On July 24, 2026, the Wall Street Journal ran an essay on sociology’s public unraveling. I’m one of multiple people quoted in it, along with Christian Smith, Kevin McCaffree, Musa al-Gharbi, Saida Grundy, Lisa Hajjar, Melissa Wilde, Jukka Savolainen, and Ashley Rubin.

While the things I was quoted as saying were accurately reported, a quotation is a sentence. An explanation is a structure — a claim, the thing it’s a claim about, what follows from it, what it rules out. You can lift a sentence out of a structure, reproduce it perfectly, set it down inside a different structure, and have it sound like it’s saying something else, without anyone trying to misrepresent anything.

Here is one of my quoted sentences: “We’re treating questions around causation as settled.” On the one hand, I stand behind the statement, knowing that charging a discipline with treating an open question as closed is serious. On the other hand, in a story about whether sociology has been captured by activists, the statement becomes a different accusation than the one I’m making. There it reads more like: sociologists have let their politics contaminate their science. That is a familiar complaint with plenty of adherents, several of them quoted in the same article. And yet, what I’m describing requires no dishonesty, no determination to distort, and no one behaving badly.

But perhaps the most important point of confusion is this: while my claim has implications for sociology, it isn’t about sociology. It’s about a premise treated as fact and where that premise got installed — in employment law, in HR practice, in K–12 curriculum, in how psychological instruments get read, in newsrooms, in higher education, and in sociology.

Appearing in a story about sociology’s troubles shrinks the explanation to something like a tenth of its size, without anyone having said a false word. Perhaps that’s the cost of being quoted in the WSJ, but it’s something I can at least try to clarify here.

The claim

The settled proposition I’m describing is this: disparities in outcomes between groups constitute evidence of discrimination (where discrimination is understood to be an expression of racism).

That is a causal claim, which makes it interpretive by its very nature. Any careful researcher knows that we never observe causation directly. Rather, we observe patterns and sequences and impose an account of how they’re related. This isn’t skepticism for its own sake — it’s the ordinary condition of causal reasoning, and it’s why a claim can be more or less plausible, better or worse supported, without ever converting into a description of what happened. This is especially the case when the subjects are human beings making choices based on motives invisible to outsiders and, sometimes, even to themselves.

The legal turn referenced in the WSJ article is worth walking through.

The idea that racism didn’t only live in the hearts and minds of individuals has a long history in academic and activist circles. In other words, the Supreme Court’s 1971 decision in Griggs v. Duke Power Co. (mentioned in the WSJ article) didn’t introduce the idea. It did, however, make two different things violations of Title VII: discrimination with intent, and disparate outcomes regardless of intent.

In my view, the Court’s decision was understandable given the context; Duke Power Co. had a history of restricting opportunities for black employees. And when its facially neutral tests arrived the day Title VII took effect in July 1965, the Court knew that proving discriminatory intent — until then what nearly all discrimination claims required — under such policies would be close to impossible. The route around that problem was to make disparate impact itself sufficient for a violation of Title VII. Chief Justice Burger was famously careful to say in the Griggs decision that good intent doesn’t redeem a practice operating as a “built-in headwind.” In fact, that carefulness may be part of why the causal question no longer needed to be asked. Once good intent is no defense, there’s no longer any reason for a court to determine what produced the disparity. The question of causality isn’t ruled out. It simply has nothing to do.

As I’ve documented elsewhere in detail, the legal equation of the two kinds of practices moved outward from employment law: into scholarship on systemic racism, into the interpretation of psychological instruments, into education, into HR, into the ordinary vocabulary of American institutional life.

The problem isn’t that the claim that disparities equal discrimination is false. I don’t know for sure that it’s wrong, and in some cases, it may well be all or partly true. The problem is that an interpretive claim got treated as a descriptive and factual one — and once that happens, disagreement with it stops registering as disagreement. It registers as denial or worse. Because the cause being named is a moral wrong, someone who disputes the causal claim isn’t heard as disputing a causal claim. They’re heard as making excuses for racism.

How do I know this premise was embedded as settled?

I know from the shadow it casts. Take colorblindness as an example. The principle that race shouldn’t determine how a person is treated was, for most of the twentieth century, the anti-racist position. It’s what the civil rights movement argued for. Today, in many left-leaning circles (including sociology), advocating for it is itself understood as a form of racism — as a mechanism by which racial hierarchy is maintained.

There’s only one way to get from one version of colorblindness to the other: that is with the claim that it is upholding racism itself. Once disparities are evidence of discrimination, then declining to adopt race-conscious policies means declining to address discrimination — and the principle stops being a competing moral vision and becomes a way of protecting the status quo. Grant the underlying premise and the conversion is automatic. Withhold it and a person committed to equal treatment is committed to equal treatment. The premise is what the idea of colorblind racism stands on, even though it never appears in the argument.

And colorblindness is only a clean case; it’s not unique. The same conversion changes a whole set of commitments. Equal treatment under law becomes an approach that entrenches advantage. Merit becomes a construct built to launder existing hierarchy. Appeals to individual responsibility become victim-blaming. And so on.

Run that kind of reclassification across enough commitments and you no longer have people who hold different views. You have people who are immoral. Within that framework, calling the conservative half of the country racist isn’t an insult. It’s an entailment.

Why did it happen?

Probably for multiple reasons. The premise that inequality is fundamentally unjust fit the political sensibilities of people positioned to adopt it. Not requiring proof of intent gave teeth to the new Civil Rights Act. Blaming systems is a more comfortable explanation for persistent racial inequality than considering factors related to individual or community level choices. Regardless of which reason was most salient, I doubt anyone experienced the moment as adopting a contested premise at all. The adoption of its moral entailments would have simply seemed like progress.

Going through the room

Consider how understanding this premise recasts the quotes of every scholar in the WSJ piece.

Kevin McCaffree says sociology misrepresents, conceals, or distorts the evidence. That’s a charge about accuracy and, in “distorts,” about intent. If I’m right, even if every sociologist in America were to report findings scrupulously, the problem would remain untouched, because the premise operates before any evidence is gathered. You can’t distort the answer to a question nobody asked. And you don’t need to distort the answer to a question you’re treating as a matter of fact. It comes out right on its own.

Christian Smith describes a sacred project, a quasi-religious impulse. It’s a claim about motive again, in a more sweeping form. But watch the remedy: Smith wonders aloud about liquidating the discipline. If sociology’s culture is really the problem, dissolving sociology should address it. And yet, we could close every department in the country tomorrow and the shadow of the Griggs-related shift would still stand. The EEOC would still operate on disparate impact, the compliance frameworks still run, curriculums still teach its moral entailments. The premise doesn’t live in sociology. Sociology is simply one of the places it landed.

Musa al-Gharbi says the discipline undermines what he calls “normie” beliefs — that religion is good, that marriage is stabilizing, that police are necessary. He’s right that it does. But the question that matters is why.

The premise supplies the answer. What those three have in common is that they’re inherited — long-standing arrangements that shaped the distribution we now observe. If that distribution is evidence of injustice, then the institutions that produced and sustain it are implicated in producing and sustaining injustice. Nothing further needs to be established about any of them. Marriage stops being a stabilizing arrangement and becomes a structure that encodes a hierarchy. Policing stops being a public function and becomes an instrument for maintaining one. Religion stops being a source of meaning and becomes a system of authority that legitimated the whole arrangement. Restoring those beliefs to good standing one at a time won’t work and isn’t really the point. They were convicted by the premise, and they’ll be reconvicted as long as it holds.

Saida Grundy says she believes in sociology as a weapon. She’s reasoning logically from the premise as described. If disparate outcomes are discrimination, then research documenting them documents an injustice, and advocacy isn’t politics contaminating knowledge — it’s the applied form of it. In this view, calling her an activist misdescribes what she’s doing: she isn’t smuggling a value into science, she’s drawing the entailed conclusion from a causal claim no one has asked her to defend, because no one is asked to defend it.

Lisa Hajjar says there’s no single Archimedean point of truth, only better and worse ways of presenting information. This deserves specific attention because, at a quick glance, it sounds like it should be on my side. It’s not.

At one level, my claim depends on a distinction between descriptive claims and interpretive ones. Dissolve that distinction — declare that every account is positioned, that description is always already interpretation — and my objection loses its grip. I want to say that a particular claim is interpretive and has been treated as descriptive. If all claims are interpretive, that’s true of every claim in the discipline, and pointing it out at any particular one carries no information.

The deeper problem with Hajjar’s point is that the no-Archimedean-point view isn’t actually held. I know of no one who actually treats all accounts as equally positioned. To focus again on sociology for a moment, in the same discipline that says there’s no Archimedean point, the account offered by a disadvantaged group is routinely treated as closer to the truth than the account offered by an advantaged one.

If every account is positioned, the disadvantaged account is positioned too, and there’s no ground from which to rank it higher. If it does have better access to the truth, then something confers that access — presumably a claim about how social position shapes perception, and about who is in a position to see the arrangement plainly. That’s a substantive causal claim about the world. It’s the kind of thing that needs an argument, and it’s exactly the kind of thing the no-Archimedean-point view says can’t be established.

Universal skepticism about objectivity turns out to leave exactly one interpretation standing. That’s not a theory of knowledge. It’s a way of protecting a claim from having to be argued for.

Melissa Wilde argues the ASA should stay politically neutral as a body. I don’t oppose that in theory. But an organization can be scrupulous in its official conduct while every claim it makes is derived from a premise no one is asked to argue for.

