Who Gets Called Racist

A passenger boards a late train. The car is nearly empty and there are two places to sit: one beside a group of young men, one at the far end. Nothing about the men is available except what can be seen from the door — age, sex, dress, manner, and the group they belong to. The choice takes about two seconds, and most people who make it make it the same way.

The decision is a prediction assembled from group features, because in two seconds there is nothing else to assemble one from. It is also unverifiable. Acting on it removes the case that would have tested it, so the passenger who walks to the far end never learns whether the estimate was right, and neither does the one who sits down.

Two operations under one word

The seat choice falls under the same word as a refusal to hire a man after a full interview. They are not the same operation. The first is an estimate produced under a time limit, and it is revisable by design: the moment individual information arrives, it is supposed to be replaced. The second has already had the individual information and did not move. One is a forecast made in the absence of evidence, and the other is a conclusion that survives evidence.

The distinction is available to anyone who uses the word and is almost never drawn while the word is in use. Every argument in this area runs under that condition.

The bounded version of the first operation states its own limits, and the limits are not decorative. Broad strokes apply where individual vetting is unavailable — a two-second judgment, or a policy operating on hundreds of thousands of people who will never be met. Where time and access permit an individual assessment, the individual assessment is better, because individuals deviate from group averages and the deviation is the thing worth knowing. Accuracy is a second limit. A generalization that tracks nothing is not a cheap estimate but a preference wearing the costume of one, and it fails on its own terms before anyone raises a moral objection to it.

What a rate licenses

The arithmetic that supports the seat choice is usually stated as raw shares, which is the form in which it establishes the least.

Group B is 13% of the population and accounts for 56% of murder offenses.
Group A is 87% of the population and accounts for 44%.

Share of offenses per share of population:
  B: 56 / 13 = 4.31
  A: 44 / 87 = 0.51
  Ratio: about 8.5 to 1

What that licenses: the per-person rate differs by roughly eight and a
half times.

What it does not license: that either rate is high. A multiple of a rare
event is a rare event. Under both figures, the passenger at the door is
looking at people who are almost certainly going to hurt nobody.

The seat decision follows from the last line rather than the first. It requires that the rates differ and that the two errors cost differently, since moving costs a minute and staying wrong costs everything. It does not require the risk to be large, and an argument for it built on the risk being large has taken on a burden the figures do not carry.

The same arithmetic caps what the decision establishes. Walking to the far end is a cheap bet at long odds, which is a reasonable bet and not a finding about the men in the seat.

What the accusation does

The accusation does not arrive at a random point. It arrives after a claim about rates, borders, or outcomes has been stated carefully, hedged, restricted to averages, and sourced. The care does not prevent it. Any account of the accusation has to explain that.

The strongest account is that the accusation was never in the business of arguing. A society that wants people not to act on group generalizations has a problem it cannot solve by argument, because some of the generalizations are accurate and an accurate generalization cannot be talked away. What can be done is to make acting on it expensive. The accusation is the price. Understood that way it is a sanction rather than a claim, and the standard complaint from the receiving end — that a moral accusation is not a rebuttal — is accurate without being an objection, since the move never offered itself as a rebuttal.

That account also explains the features that puzzle the accused. Disclaimers fail because there is no misreading to correct: the generalization was stated, understood correctly, and priced. Innocence cannot be established because the sanction attaches to the statement rather than resting on a finding about the speaker, so there is nothing about the speaker that can be produced to lift it.

Three responses are in circulation.

Denial

At full strength, the denial is evidence rather than a plea. A charge about disposition is a claim about a mental state; the person holding the mental state has better access to it than any observer; and his account of it is admissible for the same reason a report of pain is admissible. Alongside it come the standards he applies to everyone, the people he has hired and promoted and befriended, and the absence of any conduct matching the charge.

The response fails on vocabulary before it fails on evidence. The denial answers the second operation, settled hostility toward individuals, while the charge as framed covers the first as well, so a truthful denial of the second leaves the charge standing. The extended forms of the charge add a component the accused has no access to, which removes his testimony as evidence by construction. Producing individual counterexamples answers a claim about the use of aggregates by supplying individuals, which is payment in the wrong currency.

Underneath the details, entering the denial concedes the point that matters. It agrees that if the label applies, the argument is lost. The dispute has already moved from whether the claim is true to whether the speaker qualifies, which is the move the accusation was making.

