Counterfactual Fairness Audits of Multi-Step Clinical LLM Agents Require a Measured Per-Action Instability Floor

• URL: http://arxiv.org/abs/2609.03221v2
• Date: Fri, 04 Sep 2026 13:47:56 GMT
• Title: Counterfactual Fairness Audits of Multi-Step Clinical LLM Agents Require a Measured Per-Action Instability Floor
• Abstract summary: We show that a counterfactual fairness estimate reported without a per-action floor beside it cannot be read as evidence of disparity.<n>The measurements were taken with FairMedAgent, an evaluation harness for disparity in the actions of clinical adjudication agents.
• Score: 3.0704370207786265
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Counterfactual audits are the standard tool for checking whether a clinical agent treats demographically distinct but clinically identical patients differently. They report a flip rate: how often an action changes when only the patient descriptor changes. We show that this quantity is uninterpretable on its own. Re-running an identical condition ten times over sixteen vignettes (same narrative, same descriptor string, nothing varied) moved a clinical agent's action in 8.7% of outcome-vignette cells, and instability was heterogeneous across actions by a factor of eight, from 0.022 for ICU escalation to 0.179 for controlled-substance caution. No demographic contrast in our data was distinguishable from that floor. A second model gives a pooled floor of 6.7% and ranks the six actions almost identically (Spearman 0.94, exact p=0.017), so the floor is not one system's artefact. Majority-vote aggregation over five draws removes 39% of it and then flattens, and a null simulation attributes the residue to heterogeneous per-cell rates, so replication mitigates without eliminating. Any counterfactual fairness estimate reported without a per-action floor beside it therefore cannot be read as evidence of disparity. The measurements were taken with FairMedAgent, an evaluation harness for disparity in the actions of clinical LLM agents whose estimand, the within-range counterfactual flip rate, counts only flips between actions a published decision rule admits and a clinician has adjudicated. That estimand requires band adjudication, which is under way; no disparity result is claimed here. Each synthetic vignette runs a six-stage trajectory (five model-facing decisions around a deterministic environment step) under fixed-form conditions spanning race, sex, age, insurance, English proficiency, and their intersections. The harness, the floor protocol, and every analysis script are released.

