The personality test in a job application is not a quiz. In federal employment law it is a selection procedure, which puts it in the same regulatory bucket as a cognitive exam, a strength test, or a credit check — and subject to the same arithmetic. The EEOC's guidance on employment tests, issued December 1, 2007 under control number EEOC-NVTA-2007-2, lists "personality tests and integrity tests" directly alongside physical ability tests and medical inquiries, and defines them as instruments that assess "the degree to which a person has certain traits or dispositions (e.g., dependability, cooperativeness, safety)." Nothing in that framework asks whether the vendor called it a culture-fit survey. An October 2026 breakdown of the legal risks of these screens, published as employers scale assessments back up to cope with AI-written resumes, lays out three distinct ways one becomes unlawful — and they fail independently. A test can clear all the demographic math and still be illegal purely because of when it was administered.
The Law Does Not Care What You Call It
"Selection Procedure" Is the Operative Phrase
The rulebook here is older than most of the products sold under it. The Uniform Guidelines on Employee Selection Procedures were adopted by the EEOC in 1978, sit at 29 CFR Part 1607, and were issued jointly with the Departments of Labor and Justice and the Office of Personnel Management. The Bloomberg Law analysis by Alston & Bird's Anna Saraie and Martha Doty — published February 12, 2024, as AI-enhanced assessments were reviving employer interest — cites 29 C.F.R. 1607.1 as the governing hook. The doctrine underneath is older still: Griggs v. Duke Power Co., 401 U.S. 424 (1971) established that a neutral-looking practice violates Title VII if it disproportionately excludes a protected group and the employer cannot show it is job-related and consistent with business necessity. Title VII applies at 15 or more employees; so does the ADA.
So the legal question is never "is this test scientifically valid in general." It is "does this test, as you used it, for this job, screen people out unevenly — and can you prove it predicts performance?"
The Vendor Is Not Your Shield
This is the part employers get wrong most expensively. The EEOC's position, as Bloomberg Law summarizes it, is that an employer can be liable under federal anti-discrimination law even when an outside vendor built the software. The agency issued guidance on software, algorithms and AI in employment selection on May 18, 2023, and guidance on ADA compliance in algorithmic decision-making on May 12, 2022. The 2007 testing guidance already said the same thing in plainer words: relying on a vendor's assurance is not sufficient validation, and the employer remains responsible for ensuring the test is properly validated for the position and the purpose it is being used for.
Failure Mode One: Adverse Impact and the 80% Rule
The Arithmetic Is Genuinely This Simple
The EEOC's four-fifths rule asks one question: is the selection rate for a protected group less than 80 percent of the rate for the highest-scoring group? If yes, the test draws scrutiny and the burden shifts to the employer to demonstrate job-relatedness.
Run it as plain arithmetic. Say a screen passes 60% of one group and 45% of another. 45 divided by 60 is 0.75 — below the 0.80 threshold, so the employer now owes a validation defense. Nudge the second figure to 50% and the ratio is 0.83, which clears it. The uncomfortable implication for applicants: the rule is a trigger for review, not a verdict. A screen can sit at 0.79 for years and nobody checks, because nobody outside the company can see the denominators.
What It Looked Like When Someone Did Check
The EEOC's own case summaries show the scale. In EEOC v. Dial Corp., a strength test described as considerably more difficult than the actual job dropped female hires from 46% to 15%; the Eighth Circuit upheld a Title VII disparate impact violation. In EEOC v. Ford Motor Co., a cognitive test used in apprenticeship selection had disparate impact on African American applicants while less discriminatory alternatives went unused — Ford paid $8.55 million in monetary relief and had to replace the procedure. Discrimination charges involving employment testing hit 304 in fiscal year 2007, which the agency flagged as a high point, and it held a public meeting on employment testing and screening on May 16, 2007. Also worth noting: Title VII separately forbids adjusting scores, using different cutoffs, or otherwise altering results by protected characteristic — so "we just curve it" is not a fix, it is a second violation.
Failure Mode Two: Timing
When a Personality Test Quietly Becomes a Medical Exam
Here is the distinction that decides it. A personality assessment measures workplace behavior. A medical examination identifies a mental health condition or a disability. The October 2026 analysis puts the line precisely: tests that screen for anxiety, depression, or "abnormal" traits may be classified as medical examinations — and under the ADA, a medical examination cannot be required before a conditional offer of employment. Post-offer, an employer may ask disability-related questions and conduct medical exams, provided it does so for all individuals entering the same job category.
Read that twice if you are an applicant. The illegality does not depend on what the test found out about you, or on whether you were rejected. If the instrument functions as a mental-health screen and it arrived before the offer, the timing alone is the violation.