Besides, there’s a harder problem underneath. Neutrality means declining to take sides on contested questions — which requires knowing and acknowledging which questions are contested. An institution operating on a premise it no longer recognizes as a premise at all has lost the ability to make that determination. Once it quietly files the premise under settled fact, everything downstream of it will look like description rather than position. It can then be perfectly sincere in believing itself neutral while advancing a contested causal claim in every document it produces, because the claim doesn’t register as the kind of thing neutrality is about.

That’s not a failure of will. It’s a failure of sight. You can’t decline to take a side on a question you’ve stopped seeing as a question.

Jukka Savolainen probably comes closest to naming what I’m describing. He asks whether we want to learn the causes of crime or just repeat that policing is bad. That’s the right shape — a complaint about foreclosed causal inquiry. But it’s framed as a problem of research priorities and wasted resources, as though the fix were choosing better topics. The obstacle isn’t topic selection. It’s that the answer is fixed in advance, so studying the topic differently yields the same conclusion.

Finally, Ashley Rubin says activist logics are displacing scientific ones. Note what the word “logic” implies: something a person adopts and could therefore put down. On that description the fix is to swap the activist logic back out for the scientific one and the field might recover. But a settled premise isn’t a logic anyone adopted. It’s what the logic runs on. Apply the most scrupulous scientific reasoning you like to a proposition you’ve filed under settled fact, and it will carry that proposition straight through to the conclusion, more efficiently than sloppy reasoning would. Rigor doesn’t correct an unexamined premise. It transmits it.

One last item from the article before we move on. It offers as its most damning statistic that roughly 80 percent of articles in flagship sociology journals touch on inequality. It’s easy to read that as obsession — too much attention to one subject. But it’s the wrong reading. A discipline could devote every page it publishes to inequality and be in fine shape, provided the causal question stayed open inside those pages. The number isn’t a measure of how much sociologists study gaps. It’s a measure of how widely one reading of them has been applied.

Why the diagnosis matters

There are multiple reasons to care about getting the diagnosis right. (I go into those reasons in my spring 2027 book “Presumption of Guilt: How Equating Inequality with Injustice Fractured American Democracy.”) But one is that the diagnosis determines the remedy, and every remedy the article contemplates leaves the premise standing.

Depoliticize the discipline: the premise isn’t experienced as politics. Restore official neutrality: neutrality doesn’t reach what members treat as settled. Add viewpoint diversity: a premise isn’t a viewpoint; it’s the floor people stand on to have viewpoints. Teach better, publish more rigorously, recruit differently. Liquidate the whole thing. Each addresses how sociologists disagree. None addresses what has been placed outside the range of disagreement or how.

That pattern isn’t specific to sociology. Civility work, dialogue programs, civics education, depolarization — each is built to improve the conduct of a disagreement. None is built to locate a claim that stopped being visible as a claim. When one of them reaches such a claim, it’s because the right dissenter happened to be in the room, not because the tool went looking.

Named but not heard

I didn’t arrive at any of this by asking what’s wrong with sociology.

I was trying to understand the moral verdict — why disagreement on certain questions produces a judgment about the person rather than an argument. Work backward from that and you land on a claim being treated as settled, because nothing else makes the verdict intelligible. Then it becomes a question of which settled claim generates this particular set of moral commitments.

In the WSJ piece, I named the settled premise. That turns out not to be enough. Surrounded by others describing a fight about people, a sentence about a contested causal claim reads as one more complaint about politicized sociologists — which is the thing I said, converted into the thing everyone else was saying.

I suppose that is its own small demonstration. A claim doesn’t become visible by being stated. It has to be heard as the kind of claim it is — and by the time a reader reaches my three sentences, the article around them has already established what kind of story this is. The sorting happens first. The sentence arrives into a slot that was built before it got there.

Subscribe now

Appendix: A final objection

A note for readers who work in this area. Since the article ran, the sociologist Chris Maggio has pressed me on a point that deserves a fuller answer than a footnote.

Maggio took issue with one sentence in particular: “Even suggesting research into alternative sociological explanations is viewed as racist, morally wrong and victim blaming at best.” He argues this leaves readers with the impression that sociologists don’t care about identification strategies. And he points to the literature to counter that: audit and correspondence studies, including a recent meta-analysis synthesizing 114 field experiments. Those experiments exist to ask whether discrimination is present and are built to isolate it as a cause.

On one level Maggio’s objection is methodological, and on that level he’s simply right. Sociologists do sometimes build careful identification strategies, and anyone reading that sentence and concluding otherwise has been misled.

But there’s a harder objection underneath, and it’s the one worth answering. It goes like this: a field that builds experiments to find something out is a field that didn’t already know the answer. So, the existence of this literature isn’t just evidence of methodological seriousness — it’s evidence that the causal question I’m saying is closed was open.

That puts the challenge to me directly. If the premise is embedded, if the discipline treats disparities as discrimination without argument, then why run these studies at all? What is there to find out? Either the premise isn’t as settled as I say, or something else explains the research.

Here’s what I think is going on.

The settled premise says discrimination is the root cause of the inequality. What it doesn’t supply is an account of how — it has to be some process operating widely enough to produce differences at the level of a population. Multiple candidates have attempted to fill that role, including systemic/structural racism, implicit bias, and stereotype threat; each faces serious limitations.

Structural racism refers to disadvantage produced without anyone intending it — outcomes arising from how systems interact rather than from any actor’s purpose. That’s what distinguishes it from ordinary discrimination. The evidence usually offered is historical: redlining, restrictive covenants, exclusions from postwar housing programs. But those were explicit and intentional. The FHA underwriting manual said what it was doing, which makes it documented policy discrimination. The presence of structural racism is generally established by pointing at an unequal outcome and saying: this is what it looks like.

Implicit bias and stereotype threat share a slightly different problem: each was demonstrated at one scale and applied at another. The IAT shows a response-time difference in a sorting task; it’s used to help explain hiring and promotion outcomes across an economy. Stereotype threat was established experimentally — randomized manipulations of how a test is framed, with real effects on measured performance; it’s invoked to help explain achievement gaps that accumulate over years of schooling.

Different as they are, the three have something in common. Each is asked to explain more than it was ever shown to do.

The relationship among them is worth pausing on, because the literature doesn’t settle it. Sometimes structural racism is the umbrella and implicit bias and stereotype threat are mechanisms operating under it. Sometimes structural racism names something distinct from any individual-level process — that’s what “without anyone intending it” is usually meant to convey. Sometimes the terms appear side by side with no stated relation at all.

The looseness in definition has a consequence. Whatever the arrangement, the explanation tends to be invoked as cumulative — as far as I know, nobody claims implicit bias alone produces the wage gap. And when an explanation is cumulative and its components aren’t individually apportioned, each one’s shortfall can be absorbed by the others. Ask how much implicit bias accounts for and the answer gestures at structural forces. Ask what structural forces consist of and the answer gestures back. Nothing has to carry the weight alone, which means no component has to be assigned a share.

Something has to carry each explanation from where it was shown to where it’s applied, and what carries them is the settled premise. The gap is the reason to believe the mechanism operates at scale, and the mechanism is the reason the gap counts as discrimination.

Audit and correspondence studies share the scale problem — a callback difference doesn’t tell you what portion of a wage gap it accounts for — without sharing the circularity. But look at what gets demonstrated. Employers respond differently to names, at the callback stage, in the markets sampled. That’s the finding. Then it appears in explanations of labor market gaps, which requires two further claims the study never tested.

The first is about magnitude, and it’s the same shortfall as before — the difference being that here you can see exactly where it opens. The study measured whether the mechanism operates, not how much inequality it produces. A thirty percent callback penalty is compatible with discrimination accounting for five percent of the wage gap or for all of it.

The second is about kind. A callback difference is consistent with an employer acting on aversion — taste-based discrimination. It’s equally consistent with an employer using the name as a proxy for something he can’t observe and would rather know — statistical discrimination. Same result, different mechanisms, different remedies, and the design doesn’t distinguish them. What travels downstream is aversion.

Nothing bridges the raw finding to either claim except the settled premise. If disparities are discrimination, and discrimination carries the moral weight of racism, then the mechanism has to be animus and it has to be doing most of the work — because the alternatives lead somewhere the premise forecloses. Statistical discrimination implies the proxy might track something real, and asking what it tracks reopens the causal question. And if discrimination accounts for only a modest share of the gap, something else accounts for the rest, and that something needs a name nobody can offer without cost.

So the studies aren’t evidence that the causal question is open. They’re evidence that the field located the one setting where a mechanism could be demonstrated cleanly and demonstrated it. What studies in that literature don’t generally establish is the step from a callback difference to the gaps everyone is actually arguing about.

Now, it would be fair to ask here what the alternative would look like, so let me try to state it.

An open question would show up in what gets asked after a callback gap is found. Suppose the gap comes back at thirty percent — the number I threw out above. If the causal question were live, that result would generate its own next questions: A callback penalty of this size — what share of the observed wage gap could it plausibly produce? What’s in the rest, and how would we get at it? Those are the kinds of questions a finding raises when you don’t already know what it means. It would also mean asking: is this aversion or is the employer using the name as a proxy for something he can’t observe?

I’ll note here that there are designs that try to answer that last question. The main one turns on a prediction: an employer using a proxy does so because he lacks direct information, so give him more and he should need the proxy less. Build résumés at two levels of detail — thin ones with a name and a degree, thick ones with a specific work history, quantified accomplishments, references. If the gap shrinks as résumés get richer, that looks like an information problem resolving. If it holds steady, that looks like aversion.