Related papers

• The Model Proposes, the Code Disposes: A Pre-Registered Ablation of a Verifier-and-Acceptance Stage in an LLM-Orchestrated Offensive-Security Agent [0.0]
  We evaluate whether a verifier-and-acceptance stage - a model verifier whose verdicts are enforced by deterministic code - changes what an LLM-driven offensive-security agent reports.<n>We report a 15-run exploratory pilot, a pre-registered 20-run confirmatory ablation, and a pre-registered 2 x 2 factorial study with 40 runs across two deliberately vulnerable lab targets.<n>The results support a narrow conclusion: the verifier changes what the system ships, while deterministic code supplies enforcement and auditability.
  arXiv  Detail & Related papers  (2026-09-14T17:13:05Z)
• A Judge Should Know What Changed:Construct Validity for LLM-as-a-Judge Evaluation [14.867869624585886]
  We formalize construct validity for an evaluator as a two-dimensional profile.<n>We show that S and R are independent and that no scalar summary preserves all relevant comparisons.<n>These results show that high judge agreement can coexist with weak sensitivity to changes in the construct being evaluated.
  arXiv  Detail & Related papers  (2026-08-25T11:32:54Z)
• Passing Coarse Marginal Checks Can Be Cheap: Persona Mixtures and Imprecise Treatment-Response Estimates in an LLM Persona Panel [0.0]
  We study a fixed panel of sixteen lightweight persona-conditioned GPT-4.1 configurations in repeated strategic games.<n>Variation was strongly prompt-indexed, but its share depended on uncertainty assumptions.<n>Results concern one fixed model-prompt panel and do not establish human substitutability.
  arXiv  Detail & Related papers  (2026-08-02T04:10:11Z)
• KAISEN: Reproducible Subgroup Fairness Auditing for Clinical Risk Models [0.0]
  KAISEN is a five-phase audit pipeline covering subgroup stratification, disparity measurement, mechanism diagnostics, post-hoc mitigation, and drift monitoring.<n>All results are synthetic with known ground truth and do not establish clinical validity.
  arXiv  Detail & Related papers  (2026-07-30T17:57:18Z)
• RareLens: Towards End-to-End Rare Disease Care via Aligning Divergent Large Language Model Reasoning [42.80742579670377]
  RareLens is a system that supports clinical decision-making across the entire rare disease trajectory.<n>Developed and evaluated on RareBench, a real-world dataset of 157,525 cases spanning all 33 Orphanet categories and more than 7,000 conditions.<n>It achieved an area under the curve of 0.917 for screening and top-1 accuracies of 65.5% and 89.8% for diagnosis and treatment.
  arXiv  Detail & Related papers  (2026-07-25T16:58:10Z)
• Enhancing Clinician Decision-Making via Uncertainty-Aware Multi-Expert Fusion for Stroke Rehabilitation [0.32553561239735207]
  We present xAARA: an engine designed to augment rather than replace clinical judgment.<n>xAARA returns ARAT assessments with calibrated uncertainty and explanations across task, movement-phase, and movement-quality levels.<n>In 105 stroke survivors (788 exercises), xAARA achieved 94.2% task accuracy and 81.3% movement-phase accuracy.
  arXiv  Detail & Related papers  (2026-06-23T09:00:15Z)
• 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)
• Layer-Isolated Evaluation: Gating the Deterministic Scaffold of a Production LLM Agent with a No-LLM, Regression-Locked Test Harness [45.148328075418156]
  A deployed ordering agent is decomposed into a fixed taxonomy of layers.<n>The pure suite runs in CI on every change against a locked per-slice baseline.<n>We validate by controlled regression injection, degrading one layer at a time across seven non-safety layers.
  arXiv  Detail & Related papers  (2026-06-10T05:55:13Z)
• Compared to What? Baselines and Metrics for Counterfactual Prompting [39.56472929066589]
  We observe prediction flip rates on MedQA of 14.9% when we surgically change patient gender.<n>We propose a framework in which we compare differences observed under target interventions to those induced by paraphrasing inputs.<n>We find that these effects largely dissipate when we account for general model sensitivity.
  arXiv  Detail & Related papers  (2026-05-01T19:23:33Z)
• A Multi-Domain Red Teaming Framework for Safety, Robustness, and Fairness Evaluation of Medical Large Language Models [34.38896321208398]
  Large language models (LLMs) are increasingly deployed across healthcare, yet existing benchmarks fail to capture model behavior under adversarial or ethically complex conditions.<n>We developed a multi-domain red teaming framework evaluating eleven contemporary LLMs across 690 clinically grounded scenarios.<n>Results revealed substantial performance variance, with mean scores ranging from 0.791 to 0.984.
  arXiv  Detail & Related papers  (2026-04-15T14:57:55Z)
• From Black Box to Glass Box: Cross-Model ASR Disagreement to Prioto Review in Ambient AI Scribe Documentation [43.148402136307716]
  Cross-model disagreement among heterogeneous ASR systems can act as a reference-free uncertainty signal.<n>We transcribed 50 publicly available medical audio clips with eight ASR systems spanning commercial APIs and open-source engines.<n>Low-agreement regions were enriched for content disagreements, with the content fraction increasing from 53.9% to 73.9% across quintiles of high-risk mass.
  arXiv  Detail & Related papers  (2026-03-02T13:02:13Z)
• Correcting Class Imbalances with Self-Training for Improved Universal Lesion Detection and Tagging [43.06199185109424]
  Universal lesion detection and tagging (ULDT) in CT studies is critical for tumor burden assessment and tracking the progression of lesion status (growth/shrinkage) over time.<n>Prior work used the DeepLesion dataset (4,427 patients, 10,594 studies, 32,120 CT slices, 32,735 lesions, 8 body part labels) for algorithmic development, but this dataset is not completely annotated and contains class imbalances.<n>We developed a self-training pipeline for ULDT using a limited 11.5% subset of DeepLesion.
  arXiv  Detail & Related papers  (2025-04-07T15:57:03Z)

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.