ABLEIST: Intersectional Disability Bias in LLM-Generated Hiring Scenarios

Abstract

Large language models (LLMs) are increasingly under scrutiny for perpetuating identity-based discrimination in high-stakes domains such as hiring. This risk is particularly severe for people with disabilities (PwD), who already face disproportionately high unemployment and systemic discrimination. Existing research on disability bias in LLM-based hiring remains largely Western-centric, overlooking how disability in Global South regions, like India, intersects with marginalized gender and caste identities to shape AI-mediated hiring for PwD. We address this gap by conducting a comprehensive audit of six LLMs across 2,820 hiring scenarios spanning diverse disability, gender, nationality, and caste profiles. To capture subtle intersectional harms and biases, we introduce ABLEist (Ableism, Inspiration, Superhumanization, and Tokenism), a set of five ableism-specific and three intersectional harm metrics grounded in disability studies literature. Our results reveal significant increases in ABLEist harms towards disabled candidates--harms that many state-of-the-art tools fail to detect. These harms were further amplified by sharp increases in intersectional harms (e.g., Tokenism) for gender and caste-marginalized disabled candidates, highlighting critical limitations in current safety tools and the need for intersectional safety evaluations of frontier models in high-stakes domains like hiring.