The counter-charge

The second response returns the charge. At full strength this is not a debating trick but the ordinary application of a standard: a position that sorts by ancestry, allocates by ancestry, and holds different expectations by ancestry meets the word’s plain definition, and consistency requires the standard applied to one speaker to apply to the other.

The reasoning is valid and it does not accomplish what its user wants. The plain definition is not the operative one. Where the word is defined as prejudice combined with institutional power, the returned charge is answered by stipulation before any evidence is examined, and the exchange collapses into a dispute over the definition. Definitions are settled by whoever writes the curricula, the style guides, and the moderation rules. The exchange can be won at the table under a definition that is being rewritten away from it.

Declining the label

The third response accepts the label and refuses its force, granting the word without argument and returning the conversation to whether the claim is true.

At full strength the reasoning concerns where the label’s power was located. It was never in the content, which is contested. It was in the target’s own agreement that the label was disqualifying, and that agreement is voluntary and can be withdrawn without anyone’s permission. A supporting argument runs on depletion: a term applied to a border policy, a flag, and a biology textbook has been spread across too many things to carry information, and a signal that fires on everything identifies nothing.

The depletion argument holds for the half of the cost that runs through belief. The other half does not run through belief. Employment decisions, platform access, and professional licensing are imposed by people who need only apply the term, not credit it, and a term can be worthless as a description while remaining fully operative as a trigger. Withdrawing agreement removes the target’s participation in the sanction and leaves the sanction in place.

The accusation understood as a price rather than a claim is not answered here, and the page does not claim otherwise. The strongest version of its defense concedes in advance everything the three responses establish: it does not rebut, it cannot be falsified, and it does not distinguish the two operations the word covers. None of those are capacities it needs in order to impose a cost, which is what it is for.

Whether the charge runs both ways

The definition that makes the charge asymmetric is stronger than its critics allow. The word names a system rather than a feeling. Private animus exists everywhere and always has; what earned it a separate word in one direction was that it was compounded by statute, enforced by the state, and built into institutions whose outputs have not finished running. A word covering both a slur muttered on a bus and a legal order organizing a country has lost the distinction worth marking, and the power condition is an attempt to keep it.

The condition does its work at exactly the point in dispute. Whether a given population can act at scale is a question about the world, and a definition that answers it in advance converts every case of that population acting at scale into a misuse of the term rather than an instance of the thing. The condition is also applied inconsistently by the people who hold it, since the same definition is brought against individuals who occupy no institutional position at all, which the power requirement should exclude.

The deeper problem is that the condition is indexed to a distribution of offices rather than to ancestry, so it travels with the offices. Where the curricula, the hiring committees, and the moderation policies come to be held by a different set of people than held them when the definition was drawn, the definition carries the charge across to them. A term built to point permanently in one direction points that way only while the fact it depends on stays fixed.

What the condition is trying to mark is real, and the vocabulary is worse without it. Animus backed by the power to act on it and animus with no reach differ in everything except the feeling. The failure sits in the repair rather than the diagnosis: the difference was marked by editing an existing word instead of minting a second one, and an edited word drags its old force into its new boundaries.

Set the definition aside and the question becomes empirical. Measured in-group preference runs highest in one group and lowest in another, and one group records more murder offenses than murder victimizations. Since same-group killings enter both counts and cancel, the excess is interracial. Taken as given, those figures establish that the trait is distributed across groups and that the direction in which the charge is usually pointed is not the direction in which the measure is highest. They establish nothing about why.

One test proposed for the other half of the question asks for a present-day legal discrimination producing a named disadvantage, and rules out disadvantages evidenced by outcome gaps alone. The test is answerable in one direction: race-conscious hiring and admissions programs are current, legal, and explicitly differential. It is much harder to answer in the other, which is the result its construction produces. What the construction does is substitute one question for another. A claim about systemic disadvantage is not ordinarily a claim about a statute — it is a claim about mechanisms running through inheritance, private conduct, and the residue of statutes repealed within living memory. Restricting admissible evidence to current statute is a decision about what counts, and it has to be defended rather than built into the challenge.

Accounts of the asymmetry

Three accounts are offered for why measured in-group preference would run in the direction it does.

The first traces present hostility to documented history: slavery, then legal segregation. Against it stands the observation that nobody alive held or was held, that grievances of this kind ordinarily decay, and that other populations subjected to severe discrimination in the same country decayed theirs. The argument that dismisses it contains the concession that keeps it alive. A historical cause can reach the present only through transmission, and if it is reaching the present then transmission is occurring. That is not a refutation of the first account. It is the first account with its mechanism made explicit, which is what the second account already is. The other populations function as a cap rather than an elimination: they show that transmission is not automatic, which was established the moment transmission was named as the requirement, and they do not show it is absent here.

