The tool may cause unlawful disparate impact: Under nondiscrimination law, what is the primary legal risk even
A neutral tool that disadvantages a protected class can create unlawful disparate impact even when the protected attribute is not an input.
The question
An employer uses an AI resume-screening tool that, in testing, advances men at a noticeably higher rate than equally qualified women. Under nondiscrimination law, what is the primary legal risk even though gender is not an input feature?
Preparing for AIGP? Take the free 5-min readiness quiz →
- There is no legal risk at all, as long as the protected attribute itself was excluded from the features the model was trained on.Plausible because removing the attribute feels protective, but proxies can reproduce the disparity and impact still counts.
- The tool may cause unlawful disparate impact, because a facially neutral process can still produce discriminatory outcomes on a protected class. ✓Correct: disparate impact liability can arise from neutral processes that disproportionately disadvantage a protected group, regardless of intent.
- Liability can arise only where a human recruiter intentionally set out to disadvantage one particular group during the hiring decision-making process.Plausible since intent drives disparate treatment, but disparate impact does not require proof of intent.
- Nondiscrimination duties apply only to lending and housing decisions, but not to employment screening carried out by software.Plausible as those sectors are named, but employment is squarely covered and automation does not exempt it.
The trap
Believing that excluding the protected attribute from features, or lacking discriminatory intent, eliminates nondiscrimination liability. How to remember it
A neutral tool that disadvantages a protected class can create unlawful disparate impact even when the protected attribute is not an input.
How many of these would you get right?
One of 1581 AIGP questions on Certsqill. Take a free five-minute check and see your score per domain — not one number, but which section to open tonight.
Test your AIGP readiness — freeMore Understanding How Laws, Standards and Frameworks Apply to AI questions
- Copyright may prohibit or limit using the material: From an intellectual-property standpoint, which →
- Whether the team can establish lawful basis and purpose: What is the decisive issue before reuse? →
- The new counseling prediction purpose must be assessed: Which distinction controls the reuse analysis? →
- All 394 Understanding How Laws, Standards and Frameworks Apply to AI questions →
Part of the Certsqill AIGP question bank · Understanding How Laws, Standards and Frameworks Apply to AI ·
Every answer, right and wrong, comes with its own explanation.
