Instability Floors: Separating Bias from Noise in Fairness Audits of Clinical LLM Agents with FairMedAgent
Abstract
Counterfactual fairness audits of clinical language-model agents report a flip rate: how often an action changes when only the patient's demographic descriptor changes. Part of that rate is not demographic. A stochastic agent also changes its own action when nothing changes, and a flip rate cannot be interpreted without knowing how often. We measured it. Re-running one condition ten times over sixteen synthetic vignettes at default sampling changed a clinical agent's action in 8.7 percent of replicate pairs, from 2.2 percent for intensive-care escalation to 17.9 percent for controlled-substance caution, an output given no operational criteria. Across six models from five vendors, pooled floors ranged from 2.5 to 23.7 percent; in this panel neither disclosed size, vendor, nor hosting ordered them. The floor depends on the decoding configuration: majority voting over five draws removed 39 percent of it (95 percent confidence interval (CI) 18 to 64); at temperature 0 three of four locally served models showed no disagreement, but a hosted model still did. We also show that, for a binary action, the flip rate expected under no demographic effect equals the floor and a real effect adds only its square, so a flip rate inside the floor is not evidence of fairness, and direction must be tested with a signed paired test. We give a four-step reporting procedure and release FairMedAgent, the harness, with its protocol, vignettes and analysis scripts, so any team can measure the floor for its own agent.
Related
- Consumption and Investment When Welfare Floors Change over Time
- Existence of Minimizers for the Gaussian Mumford-Shah Model via Covariance Floors on the Value Distribution
- Beyond Parallel Blindness: Information Floors and Model Gaps in Block Drafting
- The Role of Uncertainty in Assessing the Fairness of Machine Learning Models
- Equality, Equity, and Causality in Fairness Research: A Commentary on Cheng (2026)
BibTeX
@misc{bellibatlu2026instability,
title = {{Instability Floors: Separating Bias from Noise in Fairness Audits of Clinical LLM Agents with FairMedAgent}},
author = {Rohith Reddy Bellibatlu and Manpreet Singh and Deepak Parashar and Rahul Joshi},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.03221},
eprint = {2609.03221},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.03221},
}
RIS
TY - UNPB
TI - Instability Floors: Separating Bias from Noise in Fairness Audits of Clinical LLM Agents with FairMedAgent
AU - Rohith Reddy Bellibatlu
AU - Manpreet Singh
AU - Deepak Parashar
AU - Rahul Joshi
PY - 2026
DA - 2026/09/29
PB - arXiv
DO - 10.48550/arxiv.2609.03221
UR - https://arxiv.org/abs/2609.03221
AB - Counterfactual fairness audits of clinical language-model agents report a flip rate: how often an action changes when only the patient's demographic descriptor changes. Part of that rate is not demographic. A stochastic agent also changes its own action when nothing changes, and a flip rate cannot be interpreted without knowing how often. We measured it. Re-running one condition ten times over sixteen synthetic vignettes at default sampling changed a clinical agent's action in 8.7 percent of replicate pairs, from 2.2 percent for intensive-care escalation to 17.9 percent for controlled-substance caution, an output given no operational criteria. Across six models from five vendors, pooled floors ranged from 2.5 to 23.7 percent; in this panel neither disclosed size, vendor, nor hosting ordered them. The floor depends on the decoding configuration: majority voting over five draws removed 39 percent of it (95 percent confidence interval (CI) 18 to 64); at temperature 0 three of four locally served models showed no disagreement, but a hosted model still did. We also show that, for a binary action, the flip rate expected under no demographic effect equals the floor and a real effect adds only its square, so a flip rate inside the floor is not evidence of fairness, and direction must be tested with a signed paired test. We give a four-step reporting procedure and release FairMedAgent, the harness, with its protocol, vignettes and analysis scripts, so any team can measure the floor for its own agent.
ER -