The second account is that the attitude is taught. Curricula present one historical practice as unique to one civilization when it was general across civilizations, coverage selects the incidents supporting a picture, and the picture is that the country is irredeemably hostile. At full strength the claim is not about bias in the abstract but about selection with a measurable output: a population holding a belief that the selected evidence supports and the unselected evidence does not.

The step from selection to attitude is the load-bearing one and it is asserted. The population exposed to the coverage is also the market for it, and coverage that sells to an existing attitude and coverage that manufactures one are indistinguishable from the fact of exposure. What this account has that the first lacks is an available mechanism. What it lacks is anything establishing which way the mechanism runs.

The third account is that the attitude pays. Organizations formed around it received large corporate transfers, programs are justified by it, and individuals earn a living producing it. At full strength the claim is not that everyone holding the attitude is paid, but that its persistence carries a subsidy, and subsidized behavior outlives the conditions that first produced it.

An explanation that discredits a position by naming what its holders earn works against every position whose holders earn, and there is a payment on this side of the argument as well. Audiences pay for accounts of the world, and the accounts that pay best are the ones the audience arrived wanting. Carried to its own end, the argument removes the standing of any monetized claim including itself.

What survives the reversal is narrower and is the part worth keeping. An incentive explains why something continues after its original cause has stopped operating. It does not establish that the thing was built to collect the incentive, and a subsidy found attached to a behavior is entirely compatible with the behavior having arrived first.

A fourth account is held in reserve, that the difference is heritable, and it arrives with its own procedure attached: confirm it by eliminating the alternatives. The procedure is sound, and it makes the reserve account wholly dependent on the quality of the elimination. An explanation reached by exhausting rivals is established exactly as far as the rivals were exhausted, and on the record above none of the three was eliminated. Each survives in some form, the first with its mechanism made explicit and the other two narrowed.

The test that separates them

One test distinguishes the two operations the word covers, and it is cheap to run. A judgment that changes when individual information arrives was an estimate. A judgment that does not change was something else, whatever produced it. The passenger who moves to the far end and then, recognizing a colleague, sits back down was forecasting. The one who does not sit back down was not.

The test runs in the other direction as well. A charge that would be withdrawn if some particular fact about the accused turned out otherwise is a claim about him, and claims about people can be argued. A charge that survives every fact about him is a claim about his category.

None of this settles whether any particular generalization is accurate, and none of it settles how common the conduct the word names actually is. What it settles is narrower. The accusation operates as a price rather than as an argument, and the definition currently attached to it disposes of the symmetry question by stipulation instead of answering it. Anyone who agrees to answer the charge is answering for both operations the word covers, including the one he did not perform.

Links

• Validity and Truth — the rule that premises stand unless overturned, which is what lets the crime figures here be used without being litigated.
• Per Capita — the rate arithmetic the seat case depends on, worked through on its own terms.
• Racial Egalitarianism - The Disparity Inference — the step from an outcome gap to a cause, which the systemic-disadvantage challenge rules out by construction.
• White Guilt — the transfer the incentive account requires, examined as a mechanism rather than as a motive.
• Racial Fatigue — the same sanction traced from the receiving end, where a full concession of the substantive claims left the label unchanged, which is the price account with a case attached.
• Moral Language as Leverage — why a moral vocabulary works on the people who hold it even when the person using it does not.

Open questions

If the label’s persuasive force falls with overuse while the institutional cost it triggers does not, what would have to change for the institutional cost to fall as well, and does anything in the persuasive depletion cause it?

The update test separates an estimate from a settled judgment by watching what happens when individual information arrives. A policy operating on a population never produces an individual encounter. What plays the role of the test there?

Sources

How To Be Responsibly Racist, 2026-06-15 — https://www.youtube.com/watch?v=oGRNJhS2bhI. Supplied the seat case, the conditions bounding group-level judgment, the accuracy constraint, the per-capita figures worked above, and the reserve account with its elimination procedure.

You’ve Just Been Called Racist, Now What?, 2025-12-26 — https://www.youtube.com/watch?v=yYIITZs0C8A. Supplied the three responses to the accusation and the depletion argument.

Why Is Black America So Racist?, 2026-06-29 — https://www.youtube.com/watch?v=WunIYGEYeD0. Supplied the symmetry question, the three accounts of the asymmetry, and the systemic-disadvantage challenge.