Model assumption
How are the estimated population shares of the types calculated?
A reproducible hypothetical scenario; the actual population shares are currently unknown.
This page explains the method and current basis for the “Population Share (Model Estimate).” We publish the assumptions, calculations, and version so you can see where the figures come from.
The actual shares of the 16 Shades types in the population are currently unknown. The percentages on this page come from a publicly specified hypothetical scenario and have not been calibrated against real population data. The figures shown here demonstrate the method. Any figures used on later results pages should correspond to the scenario version cited there.
What Does This Figure Mean?
It means that if the four tendencies are distributed in the proportions we specify, and if the stated assumptions about how they combine hold, each type would account for a certain share within this model.
For example, if a type has a model share of about 13%, it accounts for about 13% of this hypothetical distribution. That cannot be translated into “13 out of every 100 people in the real world share your type.” A model can turn assumptions into numbers; calculation alone cannot show that those assumptions resemble reality.
These types describe people's strategic tendencies in relevant situations. Names such as “Tyrant” and “Broker” are personified labels. A name alone cannot tell us that a type is rarer, and population share cannot tell us how bad or admirable someone is, or whether they will harm others.
Why Start with Four Axes?
The sixteen types are combinations of four tendencies, always in this order:
| Axis | Two poles | Question examined |
|---|---|---|
| Goal | Gain / Control | When tangible gain and decision-making power cannot both be secured, which would the person rather retain? |
| Path of influence | Strategy / Pressure | Does the person rely more on arranging information, options, and relationships, or on explicitly setting conditions and consequences? |
| Conflict closure | Instrumental / Punitive | After the external objective or remedy has been achieved, does the person still want the other party to pay an additional price? |
| Justification | Justification / Disregard | Does the person rely more on shared rules and reasons, or more readily authorize a choice through their own priorities? |
Choosing one pole on each axis produces sixteen combinations. Broker, for example, corresponds to Gain–Strategy–Instrumental–Justification, while Tyrant corresponds to Control–Pressure–Punitive–Disregard. Publishing the distribution assumptions for the four axes first and applying one calculation rule keeps the figures for all sixteen types internally consistent and easy to check.
These sixteen combinations are the classification defined by this project. There is currently no evidence that people naturally cluster into these sixteen categories. Nor does the existence of sixteen types imply that each must account for 6.25%.
What Assumptions Does the Current Demonstration Use?
Demonstration Scenario A, version type-share-scenario-0.1, recorded on 2026-09-07.
Input Proportions for the Four Axes
| Axis | Demonstration input | Scenario examined by this input | Basis for the figures |
|---|---|---|---|
| Gain / Control | 60% / 40% | Somewhat more model subjects favor tangible gain | Manually chosen demonstration parameters; not empirically measured |
| Strategy / Pressure | 55% / 45% | The two paths of influence are close to balanced, with indirect arrangement slightly more common | Manually chosen demonstration parameters; not empirically measured |
| Instrumental / Punitive | 65% / 35% | Somewhat more model subjects tend to stop once the external objective has been achieved | Manually chosen demonstration parameters; not empirically measured |
| Justification / Disregard | 60% / 40% | Somewhat more model subjects rely on shared grounds to authorize their choices | Manually chosen demonstration parameters; not empirically measured |
These figures make one imagined scenario concrete. We currently have no data showing that 60% is closer to reality than 55%, and no evidence that the relative ordering in the table matches any real population. They are not statistical findings drawn from psychology papers, a distribution of website users' assessment results, or a sample produced by AI-simulated respondents.
How the Four Axes Are Combined
For this demonstration, the four axes are provisionally treated as independent. In other words, which pole a model subject favors on one axis does not change the proportions of their tendencies on the other axes.
This assumption was chosen to simplify the calculation. The fact that the four questions can be distinguished conceptually does not prove that they are unrelated in reality. The same four input proportions can produce a different distribution across the sixteen types if different relationships among the axes are assumed.
Who Is Included in the 100% Denominator?
The denominator in this demonstration is an idealized set in which every model subject has a definite direction on all four axes and belongs to exactly one type. The sixteen types in the table total 100% only within this set.