The design is good — where it falls short, it falls short on identification, and the researchers know its limits better than I do. The trouble is what a result means once you have it. If the callback gap shrinks, that’s consistent with statistical discrimination; it’s also consistent with an employer whose aversion is simply overridden by a strong enough candidate. If the gap holds, that’s consistent with aversion. But it’s also consistent with an employer who doesn’t take the added information at face value from both applicants — which could itself be the result of statistical discrimination.

Suppose both thick résumés list a supervisory role and a strong reference. The employer reads one and thinks: solid record. He reads the other and wonders whether the title was inflated, whether the reference is a favor. He isn’t refusing to hire the second applicant because of the name. He’s discounting the evidence because of it — and if he discounts it, the extra information never reached him. He’s still working from the proxy, because as far as he’s concerned nothing reliable was added. Notice what that employer is doing. He’s using the name as a proxy for how much to trust the applicant’s own account of himself — statistical discrimination applied one level up, to the evidence rather than to the person. Which means a stable gap doesn’t even establish aversion. It’s consistent with a world in which no employer holds any animus and every one of them is discounting paperwork.

Let me be clear. I’m not saying the latter isn’t a problem. I’m saying it’s a different problem than taste-based racism, and the difference determines what would fix it. The applicant is harmed either way, which is probably part of what makes the conflation tempting: the outcome looks the same from where he’s standing. But treating the second employer as the first means addressing him with a remedy built for someone else — and being told you hold a prejudice you don’t hold is its own kind of provocation, one that tends to produce resistance rather than change.

None of this puts Maggio’s concern and my claim in tension. He can point to 114 studies of genuine causal identification, and I can grant every one, because the premise isn’t operating in place of the identification work. It’s operating on top of it, on the questions the work leaves open — and drawing its authority from the very rigor that stopped short of them.

Subscribe now

Bridge-building, the steady call to be more civil — often formalized as civil discourse — and viewpoint diversity are some of the tools concerned citizens reach for when the conversation on heated issues breaks down. Each is valuable — and each locates the trouble somewhere other than where it actually sits.

The first two treat the challenges that sit adjacent to the disagreement — in the heat surrounding it, in the lack of empathy, or in the absence of sufficient good will. Fix the conditions, they seem to assume, and the disagreement will become tractable.

Viewpoint diversity implies that if we get a wider range of positions into the room, the dominant view will finally meet real opposition. But adding a dissenter changes who is present; it doesn’t, by itself, change what everyone treats as beyond question. That shift happens only if the particular person you’ve added sees the unspoken premise for what it is and knows to surface it — rather than simply arguing the unpopular side. And that’s a function of who the dissenter happens to be, not of the diversification itself.

Our disagreements about topics related to race, gender, harm, and intent are not, at their core, relational or distributional; they are epistemic.

Regular readers of my work will find this diagnosis familiar. What I hope to do here is show what it looks like to catch that mechanism in the act — to take a moralized verdict and follow it down to the claim buried underneath. In this case, the chase began at a discussion table and finished afterward, along the line the conversation might have taken had it kept going.

The occasion was a conference.

Participants were seated at tables; each was tasked with a controversial topic. For instance, one table was discussing the death penalty, another was talking about the use of AI in teaching, and still another was tackling the Israel and Gaza conflict. The conversation at my table began on the topic of privilege. Specifically, on whether efforts to “bridge-build” are helped or hurt by acknowledging who holds privilege in a given interaction. From there the discussion moved quickly to pronouns. In some ways, the shift made sense: gender identity is seen as one of the privilege axes, and honoring someone’s stated pronouns is treated, in this frame, as a way of recognizing — and deferring to — the difference in standing between you.

As we talked through the question of using preferred pronouns, two colleagues at the table, I’ll call them A and B, said that refusing to use a person’s preferred pronouns is unacceptable. Not unwise, not impolite — unacceptable, in a way that puts the pronoun refuser outside the range of people you can reasonably talk to.

One of the conference’s skilled civil discourse facilitators was at our table, and he pushed back — gently, and pointedly, at A, whom he was sitting beside. Did it give him pause, he asked A, that much of the world hasn’t moved toward gender-neutral language at all? As supporting evidence, he pointed out that a great deal of the Global South doesn’t organize social life around preferred pronouns or treat them as owed.

It was the kind of question meant to loosen A’s firm conviction; it didn’t. A didn’t miss a beat: Yes, that’s true, he said, because those countries are living with outdated gender norms.

***

I was present for the whole exchange. The facilitator was empathetic, curious, and appeared to be genuinely trying to understand. A appeared to listen attentively and answer with sincerity. Nevertheless, A ruled the other side out of bounds, leaving nowhere else for the conversation to go.

This kind of dead-end matters for multiple reasons, including what we often take away from it. When warmth is present and the disagreement still doesn’t move, we tend not to suspect the tool — rather we suspect that someone failed to listen well enough or didn’t come in with real sincerity. The failure sends us back for more civility, more empathy, more patience — which can only produce the same result, since none of it touches what actually closed the conversation in the first place.

What closed this conversation was a claim one side had stopped seeing as a claim — and reaching it takes a different move entirely: taking the thing that feels like bedrock — in this case, refusing to use a person’s preferred pronouns is unacceptable — and asking what it stands on.

While a moral verdict like this may feel difficult to argue with — questioning one can feel like siding against the value it defends — it isn’t free-floating. Moral verdicts track deeper convictions, even if the person delivering the specific verdict can’t tell you what that is.

So, I tried an experiment. I presented A and B with three hypothetical scenarios and asked whether they’d remain in each conversation.

Scenario 1: Someone says: I don’t think gender is only a social construct, but I’ll use your pronouns anyway.
Response: Both A & B said they’d stay in the conversation.

Scenario 2: Someone says: I don’t believe gender is a social construct, and I won’t use your pronouns.
Response: Both A & B said they would leave that conversation.

Scenario 3: Someone says: I don’t believe gender is a social construct, and I won’t use your pronouns — because I think social identity, pronoun use included, is a negotiation, partly how I see you and partly how you see me.
Response: A said he’d leave. B said she’d stay.

Scenarios 2 and 3 contain the same belief about gender and the same refusal to comply. The only difference between them is the account attached to the refusal — and it split my two table mates apart. Whatever conviction was doing the work for B relaxed when a rival account of identity came into view. Whatever was doing the work for A didn’t.

As it happened, the conversation moved on before I could press further, but here is where I would have taken it given the chance.

A’s position was that refusing to use someone’s preferred pronouns is a failure of respect, full stop. And to A (based on his response to Scenario 3), no rival theory of what identity is had any bearing on it. But that only showed that the settled claim his position rested on sat somewhere I hadn’t yet reached.

Locating A’s settled claim requires pausing on his use of the word respect. In calling pronoun use “just a matter of respect,” A presents it as a small ask — a basic courtesy. So, let’s say that using someone’s preferred pronouns is like calling them by a nickname they prefer — a small kindness most people extend without a second thought.

If pronoun refusal is merely a discourtesy — on the order of ignoring a preferred nickname — then walking away, which A said he would do, becomes a wild overreaction. A breach of courtesy generally earns mild disapproval; you think a little less of the person and carry on. It doesn’t usually make you end the conversation. This logic suggests that courtesy alone can’t be the real ground. It is the name given to something heavier — and that something is what we have to find.

We can start with a question: Does a commitment to treating people with respect oblige us to honor every request about how they wish to be addressed? To take an admittedly absurd example, what if a stranger we know not to be royalty asks to be called “Your Excellency,” a title we know he hasn’t earned? I suspect that most people wouldn’t feel they’d wronged him by declining. My point here isn’t to liken the feelings associated with a person’s gender to those associated with a vanity title — my point is structural. That is, a commitment to treating others with respect, on its own, doesn’t oblige us to ratify every term a person asks us to apply to them. The “Your Excellency” case is just the cleanest way to show the line is real.

Once we see that a line exists — that respect requires honoring some requests about how a person is addressed but not others — A’s position faces a problem. He needs pronoun use on the mandatory side of that line, where refusing is a moral wrong. But for a request to land there, it has to track something understood to be a fact about the person — something real about who they are, not a preference they’ve declared.

So, what kind of claim would accomplish that? Whatever it is, it has to clear two hurdles. First, it can’t make gender a matter of choice. If gender were something a person simply elected — declared into being — it would land right back beside the vanity title, on the optional side of the line, and refusing it would be a discourtesy at most. So the fact has to be involuntary: something true of the person whether they willed it or not. But — second — it also can’t be a matter of the body. Were gender settled by biology, then the body, not the person, would have the final say; a person could even be mistaken about their own gender, the way anyone can be wrong about a fact of nature. That would strip authority from the person’s own sense of themselves, which is precisely the authority A’s position exists to protect.

Put those two requirements together and only one kind of thing satisfies both: an inner fact. Gender, on this account, is a stable, involuntary sense of oneself — not chosen, and not reducible to the body — and it is this inner sense that pronouns track, and that others are bound to recognize. That is what set A apart from B in the third scenario: the negotiated account was one B could accept and A could not, because A’s position requires an identity authored solely from within.

