The Promise
AI hiring tools promise to remove human bias from recruitment. Screen thousands of resumes fairly. Evaluate candidates on merit alone.
The Reality
Multiple studies have shown these tools can discriminate based on:
- Gender — Amazon's recruiting tool penalized resumes containing "women's" (as in "women's chess club")
- Race — Name-based screening correlates with racial demographics
- Socioeconomic status — Education and address data encode class signals
- Disability — Automated video interviews penalize atypical speech patterns and facial expressions
Why This Happens
AI learns from historical data. If your past hiring was biased (and most was), the AI learns those biases.
It's not malicious. It's math doing exactly what it was trained to do — replicate the past.
What Can Be Done
- Audit regularly — Test tools against protected characteristics
- Require transparency — Vendors should explain what signals their models use
- Keep humans in the loop — AI should assist, not replace, human judgment in hiring
- Advocate for regulation — Several jurisdictions are now requiring bias audits for AI hiring tools
The Takeaway
AI can help with hiring. But "AI-powered" doesn't mean "fair." Scrutinize the tools you use.