Methodology

How Polyaxis Works

Polyaxis is built on a simple conviction: two axes cannot describe a political mind. This page is the short version of our white paper — what we measure, how the questions are constructed, how scoring works, how we detect the tensions and contradictions in your answers, and what the honest limitations are.

1. What we measure

Polyaxis scores you on 18 independent dimensions: 11 core axes (what you believe about how society should work) and 7 style facets (how you pursue those beliefs and who you think should decide). Each dimension is bipolar — a score from −100% to +100% between two named poles, where 0 means balanced or conflicted, not apathetic.

Core axes

Style facets

2. The question bank

The evaluation contains 286 statements answered on a five-point agree–disagree scale, in two deliberately different registers:

• Conceptual items state a principle in the abstract (“Ecosystems and species have intrinsic value…”). They measure your ideals.
• Applied items put you in a concrete scenario with a real cost attached (“…even if it eliminates jobs”). They measure what you actually choose, and they carry 15–25% more scoring weight because behavior under constraint is more informative than assent to a principle.

Every axis has an equal number of items keyed toward each pole, so a tendency to agree with things (acquiescence bias) cannot masquerade as an ideology. Question order is shuffled per session with consecutive questions never probing the same axis, skipping is allowed (skipped items are excluded rather than guessed), and there is no time limit.

How the questions are distributed

The distribution above is computed from the live question bank, not hand-maintained. Coverage is audited on two levels: topic coverage (which policy domains the scenarios draw from — economics, liberty, culture, ecology, technology, infrastructure, defense, democratic process, and more) and conflict coverage (which pairs of values the tradeoff scenarios actually force against each other). The bank carries 48 deliberately designed collision scenarios spread across 24 axis pairs, each pair probed once in each pole direction, and every question carries semantic coverage tags (policy domain, latent conflict, actor level, policy instrument) that feed the audit views shipped with the database.

3. Scoring

Each answer contributes response × key × weight to its axis, and the axis score is the weighted average normalized to −1…+1. Some questions carry carefully bounded secondary loadings onto a related axis (capped at half the question’s weight so no axis is ever dominated by questions written for another). Your conceptual and applied scores are also computed separately, so you can see where your ideals and your choices diverge.

4. Value tensions — the heart of Polyaxis

Most political tests only ask what you believe. Polyaxis also measures what wins when your beliefs collide. A subset of applied items are tradeoff scenarios: each one explicitly prices one value against another (civil liberties against tech-enabled security, ecology against prosperity, local control against housing, tradition against personal freedom, and more). These items are excluded from ordinary axis scoring — choosing a side in a dilemma is not the same as holding a position — and analyzed separately:

• Every answer to a tradeoff scenario records which value won that collision and how decisively.
• Across the scenarios for a pair of values we count wins, compute a lean, and classify the result: a consistent priority (you reliably rank one value above the other), a context-dependent pattern (each side wins some scenarios — nuance, not noise), or no clear pattern.
• We then compare against your conceptual scores. If you endorse both values in principle, the scenarios forced a genuine dilemma — and the ranking they reveal is telling. If the value you profess more strongly consistently loses, we flag a contradiction with your stated ideals: possibly the most useful single insight the evaluation produces, whether you read it as self-knowledge or as a spot to revisit for consistency.

Each collision scenario is deliberately designed for its axis pair, and every reported tension is labeled with the number of scenarios behind it and a confidence level, so a single-scenario reading is never dressed up as more than it is.

5. Archetypes

Your profile is matched against 38 political archetypes — from Revolutionary Socialist to Libertarian Capitalist, Deep Ecologist to Eco-Capitalist, Moral Traditionalist to Libertarian Socialist. Each archetype is defined as a weighted set of positions across the axes; your affinity is the weighted average alignment between your scores and those positions. Matches are descriptive, not prescriptive — most people fit several archetypes partially, and the interesting information is usually in the spread, not the single top label.

6. Validation and continuous improvement

The instrument is engineered to be checked, not just believed. Beyond the design controls (balanced keying, single-construct wording, mixed agreement directions within every tension pair), we continuously compute item-level discrimination statistics, per-axis internal consistency (Cronbach’s α), and the full inter-axis correlation matrix from real responses. Items that fail to discriminate or that cross-load onto the wrong axis get reworded or retired; the question bank has already been through one full audit-and-repair cycle, documented in the project repository.

7. Honest limitations

• Self-report. Polyaxis measures your considered answers, not your behavior in the world.
• No population norming yet. Scores are positions on the instrument’s own scale, not percentiles against a representative sample.
• Scenario locality. Some applied scenarios assume institutions (federal states, private healthcare) that fit some countries better than others.
• A model, not a mirror. Eighteen dimensions is far richer than four quadrants, but any finite instrument compresses a political mind.

8. Privacy

The evaluation is anonymous: no login is required and no personally identifying information is collected. Results live at an unguessable URL that only you can share. An optional account exists solely to let you revisit your own results.