We knew from the outset that some settled claim was supporting A’s verdict. What we didn’t know was what that claim was, or where it sat. And it’s worth being clear about what kind of thing we were after: not a report of what A consciously believes, but what his verdict requires in order to stand. A position rests on whatever must be true for it to hold, not only on what the person defending it has put into words — and in the hardest cases, the load-bearing claim is precisely the one that never gets said, which is why it never gets examined. The experiment told us it wasn’t in A’s account of identity, since his verdict held firm where B’s gave way. Tracing his appeal to respect told us the rest.

Here is what changes once the claim is named. A moral verdict — “unacceptable” — closes a conversation: it isn’t the kind of thing you argue with, only submit to or offend against. But the claim underneath it is a different kind of thing. That there is an inner fact of gender, that it’s involuntary, that it owes nothing to the body, and that pronouns track it — every part of that can be disputed by reasonable, informed people. And a claim that reasonable people dispute is, by definition, arguable: something you can hold up, examine, agree or disagree with, ask for reasons about. My point isn’t to declare A wrong, but to show that what he rested on as settled is in fact in dispute and so belongs back in the conversation rather than ending it.

***

The logic we’ve unpacked here isn’t unique to one person or issue. When a question is genuinely contested — when reasonable, informed people land on different sides of it — then any verdict that puts one side beyond question is treating a contested claim as settled. That’s simply what it means to foreclose a live question: you have to promote one contested answer to the status of final. So on any contested issue, a verdict delivered as beyond dispute is a tell that a claim is buried beneath it — one the person delivering the verdict has stopped seeing as a claim at all.

That brings me back to where I started. None of the tools we reach for when these conversations break down is built to do what we just did. Bridge-building works to close the distance between people. Civility works on the tone of the exchange. Viewpoint diversity works on who’s in the room. When one of these tools reaches that buried claim that’s driving the closure, it’s because the right dissenter happened to be present, or the premise happened to surface on its own, not because the tool went looking. And yet, that looking is precisely what’s required. You have to take the moral verdict that feels like bedrock and ask what it’s standing on. Do that, and what looked like an unbridgeable moral rift turns back into a disagreement two reasonable people can actually have.

Subscribe now

As we go into the holiday weekend, I’m happy to share that my next book, “Presumption of Guilt: How Equating Inequality with Injustice Fractured American Democracy,” will be published by Pitchstone Books in spring 2027.

With that in mind, here’s a sneak peak at the (unedited) preface.

In the second half of the twentieth century, American institutions violated a constraint that pluralistic democracies cannot afford to ignore. They treated the outcome of a democratic process as settling the questions that motivated it. This recast legitimate disagreement as ignorance or moral failure—with consequences that are still unfolding.

To be sure, it can seem like what I’m calling here a “violation” happens all the time. After all, haven’t we used institutions to prohibit things like stealing—and doesn’t that prohibition reflect the determination that stealing is wrong? If democratic outcomes routinely settle the questions that motivate them, what makes this late twentieth century example any different? But this reflects a common point of confusion.

Democratic institutions don’t need to make any determination about whether stealing is wrong in order to prohibit it. They only need to determine that a failure to prohibit theft makes it difficult for people to live together without constant or near-constant conflict. The fact that there happens to be widespread consensus on the morality of prohibiting theft helps obscure the practical determination that justifies it. This can be easier to see in an example where the moral consensus more clearly breaks down.

Take drug policy. The American people do not all agree, for instance, on the moral question of whether individuals have a right to make choices about their own bodies and consciousness, even when those choices carry personal or societal risk. And yet, we have drug laws. Not only that, but people who oppose current drug policies aren’t automatically revealed as holding indefensible values. Grounding collective decisions in practical rather than moral premises lets us make collective decisions and design shared policies without labeling one side righteous and the other side immoral—or worse.

When it comes to the 20^th century case referenced above, at first—and in its narrowest form—the settled-upon answer commanded enough consensus to serve as a workable basis for policy. In its expanded form, it was absorbed almost entirely by the left. That claim—that inequality indicates discrimination—is the contested claim this book examines. Contested not because the concern behind it is unwarranted, but because it is not established fact; it is an interpretation. The problem this book traces is what happened when institutions treated it as something more than that.

As we’ll see, equation of the two naturally leads to the idea that “inequality is inherently unjust.” Where, in practice, a more specific formulation might be: “inequality is inherently unjust when it reveals that a nonwhite or otherwise nondominant group is disadvantaged.” Notably, the difference between the first and second phrasings only matters if the group seen as nonwhite or otherwise structurally powerless is advantaged and the group seen as white or otherwise structurally powerful is disadvantaged.

The very idea that the institutional settling of this question could be the driving force behind much of our political breakdown will almost certainly make some readers bristle. Given that, it probably makes sense to anticipate a few concerns up front.

On partisanship

To some readers, the fact that my analysis of the problem means focusing more on the left than on the right will make this book seem partisan in nature. Avoiding such a perception would require a sort of bothsidesism that I do not attempt. There are multiple reasons for this deliberate omission.

One is that treating problems on the left and right as equivalent but separate phenomena risks obscuring how they’re connected. It evokes a sort of scorecard mentality—one where we’re counting tick marks on each side to see who is worse—when what really matters is understanding how violations on one side produce reactions on the other. The left’s moral delegitimization of conservative positions didn’t happen in a vacuum, nor did the right’s norm-breaking response. These dynamics reinforce each other and understanding either side in isolation misses the causal sequence that produced the modern predicament.

Another reason is that the damage to democracy this has caused can’t be undone from the outside. It has to be transformed from within. That means that, in order for the country to move forward, the left has to understand what happened—that what’s been treated as settled isn’t in fact so—and there has to be a conscious decision to change.

I write this as a lifelong Democrat who has spent over twenty years in higher education—in sociology no less, one of the most left-leaning disciplines there is. That said, I don’t claim to know the right answers to any of the urgent policy questions we face as a country. I don’t know whether immigration should be more or less restricted, whether healthcare should be government-provided or market-based, or what the optimal balance is between equality and other values.

What I do know is that democracy requires seeing and treating people who reach different conclusions on these and other questions as reasoning legitimately rather than as morally deficient or ignorant. My goal is to defend the process of democratic disagreement itself—the ability to hold and articulate different views without being morally tarred. And doing so requires a clear-eyed diagnosis of its threats, regardless of who that implicates.

On singular causes, oversimplification, and obsession

Some readers will feel like my argument is an oversimplification—as in, how could so many problems be traceable to a single shift?—and they will be tempted to dismiss it as such. To those readers, I say, in some ways you’re right. After all, that’s what explanations do; they reduce complexity to comprehensible patterns. In that sense, the question isn’t whether I’ve simplified—of course I have—it’s: Have I simplified in a way that preserves important causal relationships? Generates accurate predictions? Explains things other theories can’t? Points toward a solution that might actually work?

If I’ve truly oversimplified, my explanation should make bad predictions. So, as you read, I invite you to consider whether this is the case. Or rather, does it predict: the success of Trump or someone like him? The collapse of institutional trust? The culture wars? Why we seem to have two sets of facts and two versions of reality?

I’m not claiming that the problem I describe is the only thing that matters on the modern political stage. I’m saying it helped determine which conflicts became morally loaded, how they aligned with partisan identity, and which positions got delegitimized versus treated as reasonable disagreement. The shift described here didn’t create every political conflict, but it shaped how those conflicts got framed, which side of them became “progressive” versus “bigoted,” and why they became culture war flashpoints rather than routine policy debates.

Despite my attempt here to justify my decision, there will be readers who still think I’m irrationally obsessed. To them, I would pose the following question: as you read, ask yourself whether the mechanism I’m identifying is real. Then consider whether it does the explanatory work I claim it does. If it does, then the relevant question isn’t whether I’m obsessed. It’s why this isn’t getting more attention. In other words, the test is whether it works. I’ll let the explanations and predictions speak for themselves.

Besides, in the end, maybe I am “obsessed” with this one thing. But that one thing shapes questions of how we determine what’s true; how we understand justice; how institutions function; how we talk to each other; whether democracy works; and whether we have a shared reality. So, I guess if I’m going to be obsessed with anything, a change that impacts all of these seems like a reasonable candidate.

On being wrong

One of the biggest concerns I expect from readers on the left sounds something like this: “I’ll grant that you’re describing something real. But whatever the left has done pales in comparison to what’s happening on the right. We’re talking about a difference of at least an order of magnitude. So, while you may be right that we need to put out the little fire you’re focused on, I can’t even think about problem that until the five-alarm inferno that conservatives have created is extinguished or at least brought under control.”

To this concern, I’d say the following: First, I’m not sure how useful arguments about which side is worse are. Second, and more importantly, when you try to extinguish what you see as a five-alarm fire on the right without addressing the problem I’m describing in this book, you confirm for those who lit it that they were justified in doing so.

With all this said, the argument should be held to a standard. Here are a few things that would genuinely challenge it. Show me that the change I describe didn’t happen the way it did. Show me that the mechanism I identify doesn’t actually operate the way I describe. Show me major phenomena related to social trust and democratic legitimacy that this shift and its consequences can’t explain. Show me predictions it makes that fail. Or show me a competing explanation that accounts for the same range of outcomes with equal parsimony and fewer assumptions. If there’s a better explanation for how we got here, I genuinely want to hear it.

If all you’ve got is that half the country is racist, I say, you might be right. It’s not like I can see inside people’s minds or hearts to know one way or another. Although, if that’s really the case, democracy was doomed from the start.

Who’s this book for?

