How the Index is made, and what it is not

The instrument

Twenty scenarios, each with six answers, each answer attributed to one of six axes: psychopathic, sociopathic, narcissistic, borderline, histrionic and neurotypical. An axis score is the share of the twenty answers that went to that axis, so scores move in steps of five and the six scores always sum to 100. The quiz is forced-choice: an answer that goes to one axis is an answer that does not go to the other five, so a high score on one axis mechanically lowers the rest. That matters for the correlations below.

A taker's type is their highest axis. Twenty items score into it: fifteen scenarios and five functioning items. When two axes tie at the top the tie is broken on the fifteen scenario items alone, then, if those tie too, away from the psychopathic axis; a tie involving neurotypical always goes to neurotypical. Ties used to fall to the psychopathic axis by the order the axes happened to be written in. That was a bias, it was fixed with instrument v2 on 27 August 2026, and it is stated here rather than hidden.

The items have not changed since the quiz opened on 4 April 2026 (instrument v1). If they ever change, the time series breaks; any revision will be versioned and every page will say which version its numbers came from.

What the six axes map onto

The six axes are trait profiles, not disorders. ICD-11 has no histrionic, narcissistic or antisocial category, and the DSM-5 alternative model describes personality as severity plus trait domains rather than as a list of types, so a high axis here describes a direction in somebody's answers and never a condition they have.

That is not a hedge, it is where the field went. Each axis below is written beside the trait language the two current manuals use, so a reader who knows one vocabulary can read the other. Psychopathic reads as the callous end of dissociality with the boldness that low negative affectivity produces; sociopathic reads as disinhibition; narcissistic as antagonism, grandiosity and attention seeking; borderline as negative affectivity with disinhibition, and it is the one pattern ICD-11 kept as a qualifier; histrionic as attention seeking under antagonism, noting that ICD-11 has no histrionic category at all and that it is the weakest construct in the DSM's own list; and neurotypical as the absence of an elevated domain rather than a domain of its own.

Who is in the sample

Anyone who finished the quiz on kanikarose.com or in the app since 4 April 2026, worldwide, in English. Most arrive from Kanika Batra's own channels, some from search, some from a friend. That makes it a self-selected sample of people interested in dark psychology, drawn toward a creator who is diagnosed with ASPD and clinically assessed as Factor 1 psychopathy. Every share on the Index is therefore expected to run above a community estimate, and the section on the literature below says by how much and why.

Region is inferred from the browser's timezone, so "New York" means Eastern-time takers rather than the five boroughs; regions with fewer than 30 takes are held back, and no region is ranked against another below 100. Optional one-tap fields (age bracket, gender, work area, how the taker found the quiz, and the type they expected before the reveal) are collected after the result, never required, and reported only in aggregate once each group clears its floor.

What is excluded

The live counter on the hub counts takes. The sample behind every chart counts takers, after these rules, applied on every edition:

• Rows without scores or a recognised type are dropped.
• Second and later takes by the same account are dropped from the charts (kept only for the test-retest figure below).
• Takes finished in under sixty seconds are dropped. Completion time is recorded from 18 August 2026; earlier takes have no time and are kept.
• Takes with eighteen or more of twenty answers on one axis are counted and flagged here, not excluded: on a forced-choice scenario quiz an extreme profile can be a real one.

How the numbers are computed

Percentiles are mid-rank against everyone in the sample: "higher than 91%" means 91% scored strictly lower plus half of those who tied, which is the standard definition and the only one that behaves when scores can only take twenty-one values. Percentiles are clamped to 0 to 99, so nobody is told they beat everyone.

Bands are fixed cut points on the 0 to 100 axis score: baseline 0 to 15, traits 20 to 35, pronounced 40 to 55, high 60 to 100. "High" starts at 60 so that it means at least twelve of twenty answers went to one axis. Bands are the Index's reading aid. They are not clinical thresholds and were not derived from any clinical instrument.

Editions. The histograms behind every chart are recomputed once a day at 09:41 UTC and stored; each recomputation is a numbered edition (days since 4 April 2026). If a day is missed the next read recomputes. The hero counter is live; the charts are the morning's edition; the dateline says which.

Uncertainty. Every share the Index prints carries a 95% Wilson score interval, and one set of floors decides what is shown at all: no share below 30 takes, no ranking below 100, no "top 1%" threshold below 1,000 on an axis, no reliability figure below 1,000 complete answer sets, no retest figure below 50 pairs. Where two groups' intervals overlap, the pages show them tied rather than ordered.

Reliability and structure

Internal consistency per axis is KR-20 (Cronbach's alpha for binary items) over twenty indicators of "this item went to this axis", from 7,722 complete answer sets. An axis is only as consistent as it is used: an axis that most takers put one or two answers on (histrionic averages under one) cannot read high on this statistic whatever the items are like, so read the figures beside the mean. They are published as they come out.

