How this works
This page exists so that the model can be checked rather than trusted. It describes what is measured, how the arithmetic works, where the political content comes from, and what the whole thing cannot tell you.
- {short} or {full} questions
- 31 / 92
- dimensions
- 9
- traditions
- 34
- Thinkers & figures
- 175
- Parties
- 123
Why nine dimensions
A single left–right line cannot hold the disagreements people actually have. Someone can favour redistribution and oppose surveillance; someone can be traditional in social life and egalitarian in economics. So 9 dimensions are measured separately and never collapsed into one number. The conventional left–right summary is calculated, but it is placed below the nine axes on the result page rather than above them.
How the questions are written
Every question is a concrete situation with a real trade-off, not an abstract question about a label. No ideological label appears in any question or answer option, so that people reason from the situation rather than from their self-image. Each option carries its cost — a tax rise names what it funds, a cut names what it saves — so that no answer is free.
Why the default is 31 questions
The full bank holds 92 questions, but 31 are shown by default. The reason is blunt: a long questionnaire someone abandons measures nothing at all. The short form is a curated subset that satisfies the same balance rules — both poles reachable on every axis, symmetric pull, no ordering or agreement skew — and keeps every dimension and every section. It reaches the same leading tradition as the full bank for every archetypal profile it is tested against, and for most other answer patterns the full bank's leader remains in its top three; disagreements are almost always between two traditions that sit close together. What is lost is precision: each axis is read from three or four questions instead of six to eight. Anyone who wants the finer reading can switch to the full set inside the quiz.
How scoring works
Every option carries an effect between −1 and +1 on one or more axes. A dimension's score is 50 + 50 × (Σ weight × effect) ÷ (Σ weight × largest available effect). Dividing by the largest push each question makes available is what keeps a run of moderate answers near the middle rather than at an extreme. "Not enough information" adds no evidence at all — it is left out of the calculation rather than counted as neutral.
How matching works
Your profile and every tradition, figure and party sit in the same nine-dimensional space, so matching is a distance question. A weighted root-mean-square distance is computed, and similarity = 100 × (1 − distance ÷ 60). A gap of 60 points is treated as zero similarity, because dividing by the theoretical maximum of 100 squeezes every score into the top half of the range and makes a tradition you disagree with entirely read as "50% similar".
How figures are scored
The 175 figures are scored from positions they actually wrote or enacted, with sources. Where the evidence on an axis is thinner, a confidence level is recorded and that axis counts for less in matching; where there is no evidence, the axis is left unscored rather than filled in by guesswork. These are encodings of documented positions, not guesses about what someone privately thought.
How parties are scored
The 123 parties are scored from their own published manifestos and official documents — never from a single politician's remarks, from social media, or from an opponent's characterisation. Each party carries the date of the document its position was read from, because parties change and a 2014 manifesto is not evidence about a party in 2026.
How sources are chosen
Sources are used in this order: official manifestos and documents, parliamentary records, constitutional texts, academic research and primary works, then high-quality journalism. Wikipedia is not the sole source for any contested political claim. Where a primary source exists it is used — Marx rather than a book about Marx.
How uncertainty is handled
Uncertainty is surfaced rather than hidden. Axes you answered few questions on are flagged on the result page and count for less in every comparison. If you answered fewer than 85% of the questions, the result is labelled provisional. If several traditions come back within a few points of each other, the page says plainly that no single tradition stands out instead of crowning a winner. Percentages are shown as whole numbers, because decimal places would imply a precision that is not there.
How bias is minimised
Every question records its wording decisions and its source, and that entire internal table is published on the question audit page. Automated checks run on every change: equal pull available in both directions on each axis, balanced option ordering, and the requirement that agreeing with a statement does not always mean the same ideological direction. Opposing positions are always written in their own strongest form, never assembled out of their opponents' objections.
What this model cannot do
This model cannot tell you what you should believe, or whether your views are right. It is not a measurement: the choice of dimensions, the effect assigned to each option and the position given to each tradition are all editorial judgements, published so that they can be argued with. Nine dimensions still leave a great deal out. And the quiz does not know how much you care about any issue — only which way you lean on it.
If you insisted on a single line
This is the conventional left–right summary, included because people ask for it. It is a lossy compression of the nine dimensions above, and it is the least informative thing on this page.
- Wealth and inequality35%
- The state in the economy20%
- Tradition and change (inverted)25%
- Nation and the world20%