This book is a wake-up call for progressives who feel convinced that the biggest problems we face stem from rampant bigotry, willful ignorance, and callousness on the right. This includes people who see opposition to affirmative action as racist, the desire for immigration restrictions as xenophobic, and the insistence on two genders as backwards and regressive. Many of the people holding these beliefs see themselves as defending vulnerable groups and democratic values. To them, I would say that preserving democracy requires questioning the very claim you’re using to identify bigotry itself. Ultimately, you can’t both fight injustice—at least as currently conceptualized—and maintain the foundations democracy needs.

This book is also for progressives who are aware that calling people who disagree bigots is a poor strategy, but who can’t quite put their finger on how we got here or what to do about it. It’s also for non-progressive and “anti-woke” folks who understand that you can’t rebuild and destroy democracy at the same time.

But, perhaps more than anything else, this book is for anyone who wants to push for the kind of change that can lead to democratic renewal.

Besides, if none of the groups I just described read this, perhaps it can serve as a post-mortem for a curious political scientist years from now when democracy as we once knew it is nothing more than a quaint, but distant memory.

I’m ambitious. But with good reason. There’s a lot on the line.

Subscribe now

Every semester, biology professors across the country and around the world teach the theory of evolution. And no advocate for Intelligent Design — the idea that some features of living organisms are too complex to have arisen through undirected processes and must instead reflect the work of an intelligent cause — could force an instructor to give that view a place in the course. That much is broadly accepted. But the reason it holds is widely misunderstood, and the misunderstanding drives a great deal of today’s conflict over academic freedom — including the standoff between the universities and the Trump administration, where each side sees the other’s offense and not its own.

Why can the instructor refuse? One might think it is because evolution is beyond question — settled in a way that puts the alternative out of bounds. But that cannot be the reason, because evolution is not closed in that sense; no empirical claim is. Even if the entire scientific community accepts evolution as fact, consensus is not closure. A claim that everyone accepts is still a claim; the number of people who hold an inference does not convert it into a fact. If it did, truth would be determined by a head count, and it is not. The real reason is narrower: academic freedom gives him control over what he teaches. And controlling what he teaches is not the same as closing the question he teaches about.

Since its 1940 formulation, academic freedom has been grounded in what the American Association of University Professors called the common good’s dependence on “the free search for truth and its free exposition.” Its purpose is to protect scholars from being punished or controlled by power — fired, defunded, sanctioned — for the conclusions they reach in their research, teaching, or external utterances. It is a shield between the scholar and entities with power: deans, trustees, the legislature, the state. But notice what the principle is for. It protects the search. And there is no searching for what has already been found — no free inquiry into a question one has declared closed. The freedom exists to keep questions open, which is precisely why it cannot be the thing that closes them.

That means that while academic freedom protects the biology instructor from having to capitulate when it comes to course material, it does not license him to treat a student who believes in Intelligent Design as ignorant or wrong, nor to mark down that student’s grade — provided the student has otherwise engaged the material seriously. Put simply, the biology instructor cannot deny evolution’s inferential status and claim academic freedom as his shield.

Consider what would follow if this weren’t the case. The principle of academic freedom is content-neutral; it protects every scholar alike, whatever they conclude. Suppose — and this is admittedly far-fetched, since vanishingly few biologists hold any such view — an instructor who genuinely held that Intelligent Design were correct. If academic freedom gave the biology professor authority to declare evolution settled, it would, by the same logic, give the identical authority to an instructor who believes in Intelligent Design. This symmetry says nothing about whether evolution and Intelligent Design are equally likely to be true. It only assumes that both claims come with a non-zero amount of uncertainty. Because any uncertainty at all means the question is not closed, regardless of how even or uneven the evidence is.

Now that we understand the limits of academic freedom, consider a different and harder example: a sociology instructor who teaches that opposing affirmative action is racist. The equation of a contested policy position with bigotry is a familiar feature of academic and public life, and far closer than the biology case to the conflicts now reaching the news. Is this statement protected by academic freedom?

On the one hand, the idea that a verdict about people (in this case, that those who oppose affirmative action are racist) would be protected seems theoretically plausible. Surely scholars are permitted to reach moral conclusions. A historian may teach that chattel slavery was a moral catastrophe; an ethicist may teach that torture is wrong. Academic freedom would seem to rightly protect both.

What protects these conclusions is not that they are correct, but that they rest on values — and values are understood, by everyone, to vary from person to person. No one mistakes “torture is wrong” for a finding about the world. It is plainly a moral judgment, and a moral judgment does not present itself as settled; its grounding in values is right there on the surface.

This is why academic freedom would equally protect an instructor who took the opposite view — that torture is sometimes justified, even right. That some would find the conclusion monstrous changes nothing about his protected right to defend it. The protection holds for the same reason it holds for the instructor who condemns torture: in neither case is anyone pretending the matter is settled. Academic freedom covers the moral position, precisely because a moral position never claims to be closed.

At first glance, the claim that opposition to affirmative action is racist can sound like a moral verdict of the same kind as “torture is wrong.” It is not. A moral verdict evaluates — it says a thing is wrong, or unjust, and asks you to share the judgment. “Torture is wrong” makes no claim about the world that could be checked, and its grounding in values is on the surface. “Racist” also evaluates — but it goes one step further. The label asserts that the position in question does damage along racial lines.

That extra step — the factual assertion buried in the label — is where the trouble lives, and it is worth drawing out slowly. To call opposition to affirmative action racist is to issue a verdict of bigotry, and a verdict of bigotry is justified only if there is no non-racist explanation for opposing affirmative action — because you cannot be a bigot for landing on one side of a question where a reasonable person could land on either. On values grounds alone, the verdict has nothing to stand on. A person might oppose affirmative action out of the conviction that the fair way to allocate resources is to treat everyone the same, regardless of race. No verdict of bigotry can be built on a value that a non-bigot could weigh differently.

So if the charge is to survive at all, its ground must lie elsewhere. By looking at what the verdict already assumes, we can see that the load-bearing claim is there. It is the assumption that the disparities affirmative action addresses — the underrepresentation it exists to correct — are themselves the product of racism. If equal treatment leaves those disparities standing, and the disparities are racism’s doing, then treating everyone the same perpetuates racism — and supporting equal treatment means supporting that harm, regardless of intention or values.

With that claim in hand, opposing affirmative action becomes complicity in racism. Without it, equal treatment is simply a competing principle of fairness. But look at what the ground is. The whole bigotry verdict now rests on an empirical claim: that racial disparities are caused by racism. And, like any causal inference, the claim is laden with uncertainty and rival explanations. In short, the bigotry verdict can stand only by treating as settled a contested claim about why racial inequality exists, precisely what academic freedom does not protect.

It may be tempting to think the charge levied in classrooms is often softer than “racist” — that dissent is treated not as malice but as ignorance, a failure to understand how racism works. But this softer charge rests on the very same unprotected closure. To call the dissenter uninformed is to presuppose there is a settled answer he has failed to reach — which is exactly the move in question. To diagnose disagreement as ignorance is to predict that knowledge converges on the approved conclusion — that the only reason to dissent is missing information. Supply the information, and if the person still disagrees, the ignorance explanation is spent. Only one possibility remains: they understood and refused anyway. Ignorance is not the milder cousin of the bigotry charge. It is the on-ramp.

The solution is not to forbid teaching that inequality is the result of racism or that evolution is undirected by a higher power. Academic freedom protects the instructor’s ability to teach the inference, argue it, press it as hard as she likes. The remedy is only to teach it as what it is: an inference about the world, not a settled fact — and to mark that status rather than bury it. The instructor need not pretend to neutrality she does not feel or give equal time to every objection. She need only be honest about the kind of claim she is making. A causal claim about why inequality exists is an inference and saying so takes nothing from how strongly she holds it. What she may not do is present the inference as closed or treat the student who recognizes it as an inference as having erred.

This brings us to the present moment, when academics invoke academic freedom against the Trump administration’s pressure on universities. On the central point they are right. The administration is reaching into the classroom and telling instructors what they may and may not teach — which is precisely the coercion academic freedom exists to prevent. It is the opening example playing out: only instead of a student attempting to force Intelligent Design into the course, it is the state forcing its own content in, and contested content out. When the academy says this violates academic freedom, it is correct.

What the academy fails to see is that the coercion did not arrive in a vacuum. For years, the academy closed questions it had no authority to close and waved away — as ignorant or bigoted — the people who said so. Many of those people are now in power, and they have seized on the visible effects of that closure as their justification: the universities, they say, abandoned open inquiry long ago, so why should open inquiry shield them now? Whatever the motives behind it, the accusation is not false — and its truth is the academy’s own doing. By using the shield of academic freedom to justify treating contested questions as closed, the academy built the opening the state is now driving through.

Why does this matter, if the state is still in the wrong? It matters because of what follows for the academy that wants the intrusion to stop. The state’s pressure draws its force from a real grievance — that the universities closed questions they had no business closing. As long as that grievance stands, the coercion has a justification to point to, and resistance in the name of academic freedom rings hollow, because the academy is invoking a principle it spent years misusing. The way to take the justification away is not to deny the closure more loudly. It is to end it — and ending it has to begin with admitting it was there.

Academic freedom grants no one the authority to close a question — not the state reaching in from outside, and not the scholar quietly closing it from within. The academy is right to resist the first. It will only have the standing to do so when it stops doing the second. And that begins not with a louder defense, but with an admission: that the closure was real, that it was ours, and that academic freedom never authorized it in the first place.

