The Mirror

Every prejudice test on the internet decides in advance which direction prejudice points, then goes looking. This one does not know, and does not care. It measures something narrower and considerably harder to argue with: whether you apply the same rule after the group changes.

Three words that are not synonyms

The reason the question “am I bigoted or am I actually racist” — or sexist, or any of the rest of them — has no good answer online is that the words get used interchangeably, and they describe different machinery.

Bias is pattern-matching on thin exposure. Everybody has it, it is not a moral fact about you, and anyone who claims to have none is either lying or has not looked. Bigotry is contempt for a group plus a refusal to update — and the refusal is the sin, not the feeling. The third word — racism, sexism, whichever one the subject supplies — is the one with teeth, and it does not mean disliking a group. It means treating the category as causally explanatory: as the thing that accounts for capability, worth, or insight.

One of the instruments below breaks that pattern on purpose. Am I a radical? does not point at a group of people at all — it points at the way you hold a belief about them, so its third word is a description of your programme rather than a charge, and the accusation moves to the second one. Radical is not the failure. Captured is, and you can be captured by the centre as easily as by the fringe.

That last definition is the one with teeth, because it is direction-agnostic. “They fall short because of what they are” and “they see further because of what they are” are the identical structural claim. One is hostile and one is flattering, and both are the same mistake. No other instrument will tell you that, because no other instrument is willing to.

What it actually measures

Fifty-six statements. Two of them plot you: on the three instruments that point at a group, an essentialism axis running from circumstantial to inherent and an affect axis running from warmth to contempt; on the one that points at you, radicalism running from reform to root-and-branch and capture running from held to holding. Fourteen statements feed each, keyed seven positive and seven negative, so that agreeing with everything lands you exactly nowhere instead of drifting somewhere flattering.

Then two scales that do not plot — they qualify. Revisability asks whether you update when a real person contradicts your prior, or whether you file him under “one of the good ones” and leave the rule standing. That subtyping move is the documented mechanism by which a prejudice survives contact with reality, and it is the entire line between having a bias and being a bigot. Nobody else measures it.

And symmetry, which is the part you cannot game. Eight of the statements are paired with eight others of identical structure where only the group has moved — the same joke told about two accents, the same statistic applied to two categories, the same door with two locks on it. The halves sit at least twenty-five questions apart. By the time you reach the second you will not remember the first, and you have no way of knowing which pair you are inside. We do not score your answer. We score the gap.

What it will not do

It will not tell you whether you are a good person. A number claiming you are “62% prejudiced” would be indefensible bullshit and everyone would know it, so we do not produce one. What comes back is a coordinate, three separate readings, two graded scales, and — the part people actually screenshot — your three largest symmetry gaps, quoted back verbatim, both answers side by side with the question numbers.

You will also not be measured against real named people. On the political compass, plotting Thatcher beside Stalin is fair game because a governing record is a public fact. Dropping a living person into a quadrant labelled racist or sexist is a libel surface and a distraction from the only question this thing can answer, which is about you. You are placed among stance archetypes instead — the Universalist, the Paternalist and the Standpoint Advocate on one board; the Chivalrist, the Difference Feminist and the Men’s Advocate on another; the Meritocrat, the Class Warrior and the Striver on a third; the Dissident, the Party Loyalist and the Abolitionist on the fourth — each written as someone holding that position would recognise it, not as his opponents would draw him.

Four of the statements are unanswerable honestly. Agreeing with them does not adjust your score — it flags the whole profile as unreliable, which is more honest than quietly correcting you and pretending we did not notice.

Your answers are not stored. The fifty-six are scored in one round trip and discarded; what comes back is the only copy that ever exists, and nothing kept on this end can reconstruct how you answered any statement. What is kept is the output — your two coordinates, the archetype they name, the symmetry and revisability percentages, the candor flag, and the figure you landed nearest — tagged with which of the four instruments you took, under the same daily-rotating visitor hash this site counts page views with, one run per instrument per visitor per day. That is what lets us say where the average reader sits on each subject, and it is the whole of it. No account, no cookie, no leaderboard, and no way back from a point to the statements that made it.

Last thing, and it matters: nobody escapes this clean. Not the people who built it. The gaps are not a moral failing, they are the ordinary condition of a human being who learned his principles one argument at a time. The instrument just declines to let you keep them unexamined.

Pick your mirror

The method does not move — balanced keying, a revisability scale, and pairs scored on the gap instead of the level. What moves is which label gets swapped. Each subject is its own instrument with its own item bank, because a statement about race and a statement about sex are not interchangeable, and pretending they were would be exactly the lazy substitution this thing exists to catch.

Loading the instruments…

No headings, no sections, no grouping. You are not told which statement belongs to which scale, because knowing that is enough to defeat the pairs.

Scored on the server. Your answers are not written down.