Real assessments may produce ties, near-boundary results, insufficient information, or multiple candidates. This demonstration does not estimate how common those cases are in the population, or what share of the total population can be assigned an unambiguous type. The figures in the table therefore cannot directly represent the entire population.
This distribution has not been calibrated for any country, age group, occupation, or other real-world population. A classification of everyday tendencies also cannot be converted into the incidence of harmful behavior.
How Is It Calculated?
Under the independence assumption, a type's share equals the product of the proportions for its corresponding poles on the four axes.
For example, Broker corresponds to Gain–Strategy–Instrumental–Justification:
Broker model share
= Gain proportion × Strategy proportion × Instrumental proportion × Justification proportion
= 0.60 × 0.55 × 0.65 × 0.60
= 0.1287
= 12.87%Tyrant corresponds to Control–Pressure–Punitive–Disregard:
Tyrant model share
= 0.40 × 0.45 × 0.35 × 0.40
= 0.0252
= 2.52%Every other type uses exactly the same rule. Because the two poles of each axis sum to 1, the sum of all combinations is:
(0.60 + 0.40) × (0.55 + 0.45) × (0.65 + 0.35) × (0.60 + 0.40)
= 1, or 100%The Sixteen Types in Demonstration Scenario A
All figures below are calculated values from a hypothetical scenario; the actual population shares are unknown. Display values are percentages rounded to one decimal place. This is a presentation choice and does not mean the estimates are accurate to 0.1 percentage points.
| ID | Type | Four-pole combination | Unrounded model percentage | Display value |
|---|---|---|---|---|
| T01 | Broker | Gain–Strategy–Instrumental–Justification | 12.87% | 12.9% |
| T02 | Opportunist | Gain–Strategy–Instrumental–Disregard | 8.58% | 8.6% |
| T03 | Smiler | Gain–Strategy–Punitive–Justification | 6.93% | 6.9% |
| T04 | Pursuer | Gain–Strategy–Punitive–Disregard | 4.62% | 4.6% |
| T05 | Harvester | Gain–Pressure–Instrumental–Justification | 10.53% | 10.5% |
| T06 | Raider | Gain–Pressure–Instrumental–Disregard | 7.02% | 7.0% |
| T07 | Collector | Gain–Pressure–Punitive–Justification | 5.67% | 5.7% |
| T08 | Bruiser | Gain–Pressure–Punitive–Disregard | 3.78% | 3.8% |
| T09 | Puppet Master | Control–Strategy–Instrumental–Justification | 8.58% | 8.6% |
| T10 | Agitator | Control–Strategy–Instrumental–Disregard | 5.72% | 5.7% |
| T11 | Power Broker | Control–Strategy–Punitive–Justification | 4.62% | 4.6% |
| T12 | Cult Leader | Control–Strategy–Punitive–Disregard | 3.08% | 3.1% |
| T13 | Enforcer | Control–Pressure–Instrumental–Justification | 7.02% | 7.0% |
| T14 | Decider | Control–Pressure–Instrumental–Disregard | 4.68% | 4.7% |
| T15 | Arbiter | Control–Pressure–Punitive–Justification | 3.78% | 3.8% |
| T16 | Tyrant | Control–Pressure–Punitive–Disregard | 2.52% | 2.5% |
| — | Total | — | 100.00% | 100.0% |
The display values in this scenario also happen to total exactly 100.0%. With other parameter combinations, rounded values may not total exactly 100%. Any rounding difference should be disclosed and the original calculated values retained; no individual type should be adjusted merely to manufacture differences among types.
How Much Do the Figures Change Under Different Assumptions?
We used several alternative settings to examine how strongly the figures depend on the assumptions.
| Demonstration setting | Gain / Strategy / Instrumental / Justification proportions | Broker model share | Tyrant model share |
|---|---|---|---|
| Symmetric baseline | 50% / 50% / 50% / 50% | 6.25% | 6.25% |
| Demonstration Scenario A | 60% / 55% / 65% / 60% | 12.87% | 2.52% |
| Perturbation Scenario B | 55% / 50% / 60% / 55% | 9.075% | 4.05% |
| Perturbation Scenario C | 65% / 60% / 70% / 65% | 17.745% | 1.47% |
All rows above provisionally treat the four axes as independent. The symmetric baseline shows only the mathematical consequence of assuming an even split at each pole and independence among the axes; it does not claim that the real population is evenly divided. Scenarios B and C respectively decrease or increase all four inputs from A by 5 percentage points. Those magnitudes are also manually chosen test conditions.
