There is a line in The Myers-Briggs Company's own guidance for employers that should end most conversations about MBTI in hiring before they start: the instrument "shouldn't be used in hiring," and doing it anyway is, in the company's own word, "unethical." That is not a critic talking. That is the organization whose revenue depends on the assessment saying its product does not do the thing your interviewer just asked it to do. The Myers & Briggs Foundation says the same in flatter language — it is not designed to be used for hiring purposes. JobCannon's State of Personality Testing 2027 report quotes the publisher more bluntly still: it "should never be used in recruitment or selection because it does not measure a person's skills or abilities." Never.
The Publisher Is the Prosecution's Best Witness
Validated — For Something Else
The argument in that guidance article is more precise than a flat "don't," and the precision is the useful part. The company's claim is that "an assessment can be psychometrically validated for one use case but be highly inappropriate for another use case." The MBTI instrument has been psychometrically validated for its intended use, it says, "but not for hiring and selection... it wasn't designed to predict job performance." The sanctioned uses are listed plainly: team building, conflict management, leadership development, and other non-selective purposes.
So the defense a hiring manager reaches for — "it's a scientifically validated instrument" — is true and irrelevant simultaneously. A blood pressure cuff is validated. It is not validated for picking who runs the sales team.
The Same Company Does Sell a Selection Tool
This is the detail that makes the position credible rather than performative. In the same article, The Myers-Briggs Company points to the CPI (California Psychological Inventory) as appropriate for selection when properly validated, citing its flexibility scales as an example of something that predicts job performance. It also hands employers a hard floor for any vendor: the general standard is internal consistency reliability of .70 or above, and "if you can't get your hands on documentation showing that the assessment has acceptable reliability and validity data, don't use it." A company protecting MBTI revenue in the hiring channel would not disqualify MBTI from the hiring channel and then point to a different product in its own catalogue.
The Evidence Gap Is Not "Weak" — It Is Empty
The Number That Doesn't Exist
JobCannon's report went looking for the figure everyone assumes must exist somewhere and came back without it: no peer-reviewed meta-analysis estimating an MBTI-to-job-performance correlation. Not a small one, not a contested one. None. DISC got the same result — no peer-reviewed predictive-validity coefficient located — and so did career-matching algorithms. That is a worse situation than "the correlation is low." A low number is evidence you can weigh and discount. An absent number means nobody has run the study that would let a hiring team defend the decision afterward.
What the Numbers Look Like Where They Exist
Poll-Maker's 2026 evidence review centers on the Sackett et al. 2022 reanalysis of selection-method validity. Read the bottom rows first.
| Selection method | Operational validity | Evidence note | Defensible as a screen? |
|---|---|---|---|
| Structured interview | .42 | Highest in the Sackett et al. 2022 set | Yes — recommended primary |
| Work sample | .33 | Black/White Cohen's d of .67 | Yes, with monitoring |
| Cognitive ability | .31 | Largest subgroup gap: d of .79 | Yes, biggest legal exposure |
| Integrity test | .31 | Established literature | Yes |
| Conscientiousness, work-framed | .25 | Work-framing roughly doubles generic (.24 vs .11, Shaffer & Postlethwaite 2012) | Tie-breaker, not filter |
| Emotional stability, work-framed | .23 | Same framing advantage | Tie-breaker only |
| Conscientiousness, generic | .21 | Big Five corrected ~.20 (Hurtz & Donovan 2000; Zell & Lesick 2022) | Marginal |
| RIASEC interest scales | .14 single / .23 vocation-specific | Van Iddekinge et al. 2011 | No |
| MBTI | No published figure | JobCannon found no peer-reviewed meta-analysis | No — publisher says never |
| DISC | No published figure | No peer-reviewed coefficient found | No |
Two caveats about that column of decimals. These are corrected operational validities; Watrin et al. 2023 put the conservative floor for personality measures at an uncorrected raw correlation of .14 in real applicant samples, a long way from .25. And even the leader explains well under a quarter of the variance in performance. Nothing in hiring works as well as hiring managers talk.
The Property That Makes It Unusable as a Gate
Different Letters, Same Person, Five Weeks Later
The Skeptic's August 2026 piece presses the point that matters most for selection: the type does not hold still. The figures circulating with that article put the share of people who get a completely different four-letter type on a retake roughly five weeks later somewhere between 39% and 76% — a wide band, best treated as a range rather than a point estimate. Poll-Maker states it as roughly 50% receiving a different type on retest. The two do not agree on the number; they agree on the direction, and the direction is fatal for a gate. A measurement that can flip inside a hiring cycle cannot be the thing that decides the hiring cycle.
The structural explanation is about scoring, not questions. Measured across a population, the underlying traits do not cluster into two camps — they form a normal distribution with most people bunched near the middle. Drawing a line through that hump and issuing everyone a letter manufactures a cliff where the data has a slope. The provenance fits: the questionnaire was built in the 1940s by Katharine Cook Briggs and her daughter Isabel Briggs Myers, neither formally trained in psychology, on top of Carl Jung's 1921 theory of psychological types — which Jung never tested empirically and which they did not test either.
