Is Your Hiring Assessment Defensible? A 4-Question Test
An assessment is defensible when you can show four things: that it measures something job-relevant, that every candidate is treated the same way, that it does not produce unfair outcomes across groups, and that a human remains accountable for the decision. If you cannot demonstrate all four, the assessment is a legal and reputational risk, not just a quality problem. Courts and regulators now treat hiring tools as accountable, which means defensibility has moved from good practice to necessity.
Why assessment defensibility matters now
For years, the risk of a weak assessment was a bad hire. Today the risk is a lawsuit. Employers and the vendors behind their tools are being asked to prove their hiring methods are fair and job-related.
The clearest signal is Mobley v. Workday. A federal court allowed the theory that an AI hiring vendor can be treated as an agent of the employer, which means the tool provider can share liability for discrimination. In 2025 the court granted preliminary collective certification on the age-discrimination claim. The case has since expanded rather than narrowed.
The implications go well beyond the companies named in the case. If your screening method produces unfair outcomes, "the software did it" is not a defense. You need to show the method is sound, and you need to show it before anyone asks.
What makes a hiring assessment defensible?
Defensibility rests on principles that I-O psychologists have used for decades. They are the same standards reflected in professional guidelines such as the SIOP Principles and the EEOC's Uniform Guidelines.
Most of it rests on four points. The assessment must be job-relevant, so it measures things that actually matter for the role. It must be consistent, so every candidate has the same experience. It must be fair, so it does not disadvantage protected groups. And it must keep a human accountable, so no decision is fully outsourced to a machine.
None of these is exotic. The only question that matters is whether your assessment has been checked against all four.
The 4-question test
Ask these four questions about any assessment you use. If you cannot answer them with evidence, that is where your risk is.
1. Can you show it predicts performance?
A defensible assessment measures something linked to success in the role. This is validity and it should be supported by evidence. If a vendor cannot show you validity data, treat that as a warning.
2. Does every candidate get the same experience?
Consistency is what separates a measure from an impression. The same questions, the same scoring, and the same conditions make results comparable across people. An unstructured interview, where every candidate is asked different questions, cannot offer this.
3. Have you tested it for adverse impact?
Fairness is something you have to measure. Adverse impact analysis checks whether a tool disadvantages candidates by gender, age, ethnicity, or other protected characteristics. If no one has run this analysis, you cannot claim the process is fair.
4. Can a person explain and stand behind the decision?
Accountability means a human can describe why a decision was made and defend it. A method no one can explain is a method no one can defend. Meaningful human oversight is now both an ethical expectation and, increasingly, a legal one.
Common misconceptions about defensibility
A structured appearance is not the same as validity. A grid, a score, or a polished report can still rest on subjective judgment. What matters is the evidence behind the numbers.
Automation does not create fairness on its own. An AI tool trained on past hiring can inherit the bias in those decisions, which is the heart of the concern in cases like Mobley. Fairness comes from testing.
Defensibility is not only for large enterprises. Any organization that screens candidates can face a claim, and smaller teams often have the least documentation to fall back on.
FAQ
What makes a hiring assessment legally defensible?
It must be job-relevant, applied consistently to every candidate, tested for adverse impact, and subject to meaningful human oversight. Evidence for each is what makes it defensible.
Can an AI hiring tool be defensible?
Yes, but only if it is validated, tested for bias, and kept under human oversight. Automation alone does not make a tool fair or defensible.
Do I need to validate an off-the-shelf assessment?
You need evidence that it is valid for how you use it. A reputable vendor should provide validity and adverse impact data you can review.
What is adverse impact?
Adverse impact occurs when a selection method disadvantages a protected group at a substantially higher rate. It is measured, not assumed, and it is central to defensibility.
Last reviewed September 2026.
