Audit Me If You Can: Query-Efficient Active Fairness Auditing of Black-Box LLMs

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Hartmann, David, Pohlmann, Lena, Hanslik, Lelia, Gießing, Noah, Berendt, Bettina, Delobelle, Pieter
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866914236371828736
author Hartmann, David
Pohlmann, Lena
Hanslik, Lelia
Gießing, Noah
Berendt, Bettina
Delobelle, Pieter
author_facet Hartmann, David
Pohlmann, Lena
Hanslik, Lelia
Gießing, Noah
Berendt, Bettina
Delobelle, Pieter
contents Large Language Models (LLMs) exhibit systematic biases across demographic groups. Auditing is proposed as an accountability tool for black-box LLM applications, but suffers from resource-intensive query access. We conceptualise auditing as uncertainty estimation over a target fairness metric and introduce BAFA, the Bounded Active Fairness Auditor for query-efficient auditing of black-box LLMs. BAFA maintains a version space of surrogate models consistent with queried scores and computes uncertainty intervals for fairness metrics (e.g., $Δ$ AUC) via constrained empirical risk minimisation. Active query selection narrows these intervals to reduce estimation error. We evaluate BAFA on two standard fairness dataset case studies: \textsc{CivilComments} and \textsc{Bias-in-Bios}, comparing against stratified sampling, power sampling, and ablations. BAFA achieves target error thresholds with up to 40$\times$ fewer queries than stratified sampling (e.g., 144 vs 5,956 queries at $\varepsilon=0.02$ for \textsc{CivilComments}) for tight thresholds, demonstrates substantially better performance over time, and shows lower variance across runs. These results suggest that active sampling can reduce resources needed for independent fairness auditing with LLMs, supporting continuous model evaluations.
format Preprint
id arxiv_https___arxiv_org_abs_2601_03087
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Audit Me If You Can: Query-Efficient Active Fairness Auditing of Black-Box LLMs
Hartmann, David
Pohlmann, Lena
Hanslik, Lelia
Gießing, Noah
Berendt, Bettina
Delobelle, Pieter
Machine Learning
Computation and Language
Computers and Society
Large Language Models (LLMs) exhibit systematic biases across demographic groups. Auditing is proposed as an accountability tool for black-box LLM applications, but suffers from resource-intensive query access. We conceptualise auditing as uncertainty estimation over a target fairness metric and introduce BAFA, the Bounded Active Fairness Auditor for query-efficient auditing of black-box LLMs. BAFA maintains a version space of surrogate models consistent with queried scores and computes uncertainty intervals for fairness metrics (e.g., $Δ$ AUC) via constrained empirical risk minimisation. Active query selection narrows these intervals to reduce estimation error. We evaluate BAFA on two standard fairness dataset case studies: \textsc{CivilComments} and \textsc{Bias-in-Bios}, comparing against stratified sampling, power sampling, and ablations. BAFA achieves target error thresholds with up to 40$\times$ fewer queries than stratified sampling (e.g., 144 vs 5,956 queries at $\varepsilon=0.02$ for \textsc{CivilComments}) for tight thresholds, demonstrates substantially better performance over time, and shows lower variance across runs. These results suggest that active sampling can reduce resources needed for independent fairness auditing with LLMs, supporting continuous model evaluations.
title Audit Me If You Can: Query-Efficient Active Fairness Auditing of Black-Box LLMs
topic Machine Learning
Computation and Language
Computers and Society
url https://arxiv.org/abs/2601.03087