The Impact of Disability Disclosure on Fairness and Bias in LLM-Driven Candidate Selection

TL;DR

This paper investigates how large language models (LLMs) exhibit bias in candidate selection based on disability disclosure, favoring nondisabled candidates even when other qualifications are identical.

Contribution

It reveals that LLMs tend to prefer nondisabled candidates and are biased against disability disclosures, highlighting a new dimension of fairness concerns in AI-driven hiring.

Findings

LLMs favor nondisabled candidates with identical qualifications.

Candidates not disclosing disabilities are less likely to be selected.

Bias persists even when candidates explicitly state no disability.

Abstract

As large language models (LLMs) become increasingly integrated into hiring processes, concerns about fairness have gained prominence. When applying for jobs, companies often request/require demographic information, including gender, race, and disability or veteran status. This data is collected to support diversity and inclusion initiatives, but when provided to LLMs, especially disability-related information, it raises concerns about potential biases in candidate selection outcomes. Many studies have highlighted how disability can impact CV screening, yet little research has explored the specific effect of voluntarily disclosed information on LLM-driven candidate selection. This study seeks to bridge that gap. When candidates shared identical gender, race, qualifications, experience, and backgrounds, and sought jobs with minimal employment rate gaps between individuals with and without disabilities (e.g., Cashier, Software Developer), LLMs consistently favored candidates who disclosed that they had no disability. Even in cases where candidates chose not to disclose their disability status, the LLMs were less likely to select them compared to those who explicitly stated they did not have a disability.