Haliscope vs. the Myers-Briggs Type Indicator

Where the MBTI came from

The Myers-Briggs Type Indicator was developed in the 1940s by Katharine Cook Briggs and her daughter Isabel Briggs Myers, who adapted Carl Jung’s theory of psychological types into a practical questionnaire. Neither had formal training in psychometrics — the field barely existed in its modern form at the time — and they built the instrument largely outside academic psychology.

It worked, in the sense that mattered to them: the MBTI gave millions of people a vocabulary for describing themselves, and it remains one of the most widely administered assessments in the world.

That vocabulary is why it outlived most of its contemporaries — and why the criticism below is worth making carefully. Something that shapes how millions of people describe themselves should be held to the standard its influence implies.

What the evidence says

The properties that make an assessment useful and the properties that make it accurate are different things, and the MBTI has struggled with the second.

Its test-retest reliability is contested. The publisher reports strong figures — test-retest coefficients between 0.81 and 0.86 on a sample of 1,721 people — while independent reviews have found that a large share of people retaking the instrument within weeks receive a different four-letter type. Pittenger’s 1993 review is the source of the widely cited figure that roughly half of respondents change type within four to six weeks; independent estimates across studies range from about 39% to 76%. A 2002 meta-analysis by Capraro and Capraro reports stronger reliability than the older critiques suggest.

Sources: Pittenger, D. J. (1993), “The Utility of the Myers-Briggs Type Indicator,” Review of Educational Research 63(4); Capraro, R. M. & Capraro, M. M. (2002), “Myers-Briggs Type Indicator Score Reliability Across Studies,” Educational and Psychological Measurement; The Myers-Briggs Company, reliability and validity documentation.

The spectrum and the switch

The MBTI’s central premise is that traits come in either/or pairs. That premise is not well supported. Measured across a population, traits like extraversion are continuously distributed, with most people clustered in the middle rather than at either pole.

That is also the likeliest explanation for the retest problem. When your score sits near the midpoint, a small shift in a few answers flips the letter. The four-letter code turns a spectrum into a switch.

None of this means the MBTI measures nothing. Its four dimensions correlate respectably with four of the Big Five factors, so there is signal underneath. The loss happens at the last step, when a continuous score is reported as a binary letter.

Why the MBTI can’t do career matching

The MBTI was never designed as a vocational instrument. Career lists attached to types are editorial additions, not derived results — somebody decided ENFPs make good journalists and the list propagated. The Myers-Briggs Company’s own guidance cautions against using type to screen candidates.

The vocational research tradition is separate. It runs from the Strong Vocational Interest Blank in 1927 through Holland’s RIASEC model in 1959 to O*NET, the occupational database maintained by the U.S. Department of Labor, which describes tasks, work styles, work contexts and interest profiles for over 900 occupations.

Career matching requires both threads. You need a description of the person and a description of the work, in commensurable terms. A four-letter type gives you neither at sufficient resolution. The full history of both threads is here.

What Haliscope does instead

Four differences, in the order they matter.

• Scores, not letters

  Eleven traits, each a position on a scale. If you sit near the middle of Influencing, the report says so — more useful than being told you are an extravert, and considerably more stable on a retake.

• It watches what you do, not only what you claim

  Forced-choice items ask you to predict your own behaviour. Haliscope adds open text, rankings and timed exercises. Asking someone for as many uses for a paperclip as they can manage in sixty seconds measures divergent thinking; asking whether they consider themselves creative measures self-image.

• It reads how you answer

  Each of the eleven traits carries a written definition, fixed in advance and the same for everyone. A language model reads your whole transcript against those definitions and returns a position on each: it is the scorer, held to definitions you can read, rather than a layer of decoration over one. What follows is not — every occupation’s profile is derived arithmetically from O*NET’s published data, and the ranking uses the same fixed weights and tolerances for everybody. The methodology page sets out both halves, limitations included.

• The output is a roadmap

  Career pathways at different levels of risk, fit scores for occupations you are already curious about, and the steps a move would take. Matched against O*NET occupational data rather than a list attached to a type.

Questions