# Which LessWrong Character Are You?

[Find your character](lesswrong-atlas-games.html#tribe-compass) · [Explore the map](lesswrong-atlas-explorer.html)

## An anonymous test

Answer **48 statements** in one continuous, shuffled test. Every answered statement contributes to at least one of your nine result bars. Authors, sources, and the 20 public figures used for comparison stay hidden until results.

Every statement has the same choices: **Strongly disagree, Disagree, Neutral, Agree, Strongly agree**. A left/right swipe means Disagree/Agree; use the buttons for stronger answers. Neutral is a midpoint answer. **Skip** provides no evidence. Back lets you change an answer.

No name, social-media profile, account, or saved progress is requested. Answers stay in this page’s memory and are cleared when the page reloads. Previous versions’ locally saved compass data is removed. The site does not upload your answers. The result card is sized for phone screenshots.

## How the bars work

Responses are coded as **−1, −0.5, 0, +0.5, +1**. Each statement has a signed weight on the axes it addresses. A positive weight points agreement toward the right-hand label; a negative weight points agreement toward the left.

For each axis:

```
axis = sum(answer × weight) / sum(abs(weight))
right percent = (axis + 1) × 50
left percent = 100 − right percent
```

Only non-skipped answers enter either sum. Neutral adds weight to the denominator and zero to the numerator. An unanswered axis remains unknown, not 50/50. Displayed percentages are rounded to whole numbers.

Strongest answers consistently pointing toward one pole produce **100% for that pole**. Moderate agreement produces a smaller shift. Opposing answers offset one another according to their weights. Because some questions affect multiple axes, not every possible combination of nine extreme bars is logically consistent with the questionnaire.

Each result includes “Answers behind this axis,” so you can inspect exactly which statements contributed and in which direction. Percentages describe the direction of your answers; they are not membership probabilities or confidence levels.

## Your nine axes

- Rationalist ↔ Post-rationalist
- Near-term welfare ↔ Longtermism
- Liberal/pluralist ↔ Post-liberal/order
- Voice ↔ Exit
- AI precaution ↔ AI acceleration
- Human-centered ↔ Animal/digital concern
- Institutional reform ↔ Founding experiments
- Growth/abundance ↔ Restraint/limits
- Monogamy ↔ Polyamory

The relationship bar has **four scored preference statements**: polycule openness, preference for a two-person exclusive relationship, a cuddle-puddle scenario explicitly involving another romantic partner, and exclusivity versus consensual exploration such as an orgy. These ask about your preferences. They do not infer them from sexual history or ordinary non-sexual group cuddling.

For example, strongly agreeing with the polycule and polyamorous cuddle-puddle statements while strongly disagreeing with the two exclusivity statements produces **100% Polyamory**. Reversing those four answers produces **100% Monogamy**. Answering all four neutrally produces **50/50**. Relationship answers affect your bar only, never a public figure’s chart or your closest-person match.

The axes compress overlapping ideas. EA includes near-term and longtermist approaches; animal and digital welfare can differ despite sharing an axis here. The questionnaire is an editorial model, not a validated psychological assessment.

## Your closest person

Your public-view answers are compared with **20 sourced reconstructions of public positions**. The people have not taken this test. Their documented positions use the same five-point scale. Unknown positions are omitted, never filled in as neutral, and each includes a source and a rationale. Historical changes and evidence limitations are noted in the profile.

For every statement with both a user answer and a documented public position, we calculate the absolute answer difference. We average those differences within each relevant axis, using absolute question weights, then weight each covered axis equally. Similarity is **100 × (1 − average distance / 2)**. Personal relationship statements are always excluded.

A person needs at least **five shared documented answers across two axes** to qualify. Your overall result needs at least **12 non-skipped worldview answers across five public axes**. Profiles with fewer than eight shared answers or fewer than three shared axes are marked as limited coverage. Coverage appears beside each comparison.

The nearest eligible profile is shown as “You are most like.” Exact ties remain ties. All-neutral answers, insufficient evidence, or no eligible profile above the 60/100 threshold produce an explanation instead of a forced match. This threshold is a design choice, not a scientifically validated cutoff. A match does not imply a shared identity, affiliation, or agreement with every belief that person holds.

## Explore the public profiles

After finishing, browse Holden Karnofsky, Dario Amodei, Scott Alexander, Curtis Yarvin, and the other researched figures. Their charts come from documented public writing and remain independent of your answers. A blank public-view axis means no documented position on its questions, not neutrality.

User bars and profile bars can cover different statements. The similarity score compares shared answers directly, rather than comparing the displayed bars. The original 36 statements support the public profiles; the 12 additional statements enrich your own axes without inventing new answers for those figures.

The five-response format and paired bars are inspired by [9Axes](https://9axes.github.io/). This atlas uses its own questions, weights, and matching method. Profile sources were reviewed September 23, 2026; the additional question sources September 24, 2026.

## Portraits

All 20 result profiles have local portraits. [Image sources and credits](atlas-games/portrait-sources.json) include official speaker profiles, personal websites and institutional biographies.