Keeping the four axis proportions fixed is still not enough to determine a single answer. For example, A assumes that subjects who are both Strategy and Justification account for 33% of all model subjects. A different relationship among the axes could put that joint share at 40% while preserving the marginal proportions of 55% Strategy and 60% Justification. With all else unchanged, Broker's share would then rise from 12.87% to 15.60%. Both are calculable scenarios, and we have no data for judging which is closer to reality.
These results show that the model is sensitive both to the input proportions and to relationships among the axes. The range shown here is a range of calculations produced by changing assumptions. It is not a margin of error for the actual population share or a 95% confidence interval.
How Does This Relate to My Assessment Match Percentage?
You may see several kinds of percentages in your results. They answer different questions:
| Figure | What it means |
|---|---|
| Four-axis tendency reading | The relative tendency shown by this set of answers on one axis |
| Character match percentage | The relative share of fit between this set of answers and each character under the fixed combination rules |
| Population Share (Model Estimate) | A type's share within the model under a given set of distribution assumptions |
The population share estimate is not derived from an individual's match score. Under the same scenario version, a given type always has the same model share. It is neither the probability that you belong to that type nor a measure of the assessment's accuracy.
Continuing the assessment may update your recommended type, so the type you are viewing and its model estimate may change. This does not mean the population distribution changes in response to your answers.
Which Parts Have a Basis, and Which Remain Assumptions?
| Component | Current status |
|---|---|
| Meanings of the four axes, type names, and mappings | Defined by this project's published classification |
| Inputs of 60%, 55%, 65%, and 60% | Manually chosen demonstration assumptions on this page, with no empirical basis |
| Independence of the four axes | A simplifying assumption chosen for the demonstration; not yet validated |
| Multiplication formula, values for the sixteen types, and total | Reproducible mathematical results |
| Actual population shares, differences among populations, and model error | Currently unknown |
Research on dark personality can inform discussion of related concepts. For example, research on the D factor examines a common core shared by multiple dark tendencies. Comparing its definition with this project's four axes does not allow the D factor's statistical findings to be converted directly into population shares for the 16 Shades types. The research team's definition and publication index
No mapping from other personality inventories or public survey data to this project's four axes has yet been validated. This page does not use type distributions from other assessments, occupational counts, or rates of clinical diagnoses as the numerical source for the sixteen types. Existing research that has not been validated against this model is not validation of the model.
How Would Future Website User Data Be Handled?
Assessment results accumulated organically on the website could describe “the share of each type among completed assessments for this version.” Whether those results represent any other population would still depend on where users came from and on the statistical methods used. For a voluntary online sample, the sample source and conditions for inference must be stated; a larger sample does not automatically guarantee population representativeness. AAPOR guidance on survey research methods
If such data are published in the future, they should be separately labeled “Distribution of Assessment Results on This Site.” The publication should state the time period, question and scoring versions, basic or full stage, unit of analysis, and handling of repeat attempts, and account for results that do not yield a unique type. If only submissions can be identified, the number of result records should be reported rather than described as a count of unique people.
Statistics from actual results and the hypothetical model should retain their respective sources. Publishing this page does not mean that users' item-level answers have been collected or validated, or that any new collection of those answers is planned.
How Can the Figures Be Traced and Updated?
For every published set of figures, readers should be able to find the input proportions, relationships among the axes, denominator, calculation rule, and date. When input or relationship assumptions change, the previous version and the reason for the change should be retained. Figures on results pages should link to the version of this explanation that produced them.
| Date | Version | Record |
|---|---|---|
| 2026-09-07 | Content 0.1.0 / Demonstration Scenario type-share-scenario-0.1 | First published explanation; includes four-axis assumptions, a demonstration table for all sixteen types, and a sensitivity demonstration; actual population shares remain unknown |
Publishing the assumptions makes the figures understandable, checkable, and open to discussion. Evidence is still required to determine how closely the estimates reflect reality.