The Hidden Cost of AI in Hiring

Automated hiring tools built on internet-scale training data inherit the internet's biases, and those biases fall especially hard on candidates with disabilities. A personal account of becoming paralyzed and later working in HR technology grounds a wider look at how screening algorithms, video interview platforms, and chatbots quietly exclude disabled applicants, often without anyone intending it.

Specific failures are laid out in detail:

• An AI-driven interview scheduler that could not process a request to switch from a phone interview to a video interview for captioning needs, and defaulted to quoting the ADA instead of offering a solution.
• Video interview platforms where recruiters watch only a couple of minutes of a fourteen-to-sixteen-minute recording, meaning first impressions dominate and any AI trained on that reviewing behavior learns to do the same.
• Cognitive and neurological conditions, such as Parkinson's or autism, that produce longer response times, reduced eye contact, or atypical body language, and get misread by both human reviewers and automated scoring as disengagement or lack of qualification.
• Resume gaps caused by medical recovery or disability, which get flagged as red flags rather than probed with a simple follow-up question.
• Training data sets that are rarely, if ever, tested for disability bias, even though a simple benchmark exists: checking whether disabled candidates are hired at meaningfully lower rates than everyone else.

The historical Amazon hiring-algorithm case is used as a reference point for how skewed training data produces skewed outcomes, and the same dynamic is shown to apply to disability status. Practical fixes are offered alongside the diagnosis: opening interviews by asking candidates what would help them have their best experience, building in flexibility for captioning, note-taking, or breaks, and treating unusual interview behavior as a data point rather than grounds for dismissal. Numbers on corporate DEI spending are cited to show how little of that investment currently goes toward disability inclusion compared with other diversity efforts, and why that gap represents both a risk and an opportunity for talent acquisition teams willing to act on it.

An AI has no values and no limitations on how far discrimination can go unless someone actively builds those limits in.