About

I built this because the political tests I had used all seemed to be answering a narrower question than the one I wanted to ask, and doing it with more confidence than the evidence supported.

Why

Almost every test of this kind reports a single left and right number. That number has to carry at least three separate arguments at once: who should own productive capital, how much a society should change, and how much concentrated power over a person is acceptable. Those come apart immediately in practice. A communist and a social democrat agree on ownership and disagree about the state. A libertarian and a conservative agree about the state and disagree about everything else. Collapsing them onto one line means the most interesting disagreement a person has is the one the result cannot show.

The second problem is precision. A result of "economic left: −7.43" is a claim to two decimal places about a quantity nobody has measured. It tells you where you are in a ranking and nothing about how firmly the questions placed you there. I wanted the estimate and the uncertainty together, because the uncertainty is often the more useful half.

What it does

Three axes, measured independently: Ownership: who owns and directs productive capital, and how the product is allocated. Change: continuity against openness, and how wide the moral community is drawn. Authority: how much concentrated, unaccountable power over a person is acceptable, whether it sits in a state, an employer, a platform, a party or a family.

Each is reported 0–100 with a confidence interval, so a result reads as a claim with a margin rather than a fact. The three together place you in a cube you can rotate, and you can compare that position against forty-two historical figures whose placements are each argued axis by axis. Results export as a labelled image or an animated rotation.

What it does differently

Three axes instead of one

The axes are designed to be genuinely separable rather than three views of one thing. That is what makes positions like authoritarian socialism and libertarian capitalism legible instead of contradictory, and it is why the test can tell the difference between someone who wants a different economy and someone who wants a different society.

The ownership axis measures ownership, not government

This is the decision I would defend hardest. On most scales, "more state" reads as "more left". But state ownership is not collective ownership: a firm run by a ministry is not run by the people who work in it. Because this axis asks who holds capital rather than how large the state is, state capitalism sits on the private-ownership side where it belongs, and libertarian socialism (communes, cooperatives, federations with no state at all) stays visible on the left instead of being lost.

Every item is audited for the assumptions it smuggles in

A political test can be biased by its grammar as easily as by its politics. An early draft of this bank asked how "earnings from work" should be organised and offered four ways of setting wages, which presupposes that work is paid and leaves anyone who rejects wages as a category choosing the least capitalistic of four wage arrangements. Another asked on what basis credit should be created, presuming that lending at a return is the ordinary form.

There are now automated audits for that class of error. One checks every multiple-choice item for a stem that names an economic category (wages, profit, interest, rent, price, employment) without any option rejecting that category, and the rule is that the stem gets rewritten, not just the options. Another checks the bank for duplicated and near-duplicated questions, because a repeated question is counted twice as evidence. A third verifies that the two poles of each axis carry equal item counts, so a high score reflects the respondent rather than the bank.

The full range, written seriously at both ends

The positions outside the ordinary spectrum are present as developed arguments rather than as caricatures: 355 items across 121 subject domains, with 13% at the extreme difficulty band. Marxist and anarchist positions appear as their adherents state them. Fascist and reactionary positions appear the same way: as claims about order, continuity and belonging that their holders would recognise, never as atrocity. Getting both ends written fairly is harder than it sounds, and it took several revisions.

It asks how you analyse politics, not only where you stand

54 items measure something a position score cannot see: whether you read political outcomes as shaped mainly by material conditions, class and economic power, or mainly by choices, ideas, institutions and chance. That distinction comes from Marx, and separating it from his politics is the point; it is held across the spectrum, including by conservatives who think economics drives politics, and rejected by plenty of socialists. Two people with identical views on every axis can analyse politics completely differently, and this test can tell them apart.

Three answer formats, because they fail differently

Agree/disagree invites acquiescence. True/false forces a side on questions where the honest answer may be that the premise is wrong. Multiple choice, where each option is itself a position on the axis, is the only format that lets a respondent reject a question's premise instead of picking the nearest thing to their view. All three are present, and the adaptive test serves a deliberate mix rather than whichever is most informative.

The method and the limits are published

The item bank, the model, the scoring code and the audits are all readable, and the method page states the gaps rather than only the strengths. That includes the one that matters most: the item parameters are provisional. They encode an explicit judgement about what each statement measures and how strongly, and they have never been fitted to real response data, because no such data existed until this site started collecting it anonymously.

What it does not do

It does not tell you what you are. Three axes are not the whole of politics: they leave out economics outside ownership, foreign policy, federalism, religion, and most of what people actually argue about. It does not place you in a named ideological zone, because those labels carry baggage the axes do not measure and reading "you are a social democrat" off a coordinate invites more confidence than the coordinate supports.

And it does not yet know how accurate it is in the way that matters. Accuracy measured against simulated respondents is about 2 points per axis on average for the full test, but that simulation assumes the item parameters are right, which is the very thing in question. Of fourteen checks a psychometric instrument should pass, seven have been run: internal consistency, information against a closed form, the audits above. The other seven, including test-retest reliability and validity against an established instrument, need a few hundred real respondents. The data collection is live for exactly that reason.

Who made it

Alice. Built as a personal project, published in full so the scoring can be checked and argued with: alicemow.org · alice@herkula.info. Corrections are welcome, particularly to item wording.

Comparative statements here describe this instrument's own design decisions. Where they mention a common pattern in other tests, that is an observation about the pattern rather than a claim about any particular site, and other tests are worth looking at directly.