A tool has ranked your applicants, and the top three look odd to you. Or it's offered to send the rejections for you. That's the moment to stop and take the decision back. Here's why.
The law looks at you, not the tool
Hiring laws apply to the employer. The US Equal Employment Opportunity Commission launched an initiative in 2021 on AI and hiring tools.1 Its chair said these tools "may mask and perpetuate bias".1
She also said the technology "may be evolving", but "anti-discrimination laws still apply".1
In Australia, an employer "can't take adverse action" against a prospective employee "for discriminatory reasons".2 The Fair Work Ombudsman's list of adverse action includes "not hiring someone".2 If a tool turns a person away for a reason the law doesn't allow, the business is the one that acted.
That's the plain reason a person must decide. You can't hand the responsibility to software.
What can go wrong inside a tool
Screening tools, in the UK government's words, "are often trained using historical data".3 Old hiring habits can carry over.3 It also says "many of these tests lack scientific validity whether AI-driven or not", speaking of AI-powered psychometric tests and games.3
The UK data regulator audited recruitment tools. It found some "allowing recruiters to filter out candidates with certain protected characteristics".4 A business that used such a filter could end up excluding people for reasons the law protects.
Most of these problems can't be seen from outside. You'd get a ranked list with no sign of how it was made. That's another reason to read the originals yourself.
What "a person decides" means in practice
It doesn't mean a person clicks "approve" on a list the tool made. It means:
- You set the criteria from the job, before any tool is involved.
- You read the applications, not only the summaries.
- You can explain, in a sentence, why each person went forward or didn't.
- You can change your mind if the tool and your own reading disagree.
- Nobody is rejected only because a tool said so.
The UK guide lists questions for organisations to ask. One is how the organisation will "maintain effective human oversight of the system and its outputs".3 Another asks whether human review is appropriate if the AI's role is questioned.3 A small employer can answer both by keeping the decision with one named person.
If a tool's suggestion looks wrong
Check it against the job. If you can't see a reason that fits the job, the suggestion has no weight. Where a tool leaves out a person you'd have interviewed, put them back in.
The small-employer version
Big companies buy screening systems and ask suppliers hard questions. A small employer usually doesn't need that. Use AI to help write and to find your way around, and keep every choice about a person in human hands.
Two other pages cover the steps around this one: reading applications with summaries and the official hiring rules. If you're unsure a step is lawful, ask the official body or a qualified adviser.
Reference list
- EEOC Launches Initiative on Artificial Intelligence and Algorithmic Fairness, U.S. Equal Employment Opportunity Commission
- Protection from discrimination at work, Fair Work Ombudsman (Australia)
- Responsible AI in recruitment, Department for Science, Innovation and Technology (UK)
- ICO intervention into AI recruitment tools leads to better data protection for job seekers, Information Commissioner's Office (UK)