# Do Prevalent Bias Metrics Capture Allocational Harms from LLMs?

Source: [Do Prevalent Bias Metrics Capture Allocational Harms from LLMs?](<https://www.sota2.com/research/paper/68fa74641e209131cea616a7>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-03-10T05:00:05.424Z

Authors: Hannah Cyberey, Yangfeng Ji, David Evans

Published: 2024-08-02T14:13:06.000Z

## Abstract

Allocational harms occur when resources or opportunities are unfairly withheld from specific groups. Many proposed bias measures ignore the discrepancy between predictions, which are what the proposed methods consider, and decisions that are made as a result of those predictions. Our work examines the reliability of current bias metrics in assessing allocational harms arising from predictions of large language models (LLMs). We evaluate their predictive validity and utility for model selection across ten LLMs and two allocation tasks. Our results reveal that commonly-used bias metrics based on average performance gap and distribution distance fail to reliably capture group disparities in allocation outcomes. Our work highlights the need to account for how model predictions are used in decisions, in particular in contexts where they are influenced by how limited resources are allocated.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Essay Grading on ICNALE Ishikawa 2013 (test)](<https://www.sota2.com/research/sota/essay-grading-on-icnale-ishikawa-2013-test.md>): Essay Grading; ICNALE Ishikawa 2013 (test); Pearson Correlation (ΔDP): 0.94.

- [Resume Screening on GPT-3.5 Generated Resumes (test)](<https://www.sota2.com/research/sota/resume-screening-on-gpt-3-5-generated-resumes-test.md>): Resume Screening; Resume Screening real job listings GPT-3.5 generated resumes (test); Pearson Correlation (ΔDP): 0.86.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2408.01285>)

[PDF](<https://arxiv.org/pdf/2408.01285.pdf>)
