Tue, 08 Sep 2026 17:08:53 GMT
• Title: The Audit Decides the Verdict: Instrument Effects Rival Demographic Bias in LLM Decision Audits
• Abstract summary: A charitable-aid benchmark reports that the same models favor minority applicants when rating requests one at a time and penalize some when ranking side by side.<n>We test whether that reversal generalizes to hiring, lending, and medical triage.
• Score: 2.347309464574674
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Whether a language model looks demographically biased can depend on how the audit asks its question. A charitable-aid benchmark reports that the same models favor minority applicants when rating requests one at a time and penalize some when ranking side by side. We test whether that reversal generalizes to hiring, lending, and medical triage: 40,726 requests to five models, applications differing only in the applicant's name, and a primary test fixed before collection. It does not. None of 36 planned contrasts survives correction. The rating advantage keeps its sign at roughly half the published size, and a precision extension bounds any hiring ranking penalty below the published effect, though the lending and triage ranking floors sit above that margin, so the exclusion is conclusive for hiring ranking and for rating in all three domains only. Planted disparities tracking their injected sizes and a directional replication on the original aid materials bound these nulls. The audit is livelier than the demographics: models recognize transparent audits nearly always, tie every identical-content comparison whether the varying detail is race or a hobby, and reward first-listed candidates as much as any demographic effect we measure. Audit verdicts reflect audit construction more than demographic bias.

Related papers

• FairFund-Bench: Evaluating Distributive Bias in LLM Resource Allocation [0.0]
  We introduce FairFund-Bench, a benchmark that systematically varies key features of previous audit designs.<n>We show that this disagreement can arise from differences in audit format and introduce FairFund-Bench.<n>The benchmark scores models on four criteria (demographic bias, deservingness alignment, cross-task consistency, and cross-context consistency) is publicly available and can be readily adapted to other substantive domains.
  arXiv  Detail & Related papers  (2026-07-31T01:29:16Z)
• The C-index illusion: discrimination without calibration in published survival models [0.0]
  We reproduce three published survival-ML models across three structurally distinct domains.<n>We test five pre-registered hypotheses under a Holm-corrected family-wise error rate.
  arXiv  Detail & Related papers  (2026-07-21T19:13:54Z)
• AgentFairBench: Do LLM Agents Discriminate When They Act? [2.3004655342211078]
  AgentFairBench is a cheap, reproducible, multi-domain benchmark for demographic disparity in the actions of LLM agents.<n>It spans three regulator-anchored domains: hiring, lending, and medical triage.<n>A NumPy-only harness computes counterfactual flip rate, mean absolute score difference (MASD), action-rate disparity, and tool-invocation disparity.
  arXiv  Detail & Related papers  (2026-06-15T13:50:26Z)
• An External Fairness Evaluation of LinkedIn Talent Search [55.18656975953939]
  We conduct an independent, third-party audit for bias of LinkedIn's Talent Search ranking system.<n>We focus on potential ranking bias across two attributes: gender and race.<n>Our analysis reveals an under-representation of minority groups in early ranks.
  arXiv  Detail & Related papers  (2025-11-13T19:10:49Z)
• Hearing the Order: Investigating Selection Bias in Large Audio-Language Models [51.69003519291754]
  Large audio-language models (LALMs) are often used in tasks that involve reasoning over ordered options.<n>In this paper, we identify and analyze this problem in LALMs.
  arXiv  Detail & Related papers  (2025-10-01T08:00:58Z)
• Reference-Free Rating of LLM Responses via Latent Information [53.463883683503106]
  We study the common practice of asking a judge model to assign Likert-scale scores to free-text responses.<n>We then propose and evaluate Latent Judges, which derive scalar ratings from internal model signals.<n>Across a broad suite of pairwise and single-rating benchmarks, latent methods match or surpass standard prompting.
  arXiv  Detail & Related papers  (2025-09-29T12:15:52Z)
• Robustly Improving LLM Fairness in Realistic Settings via Interpretability [0.16843915833103415]
  Anti-bias prompts fail when realistic contextual details are introduced.<n>We find that adding realistic context such as company names, culture descriptions from public careers pages, and selective hiring constraints induces significant racial and gender biases.<n>Our internal bias mitigation identifies race and gender-correlated directions and applies affine concept editing at inference time.
  arXiv  Detail & Related papers  (2025-06-12T17:34:38Z)
• Auditing for Bias in Ad Delivery Using Inferred Demographic Attributes [50.37313459134418]
  We study the effects of inference error on auditing for bias in one prominent application: black-box audit of ad delivery using paired ads.<n>We propose a way to mitigate the inference error when evaluating skew in ad delivery algorithms.
  arXiv  Detail & Related papers  (2024-10-30T18:57:03Z)
• The Decisive Power of Indecision: Low-Variance Risk-Limiting Audits and Election Contestation via Marginal Mark Recording [51.82772358241505]
  Risk-limiting audits (RLAs) are techniques for verifying the outcomes of large elections. We define new families of audits that improve efficiency and offer advances in statistical power. New audits are enabled by revisiting the standard notion of a cast-vote record so that it can declare multiple possible mark interpretations.
  arXiv  Detail & Related papers  (2024-02-09T16:23:54Z)
• Large Language Models are not Fair Evaluators [60.27164804083752]
  We find that the quality ranking of candidate responses can be easily hacked by altering their order of appearance in the context. This manipulation allows us to skew the evaluation result, making one model appear considerably superior to the other. We propose a framework with three simple yet effective strategies to mitigate this issue.
  arXiv  Detail & Related papers  (2023-05-29T07:41:03Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.