Correlations between axis scores (Pearson, n = 7,722). Because the six scores sum to 100, they are negatively correlated by construction; read the matrix for which axes trade off most and least, not as evidence of distinct factors.

Test-retest. Not enough yet. 6 people have taken the quiz twice on the same account; the figure publishes at 50 pairs, and the 90-day second sitting will build it.

Beside the literature

The Index will read higher than every row above, and it should. People who take a dark-psychology quiz on a dark-psychology creator's site are not a community sample; the meta-analysis itself finds self-selected and organizational samples running four to seven times the community rate. Our high band is also a reading aid on a self-report screener, not a PCL-R assessment. Read the Index as an audience portrait beside the literature, never as a rival estimate of prevalence.

The Index's own high-band shares, same edition, with intervals:

• Psychopathic3% (3 to 4%)
• Sociopathic0%
• Narcissistic0%
• Borderline0% (0 to 1%)
• Histrionic0%
• Neurotypical26% (25 to 27%)

"High" is 60 or more on the axis, a reading aid on a self-report screener. Factor 1 and Factor 2 are terms from Hare's PCL-R, not the DSM; the DSM diagnosis is antisocial personality disorder. Nothing in this table validates the quiz against either.

What this is not

• Not a diagnosis. Only a clinician can make one, and only with a proper assessment.
• Not the PCL-R, the LSRP, the SD3 or the TriPM, and not validated against any of them.
• Not a population estimate. Every share is a share of people who took this quiz.
• Not clinical thresholds. The four bands are fixed cut points chosen for reading, stated above.
• Not independently reviewed yet. A named methodological review is being sought and will be published here verbatim, reservations included.

Disclosure

The eight elements a careful journalist or fact-checker asks for, in the shape of the AAPOR Transparency Initiative's disclosure standard, stated here in one place.

• Sponsor and conductor. Designed, run and funded by Kanika Batra / kanikarose.com. No external sponsor.
• Instrument. The Dark Mirror: a 20-item scenario quiz, five options per item, six axes scored 0 to 100 in steps of five. The full item logic is described under “The instrument” above.
• Population under study. Visitors to kanikarose.com who chose to take the quiz. Nothing broader is claimed, anywhere.
• Sample design and recruitment. Self-selected, non-probability. Takers arrive from search, social media and press coverage; there is no sampling frame, no quotas and no weighting.
• Mode. Self-administered online, any device.
• Dates of collection. Continuous since 2026-04-04; each edition states its own compute date and count.
• Sample size. Printed on every page beside every figure, per edition.
• Precision. No margin of error is reported, deliberately: classical margins of error do not apply to a self-selected sample, and AAPOR's own guidance says non-probability samples should not report one. The Wilson intervals used on the statistics and where pages describe sampling variability within the people who took the quiz, and nothing more. They are not a claim of representativeness.

Change log

• 2026-04-04 Quiz opens; instrument v1.
• 2026-08-16 Index launches: daily editions, mid-rank percentiles, four bands, regions at 30+.
• 2026-08-18 Snapshot rules v2: retakes and sub-sixty-second takes excluded, exclusion ledger, Wilson intervals and floors, reliability figures, completion time and optional gender, work area, how-found and expected-type fields collected. This page published.
• 2026-08-25 Answer order shuffled per take. Options had been shown in the same axis order on every scenario since v1; the order is now drawn once per take. Scoring is unchanged, so editions before and after remain comparable, with a note that position effects in earlier takes were fixed rather than random.
• 2026-08-27 Instrument v2: answer coverage rebalanced. In v1 the psychopathic, sociopathic and neurotypical options appeared in all twenty scenarios, narcissistic and borderline in seventeen, histrionic in six, so the axes had unequal ceilings and histrionic could never be a highest axis. Each dark axis now appears in exactly sixteen scenarios (uniform 80 ceiling) and neurotypical in all twenty. Tied verdicts, which previously fell to the psychopathic type by an implementation accident, now resolve to neurotypical when it ties and otherwise to the axis leading on the scenario questions. Twelve options were rewritten or replaced in the same pass to carry clearer trait signal. Scoring arithmetic is unchanged (answers over twenty, steps of five), so per-axis curves remain comparable, with the note that dark-axis distributions and primary-type shares shift from this date.
• 2026-09-02 Country recorded from the connecting address at submission (two-letter code only; the address is never stored); takes before this date carry no country, so country pages count from here. Age bracket and work area now asked on the screen before the reveal as well as after it. New cuts published: by country, by age, by gender, by work area, and expected type against the type scored, each counted only among takers who answered, with the same floors as everywhere else. Kanika's own result, sat under the same twenty scenarios, is marked on the curves and shown beside each taker's scores.

Cite this

Aggregates on every Index page and the downloadable histograms at the data page are licensed CC BY 4.0: use them, with a link back to the edition you used. Individual takes are never released.

Numbered, dated facts for quoting live at the statistics page. Written and read by Kanika Batra, diagnosed ASPD, clinically assessed as Factor 1 psychopathy.