And It Is Gameable, Which Candidates Already Know
Faking is measurable. Birkeland et al. 2006 found applicants inflated conscientiousness by about 0.45 standard deviations versus non-applicants, and emotional stability by about 0.44. The mitigation Poll-Maker offers is real but thin: "because nearly all candidates inflate, rank ordering suffers less than absolute scores do." On the long-running Hacker News thread defending MBTI against its harsher critics, one commenter describes the workplace version without any psychometric vocabulary at all — during employment testing, they "chose a personality at the outset and 'achieved' that outcome." The items are transparent enough that anyone who has read a type description once can steer. A screening question about type does not measure your personality. It measures your guess about which personality the company wants.
So Why Is It Back in the Funnel?
The Adoption Numbers Are Smaller Than the Panic
JobCannon cites SHRM 2024 data (n=2,366) showing 54% of US organizations use pre-employment assessments of some kind — but among those employers, only 14% named personality tests specifically. UK CIPD 2026 data (n=791) puts personality or psychological questionnaires at 20% of organizations that recruited in the past year. Employers who skip them gave reasons: doubts the tests measure needed skills (38%), ROI concerns (36%), bias and discrimination worries (36%), thin resources (36%), negative applicant reactions (19%). The volume figures are where it looks enormous, and those are self-reported vendor numbers that do not add up to anything — the Myers & Briggs Foundation claims 2 million-plus MBTI tests a year, 16Personalities 1.5 billion cumulative worldwide. Candidates mostly do not mind: Criteria's 2025 survey (n=2,000+) found 86% of job seekers liked assessments or felt neutral.
The Defense That Deserves a Hearing
The commenter who opened that Hacker News thread makes a point MBTI's fans usually mangle: "It's not super predictive, but not to be honest, nothing else in psychology is either." The evidence offered — a tech company poll returning 70% INTJ and 25% INTP among engineers against roughly 2% INTJ in the general population — is a self-selected poll of self-reported types, so treat it as an anecdote wearing a percentage. The sturdier defense in that thread abandons prediction entirely: MBTI's value is shared vocabulary, a common frame for discussing how colleagues differ. Others there push on permanence, arguing personalities are malleable and the result is a snapshot of current context. Both arguments land where the publisher does. A snapshot and a shared vocabulary are development tools, not gates.
What to Actually Say in the Room
The Script
If type comes up in an interview, you do not need to litigate psychometrics. You need one question and one redirect.
- Ask where it goes: "Happy to talk about working style — is this part of the hiring decision, or onboarding?" The publisher's own line separates those two uses, and the answer tells you which conversation you are in.
- Cite the owner, not the critics: the instrument should never be used in recruitment or selection because it does not measure skills or abilities. Quoting the vendor is harder to argue with than quoting a skeptic.
- Offer the better tool: "I'd rather show you the work," then volunteer a structured behavioral example. At .42, the structured interview beats every personality measure on the board.
- Answer descriptively if you would rather not fight: give the behavior, skip the letters. "I do my best thinking before the meeting and my best arguing in it" is more informative than INTJ.
- Know the legal shape: Karraker v. Rent-A-Center (7th Cir. 2005) held MMPI-based testing to be a medical examination under the ADA. Newer rules bite on automated screening — NYC Local Law 144 (enforced July 5, 2023) requires annual independent bias audits and public disclosure, California regulations effective October 1, 2025 add anti-bias testing expectations and four-year data retention, and Illinois HB 3773 took effect January 1, 2026.
What Is Still Unknown — Including One Fact That Cuts the Other Way
JobCannon explicitly declines to provide market share by framework or retest reliability figures for any framework, so nobody can tell you how many companies still screen on type. The regulatory wind is not uniformly behind candidates either: the EEOC's National Enforcement Plan of June 2026 deprioritized disparate-impact theories. And the awkward finding critics rarely mention — on subgroup differences, personality testing is the cleanest predictor available. Conscientiousness shows a Black/White Cohen's d of −.07, effectively no gap, against .79 for cognitive ability and .67 for work samples. That is a genuine argument for personality measures in hiring. It is not an argument for MBTI, which has no validity coefficient to set beside that clean profile. Poll-Maker's rule is the one to keep: personality is a tie-breaker, not a filter, and MBTI, DISC and the Enneagram are to be avoided for selection outright.
The useful thing about this controversy is that it does not require picking a side in the decades-old fight over whether MBTI means anything at all. The publisher, the skeptics, the 2026 evidence reviews and the people defending the test on Hacker News converge on the same boundary: development yes, selection no. An employer who crosses it is not disagreeing with critics — they are disagreeing with the company that sold them the test. Expect that gap to get expensive as bias-audit rules mature, and in the meantime, when you decline to be screened on four letters, know you are standing exactly where the trademark holder stands.
-EditorZ
Photo by Vitaly Gariev on Unsplash
Post a Comment