Subscribe now

I’ve spent a fair amount of time arguing that some of the most popular cures for our political moment don’t reach what’s actually wrong. I’ve made the case about civility — that lowering the temperature does nothing if the people being civil still view disagreement as evidence of a moral or intellectual defect. I’ve made it about civil discourse initiatives — that teaching people to talk across difference assumes the problem is a communication breakdown, when the problem is that one set of views has been ruled out of bounds before the conversation starts. And I’ve made it about viewpoint diversity — adding conservatives to an institution which codes their positions as bigotry doesn’t diversify the institution’s thinking; it just gives it more people to diagnose.

Each of these is a remedy aimed at a surface — civility at tone, civil-discourse initiatives at process, viewpoint diversity at composition. None of them touches the thing festering underneath.

There’s a fourth remedy I haven’t taken on directly, and it’s one I regularly see proposed confidently. The idea seems to be that the answer to polarization, to democratic decay, to the whole grinding mess we’re in, is for more people to move toward the middle. The solution, in this line of thinking, seems to be to find the reasonable people on both sides, split the difference, and rebuild from there. The center gets treated as a kind of safe ground — the place the fever breaks.

I want to argue that there’s nothing magical about the center. And the reason it isn’t magical is the same reason the other three remedies miss. It’s pointed at the wrong thing entirely.

Moderation is not humility

The center looks like it operates at the level that matters — not tone, not process, not headcount, but substance. The centrist appears to be doing the real work: weighing the evidence, taking the strongest points from each side, landing on the merits. Set against the partisan who appears to take everything on faith, the centrist looks epistemically responsible. That appearance gives the center a degree of moral prestige.

But this rests on treating two different things as if they were one. The first is moderation — a property of your position. It describes where you land relative to the poles: somewhere in between. The second is epistemic humility — a property of how you hold a position, any position. It describes whether you treat your assumptions as contestable, as something a reasonable and informed person might reject.

Moderation and epistemic humility are not the same; they are barely even related. You can hold a moderate position with certainty — certainty that the truth obviously lies in the middle, that anyone out at the poles is a zealot or a fool, that your own balance is simply what clear sight produces. And you can hold a position far from the center with real humility — aware that you might be wrong, able to say why a thoughtful person would land elsewhere, treating the people who disagree as reasoning rather than defective.

Moderation is about where you stand. Epistemic humility is about how. They run on different axes. And it’s the second one — the how — that democratic health actually depends on.

The center has its own version of the move

I’ve described elsewhere (here, here, and here, to name a few) what I take to be a central source of contemporary political dysfunction: a contested position came to be treated as settled, so that disagreeing with it became evidence of a defect rather than a difference of view. Once that happens, dissent can’t be engaged on the merits — there’s no reasoning to meet, only a character to diagnose.

And yet, listen to a thoughtful centrist on the people at the poles — and I mean a genuinely thoughtful one. He doesn’t say they’re stupid or malicious. More likely, he's sympathetic: as he sees it, they've been captured by their information environment, sorted into echo chambers, carried along by currents larger than themselves. He might be quite generous about it. But notice what the generosity does. It explains the strong conviction at the poles rather than engaging it — treats the holding of an intense view as a condition to be accounted for, with a cause, rather than a position that might have reasons worth meeting. The sympathy can be real. It is also a way of not having to take the view seriously, because a view you’ve explained by your understanding of its origins is a view you no longer have to engage with.

So the centrist is not necessarily standing outside the dysfunction. He might be — but nothing about occupying the center puts him there. He can just as easily be committing the very thing that produced the current crisis — explaining the disagreement in his own terms rather than meeting it — while being, by every visible measure, fair, warm, and well-meaning.

None of this is to say the center has nothing to recommend it. Moderate positions, to the extent they favor incremental change, can be wiser when stability is what you're after — rapid change is often destabilizing, and sometimes that's a decisive cost. But that's a conditional advantage, not a standing one. Stability is itself a value, weighed differently by different people for different reasons. The case for incrementalism is strong only once you've granted the importance of the thing that depends on it.

A politics of the middle may also lower the stakes of being wrong, if moderate errors prove more recoverable than radical ones — which matters, of course, only to the extent you weight recoverability over getting it right faster. It may even be that people drawn to the center are, on average, somewhat likelier to hold their views loosely. But notice that this, if it’s true, is a fact about a correlation — not about the center itself. The humility would be the thing doing the work, and the humility can travel anywhere. The center doesn’t manufacture it; it just, sometimes, attracts it. What I’m denying is the stronger and more common claim: that moving toward the middle is itself the remedy. That the center is where the fever breaks.

Polarization was never the disease

Distance was never the disease. Consider what a healthy democratic disagreement actually looks like. Two camps can be far apart on an issue — genuinely, deeply opposed — and the polity is fine, so long as each side holds its position as contestable — as something a reasonable person could reject, held by people who are reasoning rather than defective. That’s not dysfunction. That’s democracy doing the thing democracy is for — managing disagreement among people who don’t have to resolve who’s morally correct in order to live together.

Now consider the opposite. A polity can be highly unpolarized — broadly agreed, consensus everywhere — and be epistemically dead. Everyone lands in the same place; dissent from the consensus is treated as a defect. There’s no distance to speak of. And it is far sicker than the polarized case, because the thing democracy requires — that contested questions remain contestable — has been extinguished. It just doesn’t look sick, because we’ve been trained to read agreement as health and distance as disease.

The variable that matters isn’t how far apart people are on some ever-shifting political spectrum. It’s whether they treat the assumptions on which their position rests as open to being challenged. And that variable runs perpendicular to the left-right axis. You can move someone all the way to the center and not change it at all — you’ll have relocated their certainty, not loosened it. The center is a coordinate on a line. The disease isn’t on the line.

Nothing magical

I get that there is something seductive about the center; I understand the pull. It can feel like wisdom. It can feel like the exhausted, reasonable alternative to two camps hellbent on screaming past each other. But wisdom isn’t a location. The reasonable person isn’t reasonable because of where he ends up. He’s reasonable because of how he holds what he believes — because he can say why someone might see it differently, and mean it.

That person might be a centrist. He might also be quite far from the center. The center has no monopoly on him, and no special claim to produce him. There is nothing magical about the middle of a line that was never the right thing to measure.

Subscribe now

The most commonly voiced worries about AI and democracy tend to go something like this: AI will flood the information environment with deepfakes, bot swarms, and micro-targeted manipulation. Voters won’t know what’s real. Trust will collapse. Democracy depends on a shared set of facts, and AI is dissolving that foundation.

Most of these worries are legitimate as far as they go. But they misidentify where the deepest damage happens—and therefore misidentify the solution.

The fabrication problem is real. A deepfake video of a candidate committing a crime is identity fraud. And yet, we have tools designed for that: things like provenance standards, watermarking, platform accountability, criminal liability. These tools are imperfect and lagging — but they’re at least aimed at the right problem.

The harder problem is the contestable premises that AI platforms embed in virtually everything they produce, while presenting interpreted output as neutral or factual.

When an AI summarizes a policy debate, it makes choices — based on whatever its training data treats as mainstream — about which framings count as reasonable and which count as fringe. Consider a concrete example: When Chat GPT answers a question about disparities in hiring or arrests or test scores, it makes choices about whether to treat “disparate outcomes indicate discrimination” as an established empirical finding or a contested causal claim. Similarly, when it describes a historical event, it makes choices about which moral categories apply.

This isn’t a bug in any particular AI platform—it’s a structural feature of how they work. They take in language, which is saturated with contested moral and causal premises, and produce output that inherits those same premises. This means the question isn’t whether AI systems will encode contested commitments as facts. They already do. It is whether we will teach it to make those commitments visible to the people using them.

Right now, they’re not. And that’s devastating for democracy.

Consider what happens when a citizen asks an AI system about affirmative action, or immigration, or policing, or gender. The system can produce fluent prose that sounds authoritative and balanced. And yet, embedded in that prose are answers to questions the user never asked and probably can’t see: Is group membership the right unit of moral analysis? Does intent matter or only impact? Is equality a matter of treatment or outcomes? Are institutions obligated to pursue substantive justice or to remain procedural? These kinds of contested premises drive an enormous number of political disagreements in American life.

Importantly, there aren’t infinitely many of such contested premises. If we sat down and tried to list the load-bearing premises behind contemporary political conflict, I suspect we’d get to twenty or thirty before the list started repeating itself. Maybe fewer. In other words, the set is finite and it’s mappable.

Imagine an AI that, when asked about a politically charged topic, surfaced its assumptions the way a good academic paper is supposed to surface its priors. This response treats disparate outcomes as evidence of discrimination — a contested causal claim. This response assumes that historical injustice creates ongoing institutional obligations — a contested normative position. This response frames the question in terms of group identity rather than individual circumstance — a contested explanatory choice.

That kind of transparency changes the user’s relationship to the output. Instead of absorbing AI-generated content as background fact, the user encounters it as a position — one that can be examined, challenged, or rejected.

This approach scales to almost every concern about AI and democracy. Deliberately false information gains traction precisely because trust in legitimate sources is already depleted — and that depletion was itself produced and sustained by the same kind of unmarked loading. Bot swarms manufacture the appearance of broad agreement, banking on the unstated premise that consensus signals truth. Make the premise visible and the swarm loses its grip. Micro-targeted political messaging exploit unstated assumptions about what will fire up a particular audience. Making those assumptions visible to the recipient strips the message of its force. Manipulation that has to announce itself isn't really manipulation anymore — it's an argument the recipient can accept or reject on the merits. Finally, there’s another useful side effect of making assumptions clear: if explicit-assumption architecture becomes standard, content that lacks it stands out. Targeted lies and fabricated material become easier to flag, not harder, because they fail to do what legitimate AI output is expected to do.