Target Paid $2.8 Million for Exactly This Combination
The EEOC's action against Target is the clearest worked example available. Four employment assessments were examined. The agency alleged three had disparate impact on hiring by race and gender in violation of Title VII, and that one amounted to an unlawful pre-employment medical examination under the ADA. Thousands of individuals were affected, and the matter resolved for $2.8 million, with Target discontinuing the tests during the investigation, modifying its applicant tracking systems to address potential adverse impact, and hiring an outside consultant to train staff on recordkeeping, ADA medical exam requirements and disparate impact analysis. Note the shape: both failure modes at once, in the same hiring funnel, at a company with a full legal department.
Failure Mode Three: Never Validating the Test Against the Job
Built for Development vs. Built for Selection
The Uniform Guidelines outline three methods for demonstrating job-relatedness, and the business-necessity defense requires evidence that the test predicts performance in the specific role — as the October 2026 piece phrases it, not merely that it "feels relevant." That standard sorts the product market cleanly, and most assessments on the market were never built to meet it.
| Instrument | What it is built to do | Defensible as a hiring screen? | Where it breaks |
|---|---|---|---|
| DiSC | Describe communication preferences | No — approved for team development only | No selection validation; using it to reject people has no business-necessity backing |
| PXT Select | Job-fit selection, with validation research | Yes — designed for hiring decisions | Still needs a role-specific job profile and a four-fifths audit of your own results |
| MBTI / 16-type instruments | Self-understanding, shared vocabulary | No — the publisher's own materials say not to use it for hiring | Commenters on Hacker News cite research that as many as 75% get a different result on retest, and no demonstrated predictive power for job performance |
| Anything screening anxiety, depression, "abnormal" traits | Reads as a mental health screen | Not pre-offer, at any validation level | ADA timing — the violation exists before the scoring question arises |
The Vendor Telling You Not To Do It
The Hacker News thread on why companies are interested in Myers-Briggs types contains the detail that should end the argument: the Myers-Briggs Company's own marketing says not to use the instrument for hiring. Regulars there report companies doing it anyway, sometimes without disclosure, including an account of a sales firm administering combined IQ and personality tests with no certification behind them. The same discussion notes MBTI correlates with Big Five traits in the 0.45–0.72 range while converting continuous traits into binary categories, and cites Adam Grant's 2013 Psychology Today piece as the standard objection. A test the publisher disclaims for selection is, by definition, a test with no selection validation to put in front of the EEOC.
Why Autistic and Depressed Applicants Get Filtered First
The Trait Items Are Sometimes Symptom Items
Combine the three failure modes and the pattern is structural rather than malicious. Items probing sociability, enthusiasm, assertiveness in groups or emotional steadiness sit at the exact point where a workplace-behavior question and a disability question become indistinguishable — which is why the October 2026 analysis treats anxiety and depression screening as the bright line for medical-exam reclassification. An autistic applicant answering honestly about small talk, or someone managing depression answering honestly about energy levels, produces a low score the employer reads as poor fit. And because disability status is rarely disclosed at the application stage, these candidates are the hardest group for a four-fifths audit to see at all.
Accommodation Exists, and Mostly Nobody Offers It
It is not hypothetical. In EEOC v. Daimler Chrysler Corp., a learning-disability accommodation case, the settlement covered 12 identified individuals and required provision of readers as an accommodation on pre-employment tests. The EEOC's best-practice list also warns against tests being administered casually by untrained managers and tells employers to identify equally effective alternatives with less adverse impact. Both of those are, in practice, accommodation rights read from the employer's side.
What an Applicant Can Actually Use
- Ask when, not whether. Best practice per the October 2026 guidance is to administer selection tests after a conditional offer. A mental-health-flavored screen before an offer is the cleanest ADA exposure an employer can generate.
- In New York City, there is a paper trail. Local Law 144, effective January 1, 2023, requires annual bias audits and candidate notice for automated employment decision tools.
- In Illinois, consent is required before AI analyzes a video interview, including notice and explanation. Maryland and Colorado have enacted similar AI hiring provisions.
- In California, the results are your data. Under the CPRA, personality test results qualify as personal information, carrying rights to know what is collected, to delete it, and to opt out of certain uses.
- Vendor branding is not a defense — for them. If the screen was built by a third party, the employer still owns the liability.
The near-term direction is set by volume: as AI-written applications push employers toward more automated filtering, the exposure concentrates in screens nobody validated and nobody timed correctly. The compliance checklist is unglamorous — job profile before posting, test after the conditional offer, identical procedures for every candidate in the same role, outcomes audited by demographic group against the four-fifths ratio, vendor validation research verified, assessment data kept separate from interview notes. What stays genuinely unknown is how often any of it gets audited absent an EEOC charge, because the selection rates that decide the 80% question live entirely inside the employer's own systems. If you were screened out by four letters, you will probably never see the number that did it.
-EditorZ
Photo by Zulfugar Karimov on Unsplash
Post a Comment