AI and Disability Employment: Risks, Opportunities, and the Fight for Algorithmic Fairness

The Double Edge of AI in Employment

Artificial intelligence is transforming every stage of employment — sourcing, screening, interviewing, performance management, career development. For disabled people, this transformation cuts both ways: AI can be a powerful accessibility tool, or an efficient mechanism for systematic exclusion.

AI-Powered Hiring: The Discrimination Risk

Automated Resume Screening

Many employers use automated tools to filter applications. The risks for disabled candidates include:

• Gap penalties: Algorithms may penalise career gaps — disproportionately affecting people with disabilities who may have periods of health-related absence or delayed career entry
• Credential bias: Overweighting degrees and prestigious institutions disadvantages people whose education was disrupted by disability
• Keyword optimisation: Candidates need to "game" keyword matching, disadvantaging those with cognitive or learning disabilities
• Proxy discrimination: Even without explicit disability data, AI can infer disability from patterns (gap years, rehabilitation centre addresses, disability-related volunteering)

Video Interview Analysis

Some AI video interview tools analyse features such as tone, word choice, facial movement or body language. Where they do, the risks include:

• Facial expression analysis can fail for people with facial paralysis, cerebral palsy, or conditions affecting facial muscle control
• Eye contact scoring can penalise blind and visually impaired candidates, autistic candidates, and those with social anxiety
• Speech analysis can disadvantage people who stutter, use AAC devices, or have speech impairments
• "Enthusiasm" metrics can disadvantage people with flat affect (associated with some autistic people, depression, or certain medications)

Legal status: In 2023, the US EEOC settled its first lawsuit over discriminatory AI hiring software — a case about age discrimination, where software automatically rejected older applicants. New York City requires bias audits of automated employment decision tools, and the EU AI Act classifies AI used in recruitment and employment decisions as "high-risk".

Gamified Assessments

Some platforms use cognitive games to assess candidates:

• Reaction time tests disadvantage people with motor impairments
• Memory games disadvantage people with cognitive disabilities, brain injuries, ADHD
• Untimed does not mean accessible: Many "untimed" assessments still measure speed implicitly

AI as an Accessibility Enabler

Assistive AI Technologies

• Real-time captioning: AI-powered live captions in phones and meeting platforms improve meeting accessibility for deaf and hard-of-hearing workers
• Text-to-speech and speech-to-text: Dramatically improved by neural networks, both in dedicated software and in built-in operating system features
• Screen reader intelligence: AI helps screen readers navigate complex web applications, interpret images (alt-text generation), and summarise page structure
• Cognitive assistance: AI task managers, reminder systems, and text simplification tools support people with learning disabilities and brain injuries
• Communication aids: AI-powered AAC (Augmentative and Alternative Communication) predicts phrases, learns personal vocabulary, and generates natural speech
• Visual description: AI describes images, scenes, and documents for blind and visually impaired users

Workplace Accommodation AI

• Smart scheduling: AI systems that optimise schedules around energy patterns, medical appointments, and sensory needs
• Environmental controls: AI-powered lighting, temperature, and noise management responding to individual needs
• Meeting accessibility: Auto-captioning, AI meeting summaries, action item extraction — reducing cognitive load
• Document accessibility: AI-powered PDF remediation, automatic alt-text, content simplification

Algorithmic Fairness and Disability

The Technical Challenge

Much algorithmic fairness work focuses on characteristics like race and gender. Disability presents particular challenges:

• Disability is heterogeneous: "Disabled people" is not one group — accommodations for a blind person may be irrelevant for a wheelchair user
• Disability is non-binary: Many conditions are episodic, fluctuating, or invisible
• Disability data is scarce: Many people do not disclose disability, so training data lacks labels
• Fairness metrics conflict: Different statistical definitions of fairness can pull in different directions

Regulatory Frameworks

• EU AI Act (Regulation (EU) 2024/1689): Lists AI systems used to recruit or select people, filter applications, evaluate candidates, or make decisions on promotion, termination and performance monitoring as high-risk. Providers must examine training data for possible biases and take measures to detect, prevent and mitigate them; systems must be transparent enough for deployers to interpret their output, and designed so they can be effectively overseen by humans.
• NYC Local Law 144: Since enforcement began on July 5, 2023, employers and employment agencies may not use an automated employment decision tool unless it has had a bias audit within one year of use; audit information must be made public and candidates must be notified.
• US federal guidance: In January 2025 the EEOC removed its AI technical-assistance documents, including its May 2022 document on AI and the Americans with Disabilities Act. The underlying laws — including the ADA — still apply to employers' use of AI tools.

Best Practices for Employers

1. Audit your AI tools: Require vendors to demonstrate disability-inclusive testing
2. Offer alternatives: Always provide a non-AI pathway through the hiring process
3. Proactive accommodations: Ask about accommodation needs before, not after, automated assessments
4. Human review: Never let AI make final hiring decisions without human oversight
5. Monitor outcomes: Track application-to-hire ratios by disability status (where disclosed) to detect disparate impact
6. Involve disabled people: Include disabled employees and disability organisations in AI procurement decisions

Resources

• EU AI Act, Annex III (high-risk AI systems): https://artificialintelligenceact.eu/annex/3/
• NYC Department of Consumer and Worker Protection — Automated Employment Decision Tools: https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page
• EEOC: iTutorGroup settlement press release: https://www.eeoc.gov/newsroom/itutorgroup-pay-365000-settle-eeoc-discriminatory-hiring-suit
• Cooley: "Gone but Not Forgotten: Federal Laws Still Apply Despite AI Guidance Disappearance Act" (2025): https://www.cooley.com/news/insight/2025/2025-02-21-gone-but-not-forgotten-federal-laws-still-apply-despite-guidance-disappearance-act