To be sure, making contested premises legible doesn’t solve the deepfake problem. Nothing in the framing layer addresses identity fraud. But it solves, or at least addresses, almost everything else on the list of AI-and-democracy concerns.

Democracy doesn’t require that citizens agree on contested questions. It requires that the institutions mediating public life — including, now, AI systems — remain legible enough that citizens can see what they’re being asked to accept. The danger isn’t that AI takes positions. It’s that AI takes positions invisibly, at scale, in the voice of neutral authority and factual claims.

The fix isn’t to strip the positions out. It’s to put them on the surface, out in the open, where they belong.

Subscribe now

I created a Claude project that draws upon my work to surface the assumptions behind politically contested/contestable claims. This morning, I tested it on a NYT op-ed titled “The Town That Reveals All of Trump’s Bad Economic Ideas.” Below is what it gave me. (“Cat 1” refers to facts, “Cat 2” to interpretive and causal claims, and “Cat 3” to values and goals.)

The op-ed’s central move is using Hickory’s recovery to argue Trump’s second-term policies are “counterproductive” — but the recovery it documents runs through mechanisms (aging retirees spending Social Security, immigrant labor filling service jobs, organic industrial repurposing, municipal investment) that are arguably largely independent of the federal policies being criticized. The op-ed’s case against Trump rests primarily on prospective harm to those mechanisms, not demonstrated reversal. That gap is where most of the divergence lives.

Cat 2 — Causation (the load-bearing weakness)

• The recovery’s mechanisms — retiree consumption, immigrant labor, fiber optics absorbing textile skills, data centers using legacy water infrastructure, a $40M municipal bond, an Appalachian State campus — are plausibly not mechanisms tariffs or deportations created or accelerated. The op-ed’s case against Trump is arguably really: “these mechanisms are fragile and his policies threaten them.” Whether that threat is materializing is shown thinly: Shuford’s tariff bill, closed Hispanic businesses, one deportation, fear-driven absenteeism. Suggestive, not demonstrated.

• The economists’ counterfactual — slower Chinese import growth → muted rightward shift → Trump loses 2016 — is a heavily underdetermined causal chain presented as analysis. A reader could accept the China shock findings and reject this specific counterfactual.

• “Job growth in China-shocked areas now outpaces less-affected areas” is consistent with recovery, but also plausibly with mean reversion from a deeper hole. Framing implies the former.

Cat 1 — What counts as “recovery” (and who counts as the recovering thing)

• The piece concedes wages are lower ($17.60 health care support vs. $22.15 production) and quotes Hanson: “a jobs recovery, not a good jobs recovery.” But the headline calls it recovery anyway.

• Deeper: the op-ed treats “Hickory” as the unit that recovers — a region, an aggregate. But a region isn’t a moral patient; only the people in it are. The shift from “the workers displaced in 2001-2013 suffered badly” to “the region recovered (via aging retirees, immigrant newcomers, and young Charlotte commuters)” is arguably a category change that goes unmarked. On a more individualist frame, this isn’t recovery — it’s replacement. The displaced workers largely didn’t recover; different people moved into the geography. Whether that counts as the same story having a happy ending is a contested philosophical question the op-ed silently resolves.

• “China shock” as a measured phenomenon is treated as settled. There’s economics literature pushing back on the magnitude and on whether the 2.4M figure properly accounts for offsetting gains. Not surfaced.

Cat 3 — Whose recovery counts

• The implicit success criterion is regional aggregates (population, median income, employment levels). An alternative frame — did the families displaced in 2001-2013 recover? — yields a much darker picture (disability, overdoses, kids steered away from factory work, lost intergenerational skill). The op-ed acknowledges these but doesn’t let them disturb the recovery narrative.

The Cat 4 political prescription (Democrats can win these areas by opposing Trump) rests on the Cat 2 causal claim that his policies are actively harming the recovery. If that claim is weak, the prescription weakens with it.

Subscribe now

This morning I read a New York Times piece about Pierre-Édouard Stérin. I’d never heard of him before. Apparently, he’s a French billionaire funding candidates and think tanks to make France, in his words, “less Muslim, more Catholic, and more capitalist.” I confess, I’m not particularly interested in Stérin himself. I am, however, interested in the use of the term “far right.” This morning’s article used it nine times. That alone wouldn’t be worth remarking on—after all, the term is pervasive. We regularly hear it used with characters like Nick Fuentes, parties like the AfD in Germany, and of course President Trump himself. And yet, I have yet to see anyone stop to seriously consider what our use of the term is costing us.

My use of “far right” as a motivating example need not be read as a failure to see the parallel of “far left” (“radical left” works much the same way). Nor is it a denial that both terms tend to function as a sort of lazy and convenient shorthand. While each may have the advantage of saving us a few words, both need to be dropped from discourse and, more importantly, recognized as obstacles to clear political reasoning.

“Far right” and “far left” trick us, in part because they feel like straightforward descriptions. As if there’s an obvious spectrum of political opinion, and these terms simply locate what falls outside its acceptable range. The Overton Window is likely the version of this idea most people have encountered: the notion that at any given moment, there’s a corridor of sayable positions, and what lies beyond it can be comfortably labeled “extreme.” The terms “far right” and “far left” essentially import the same logic into everyday language. They imply that there’s some center domain, however defined, that contains the natural zone of reason in a given place and time. The implication is that positions outside that domain don’t require engagement—they require a label. But neither “far left” nor “far right” is, in fact, a descriptive term. Each is a verdict—delivered in place of an argument.

They draw a line and declare everything past it deviant—not objectionable in ways we can specify, not advocating for tradeoffs we find unacceptable for specific reasons, but simply beyond the consideration of morally serious people. The label does the work that argument is supposed to do.

That substitution has consequences. For starters, positions that get labeled rather than engaged don’t disappear—they find other channels, other audiences, and often grow. The label doesn’t resolve disagreement. It forecloses it. And foreclosed disagreements have a way of returning, louder and less tractable than before. The second consequence may be less visible but is equally corrosive: the terms degrade the thinking of whoever uses them—and, as with the NYT piece, whoever reads them uncritically as well.

I focus on “far right” because the term is doing more work in our political conversation than its counterpart. In the current context, “that’s far right” does the same work as “that’s racist.” Both convert a position that could be argued with into one that can only be condemned. The argument is over before it begins.

And yet, democracy is built on a specific premise: that reasonable people, weighing real and competing goods, will reach different conclusions. Liberty against security. Individual autonomy against collective welfare. Continuity against change. Cultural cohesion against pluralism. There are genuine losses on every side of these contests. The person who weights them differently than you isn’t necessarily confused or malicious. They may be making a different but recognizable human judgment about what matters most.

That’s not a squishy, feel-good observation. It’s a structural requirement. Democracy doesn’t work if we pre-decide which weightings are legitimate before the contestation begins. It works by letting those weightings compete, negotiate, and produce outcomes that participants can accept even when they lose—because they were heard, because the process was fair, because their position was engaged rather than dismissed.

“Far” short-circuits that. It doesn’t say: here is what this person is claiming, here is the tradeoff they’re making, and here is why I disagree with it. It says: this position has left the space where democratic reasoning applies. And once that move is made, the appropriate response is no longer argument. It’s exclusion.

Consider what the Times piece actually tells us about Stérin. He wants less immigration, particularly from Muslim countries. He wants abortion banned. He wants to privatize education and healthcare and slash the welfare state. He wants to ban Muslim dress in public spaces. These are positions. They involve contestable empirical claims and genuine value tradeoffs—about national identity, religious pluralism, the role of the state, the relationship between culture and citizenship. People who hold versions of these views represent, in various combinations, the sincere convictions of hundreds of millions of people across Western democracies.

Are those positions ones that most readers of the NYT piece likely find objectionable? Almost certainly. And yet, labeling them “far right” doesn’t tell us what specifically is wrong with the reasoning behind them. It tells us the conversation is over. But ending the conversation isn’t the same as winning it.

None of this means Stérin’s positions are beyond criticism. The same standard applies in every direction—his critics’ arguments are no more immune to scrutiny than his own. It means that how we criticize matters. If his case for limiting Muslim immigration relies on a causal story about cultural cohesion that doesn’t hold up, say that. If his position on abortion imposes one set of values on people who hold competing ones with equal sincerity, make that argument. If his vision of a privatized welfare state would produce specific harms to specific people, name them. These are criticisms with content. They can be evaluated, contested, and answered. “Far right” can’t be any of those things. It’s not a criticism. It’s a dismissal dressed up as one—and dismissals don’t constitute an argument.

Now here’s where the predictable objection likely arrives: what about figures who go beyond tradeoffs entirely, into explicit claims about who belongs in democratic life? I mentioned Nick Fuentes earlier. He’s said women shouldn’t be allowed to vote, that black people should “for the most part” be imprisoned, and that white men should run the country, the companies, and the household. Shouldn’t we use the term “far right” to ostracize him? Not only that, but don’t we have a moral obligation to do so?

It’s an understandable concern—but working through it makes clear what’s wrong with this line of thinking.

