Miraai Personality

What a candidate is like at work, in ten minutes.

Candidates see pairs of pictured statements and choose which one is more like them, and by how much. Both statements in a pair are reasonable, so there is no flattering answer to reach for, and what comes back is closer to how the person actually works.

Try a pair Book a demo

The personality suite

One model, four ways to use it.

Every form reads the same six factors, so a short screen, a full profile, Edge and Needs all sit together in one report and in one sitting when you want them to.

Miraai Traits

The full profile

Reported at factor and trait level, with twenty-four facets behind them that give a report its detail. It takes ten minutes, in pictures or in text.

Ten minutesFactors and traits

Miraai Factors

The short form

A shorter form scored by factor and trait, for early screening when the time a candidate spends matters most.

Half the lengthEarly screening

Miraai Edge

How traits hold up under pressure

Short work situations that show whether a strength stays useful or starts to cost the team, at both ends of each trait.

Work situationsAbout 12 minutes

Miraai Needs

What someone needs from the work

Sixteen work needs, from discovery and challenge to security and balance, so fit is part of the answer rather than a guess.

16 needs

The model

Six factors, twelve traits, twenty-four facets.

The five big personality factors plus honesty, each split into two traits and each trait into two facets. Both ends of every scale are written as a way of working rather than a weakness, because a person low on extraversion is quiet, not deficient, and the report says so.

Openness

GroundedExploring

Intellect

PracticalConceptual

Openness to experience

ConventionalInquisitive

Extraversion

QuietProminent

Enthusiasm

ReservedLively

Assertiveness

UnassumingAssertive

Agreeableness

DirectWarm

Compassion

ImpartialCompassionate

Politeness

DirectPolite

Conscientiousness

EasygoingDriven

Industriousness

RelaxedIndustrious

Orderliness

FlexibleStructured

Emotional Stability

SensitiveSteady

Engagement

CautiousAssured

Regulation

ReactiveComposed

Honesty-Humility

StrategicGenuine

Honesty

DiplomaticCandid

Humility

AmbitiousHumble

Miraai Edge

What a person does with a trait when the work gets hard.

A personality profile tells you how much of a trait someone has, but not what they do with it under pressure. Edge measures the same traits by putting the person in short work situations, and it looks at whether they recognize when a trait is working against them and whether they correct for it without losing what the trait gives them.

None of this asks anyone to be moderate. Effective people can sit at the far end of a trait, and what sets them apart is that they know what it costs and they manage it, so instead of "high on conscientiousness" a report can say "thorough, and knows when to stop."

Every strength has two edges.

Too little of a trait costs performance, and so does too much. Edge measures how a person handles both ends at work.

Managing the edge.

The risk is never a strong trait but its overdone form, and Edge scores whether a person sees the edge coming and keeps the strength working.

Two reasonable answers.

One work situation and two responses that sound equally good, where one is the effective call and the other is the trait overdone.

Measured where it happens.

Personality measured in work situations predicts performance about twice as well as general personality questions (Shaffer and Postlethwaite, 2012), and every Edge item is one.

Miraai Needs

What someone needs from the work to do their best.

Measured with the same pictured pairs and reported beside the personality profile, so a manager sees what the person is like and what the role has to give them in return.

Discoverydrawn to discovery work

Creationdrawn to creative work

Practicaldrawn to hands-on work

Achievementdrawn to goal-driven work

Challengedrawn to high-stakes work

Competitiondrawn to contest-based work

Influencedrawn to directing roles

Recognitiondrawn to visible, esteemed work

Affiliationdrawn to team-based work

Servicedrawn to helping work

Integritydrawn to principled work

Structuredrawn to structured work

Autonomydrawn to self-directed work

Securitydrawn to stable employment

Rewarddrawn to financially rewarding work

Lifestyledrawn to balanced work

The evidence

Two formats, one score, and the study in public.

The personality assessment runs as pictured pairs or as plain text, and the validation study behind it is published in full on the Open Science Framework, with its analysis log. No AI model scores, ranks or rejects a candidate.

The validation study on the Open Science Framework →

r = .84between a person’s pictured and text-only scores, in 152 people who took both

r = .83average two-week stability of the factor scores, in 313 people who came back

No pairbehaves differently between the two formats beyond the negligible boundary

Source: the Miraai Validation Study, 2026. Read it on the Open Science Framework

Reliability

Factor scales α .82 to .93, and .88 to .95 on the Thurstonian model matched to the forced-choice format; trait scales α .74 to .89, mean .82. Internal consistency on 790 completed forms from working adults.

Stability

Two-week test-retest .77 to .88 on the factors and .70 to .87 on the traits, in 313 people who came back, half in each format.

Two formats, one scale

Pictured and text formats are invariant at the configural and metric levels (ΔCFI at most .001 on every factor), their pair parameters correlate .92 for discrimination and .97 for thresholds, and no pair shows differential functioning by format. The negligible boundary is McFadden’s ΔR² .035; the largest format effect is .004.

Fairness

By gender, age band and education, 0 of 96 pairs exceed the negligible DIF boundary (largest ΔR² .019). Mean subgroup differences are small: gender mean |d| .18, largest .41; age mean |d| .09, largest .25. Among 15,069 applicants under real stakes, reported in the validation study’s operational chapter, first- and second-language candidates differed by no more than a tenth of a standard deviation on any factor.

What the scores relate to

Conscientiousness .42 with citizenship toward the organization and −.42 with counterproductive behavior toward it; Agreeableness .48 with citizenship toward individuals; Honesty-Humility .30 with organizational citizenship and −.20 with counterproductive behavior; Emotional Stability −.28 with counterproductive behavior. n ≈ 790.

Norms

Scores are reported as stens against a reference distribution of 790 completed forms from working adults, with a second, operational applicant norm of 15,069 people. The reference statistics, sten conversions and confidence bands are in the validation study, the assessment’s technical manual.

What an AI produces when told to look like the strong candidate

Mean factor score · sten 1 to 10

What an AI produces when told to look like the strong candidate · Mean factor score · sten 1 to 10
	Honest (AI as itself)	‘Ideal candidate’	‘Career coach’	‘Best for a sales-manager role’
Honesty-Humility	7.6	6.8	7.2	3.1
Emotional Stability	7.3	8.3	8.5	8.8
Extraversion	5.5	8.0	6.7	10.0
Agreeableness	7.1	8.4	7.1	7.4
Conscientiousness	6.7	8.2	8.2	8.4
Openness	7.7	7.2	5.8	7.1

We pointed language models at our own test and told them to fake. Asked to be the best candidate for a sales-manager role, they pushed Extraversion to the top of the scale and paid for it in Honesty-Humility, because every pair makes them choose, and that trade is what gives the attempt away. Mean factor score on a scale of 1 to 10 where people average 5.5, the dotted line. Each line pools three language models (Claude Sonnet 4.5, Gemini 2.5 Flash and Qwen2.5-VL-72B) answering all 96 pairs in both formats, twice each: 48 complete forms in all. Instructed faking raised the profile, so faking is reduced here, not removed.

Fair in a second language

More than fifteen thousand applicants to a government scholarship program took it under real stakes, most of them in a second language, and first- and second-language candidates differed by no more than a tenth of a standard deviation on any factor.

Reports a manager can use

Hiring reports come with the interview questions to ask next, development feedback is written for the person rather than the file, and scores map onto your own competency framework so the report reads in your managers’ words.

Next step

Send us one job description.

We'll build the first scene for it and show you, so you can look at the real thing instead of reading about it.

Book a demo Ask about a custom build