Fuentes’s statements can be disputed on the grounds that they’re explicit claims about who belongs in democratic life and who doesn’t. That’s a specific charge—and it deserves to be stated with precision. “Far right” doesn’t do that. It buries the charge in a geographic metaphor and lumps Fuentes together with anyone else who has wandered past an invisible, and ever-moving, line. The specific accusation—that he is advocating the explicit removal of certain groups of people from democratic standing—gets lost in the label. That accusation deserves to be made directly: that claims about who belongs in democratic life are incompatible with the basic commitment to equal standing that democracy requires.

That said, a harder objection remains: even if Fuentes’s views aren’t formally about democratic exclusion—even if they’re just about how society should be organized—surely democracy isn’t obliged to dignify them with argument? Surely some views are simply too repugnant to engage?

That feeling is understandable. But notice what it’s asking for: the right to remove views from democratic contestation rather than argue against them. That is precisely the move that makes democracy fragile. The line between “too repugnant to engage” and “uncomfortable to engage” is drawn by whoever controls the labeling. And once we grant that power, we’ve conceded something more important than the argument we were trying to avoid.

This pattern—reaching for a label that condemns when a precise charge is available—has a name. There’s a word for using moral righteousness as a substitute for argument: sanctimony. And sanctimony has a democratic cost. It signals to everyone on the receiving end of the label that the game is rigged, that the contestation isn’t real, that the outcome was decided before the argument began. That signal is corrosive. It confirms, for those who already distrust democratic institutions, that their distrust is warranted.

The left-right spectrum was always a simplification. But “far” turns it into something worse: a sorting mechanism that separates legitimate disagreement from pre-disqualified deviance, with the cut point set by whoever holds cultural authority. The center gets to feel like the zone of reason by default. Everyone past the line gets to be dismissed without engagement. And the actual arguments—the specific claims, the real tradeoffs, the genuine goods in tension—never have to be named.

We’d have more substantive politics and a democracy that earns its name if we did the harder work of saying specifically what we mean—rather than reaching for a label that tells us we don’t have to.

Subscribe now

A colleague emailed me this week; someone I hadn’t heard from in a while. He was curious if I’d written anything about Iran (I hadn’t). He told me he has students serving in the military. That he’s scared. And that he’s angry.

Since he reached out, I’ve been sitting with the question of whether I have anything useful to say on the topic. Is it the time to remind people to avoid the Certainty Trap? That, regardless of your opinion on the U.S. and Iran, it’s important not to demonize or morally condemn people who disagree? I could write that piece and write it honestly—given it’s what I actually believe.

And yet, when I sat down to write, I realized that, while that statement is true, it’s not really a description of what I think or feel when I see the headlines. So, while I don’t know if what follows will be useful, it will, at the very least, be honest.

***

What I don’t feel in this moment is outrage. Not at Trump, Hegseth, the Iranian regime, or anyone else for that matter. And I wonder what that says about me; whether I’m not feeling the one thing I’m supposed to—the thing that a good person would—at this moment.

Here’s what I do feel.

For starters, I feel scared. The reasons for this are probably obvious. Like so many people, I worry that this conflict will escalate to an apocalyptic, existential, World War III event. I fear what the future holds for my teenage kids. What kind of world will they inherit? Could this lead to a draft that affects them? What does real economic destruction look like? I get lost in the sheer number of things I take for granted every single day that could vanish in the blink of an eye. I go to the grocery store and expect that the shelves will be stocked. I flip a light switch and assume the electricity will come on. And if, God forbid, I have to call 911, I assume that someone will answer and someone will come. But what if everything I assume will always be there is actually deeply, heartbreakingly, fragile?

And I feel sad. I feel sad for the people everywhere—including in Iran and Israel—who are going about their lives and already live with the answer to the question I just posed. Yes, everything really is that fragile. Life can actually be that fragile.

I also feel something harder to admit; it’s a thought, a question really, about whether war is just something humans do to each other, whether it’s unavoidable. And I’m not sure if that thought is wisdom or resignation.

Last, I feel a kind of dizzying awareness of my own undeserved luck. I’m watching the headlines roll in from my comfortable desk, in my house, in a country that isn’t on fire. Not because I did something good or right. Just because of where and when I was born. And I don’t know what to do with that. I don’t know that it’s helpful (or necessarily accurate) to lean into the idea the circumstances of birth determine everything in life. And I also can’t shake the recognition that I didn’t earn THIS. Where exactly the line is between luck and agency, between the randomness of birth and the choices that follow from it—I genuinely don’t know. And this moment makes that uncertainty harder to sit with, not easier.

But feelings aren’t verdicts. And if a verdict is what people want to hear, I can’t provide one.

To be sure, it may well be true that Trump is surrounded by sycophants and “yes” men and that he’s chased away all the people who would have reined in his baser impulses. And yet, even if that’s the case, that doesn’t change the fact that I don’t know the right answers to questions like: Was the alternative—continued diplomacy, containment—clearly better? Was the diplomatic track going anywhere? What would “responsible” or “reasonable” action have looked like?

If you read those and think you know the right answers—the one decision that a competent, decent administration would obviously have made—I’d ask you to sit with that certainty for a moment. Not because I’m defending what happened or what’s going on. But because what I see when I look at this honestly isn’t a clear choice between right and wrong. It’s tradeoffs. Painful, uncertain, potentially catastrophic tradeoffs, with no option that doesn’t carry serious costs. That’s not a defense of anyone. It’s just what is.

Maybe this is a Certainty Trap piece after all. Just not the one I thought I'd write.

Subscribe now

What I’m about to say won’t satisfy anyone. I’m going to say it anyway.

Two people have been killed by federal agents in Minneapolis. One an ICU nurse, another shot in her car. Protests have nearly shut down the city. The left says we’re watching fascism unfold. The right says we’re watching law enforcement do its job.

Here is what I know:

Democracy was never meant to settle moral questions. On the contrary, it rests on the shared understanding that determining moral truth is not the project we are engaged in. It gives us a way to act collectively—pass laws, enforce borders, build institutions—without resolving questions that human beings have never resolved, like: What do we owe strangers? How should we weigh the needs of citizens against those who want to become citizens? What does justice require when outcomes are unequal?

These are real moral questions. And they don’t have obvious answers. Reasonable people have disagreed about them for centuries and will disagree for centuries more.

In the late 1960s and early 1970s, the left stopped treating a particular disputed moral question as open. Not through argument—through institutions. Through legal doctrines that assumed the answer. Through policies and social norms that made disagreement socially costly. Through a slow redefinition of what it meant to be a decent person, until holding the wrong view became evidence of bigotry rather than evidence of reasoning. It started with: Is inequality unjust? (the answer was determined to be yes) and became a template.

Once democracy—via institutions with epistemic and democratic authority—could be seen as settling one contested moral question, it could settle others. Climate. Immigration. Trade. The scope of government. What counted as violence. What counted as speech. The questions closed not because anyone explicitly banned deliberation, but because deliberation requires believing your opponent is a legitimate moral agent. If their position is diagnostic of bigotry or stupidity, there’s nothing to deliberate about.

Then came Trump. You can argue about what he is. What I can tell you is what his victory meant: the moral questions that had been treated as closed weren’t closed after all. Calling something racist didn’t end the conversation. Trump stood, and still stands, for the millions of people who had never accepted the settlement of these questions in the first place. They had only been silenced by it. When you treat contestable positions as settled and opposition as morally disqualifying, you don’t resolve their disputed nature. You just stop discussing them. And the people you’ve excluded don’t vanish. They wait.

To be sure, Trump could have reopened those questions through argument and deliberation. He didn’t. He imposed different answers. Deportations. Tariffs. Dismantling. He didn’t offer reasons. He simply acted.

Someone on the left will object here: But our answers made the world better. Theirs make it worse. That's the move. That's always the move. Both sides believe this. Again: democracy wasn't designed to answer these questions at all. It was designed to function despite the disagreement being irresolvable.

Now both sides are locked in a fight that cannot end, because neither side is actually asking the questions that need to be asked. The left treats enforcement as self-evident cruelty—not a policy to be debated but an evil to be resisted. The right treats enforcement as self-evidently justified—not a policy to be calibrated but a battle being won.

But the questions are still there, quietly waiting for our engagement: What do we owe people who come here without authorization? What level of enforcement is proportionate? Under what circumstances do we accept costs to pursue legitimate aims? What obligations do nations have to each other? When does pressure become recklessness? What do we owe our own citizens that we don’t owe others? How do we weigh economic pain now against strategic benefit later? What does proportionate force look like when people resist?

These aren’t being deliberated. They’re being fought over. The winner’s answer gets imposed. That’s not democracy. That’s just power.

I watch some of the footage from Minneapolis and I see something I don’t hear nearly enough: both sides have abandoned the democratic project. The left abandoned it first, by treating moral questions as settled and opponents as illegitimate. The right has abandoned it now, by treating opponents the same way—not as fellow citizens to engage with, but as enemies who deserve no consideration, no lament, no restraint.

Two Americans are dead. The president calls one a terrorist. The left calls the agents murderers. No one asks: how did we get to a place where we cannot even mourn together?

I am confident in my assessment of how we got here. However, I’ll be the first to admit that I don’t know how we get out. Maybe understanding what broke is the first step. Maybe it’s the only step available right now. Not a program. Not a solution. Just seeing the mechanism clearly. A commitment to stop making it worse. Remember that you hold a contestable view on an open question—and so does the person across from you.

Democracy has always required a category: “good person who disagrees with me.” The left emptied that category when they turned disagreement into diagnosis. The right keeps it empty.

Now there is no one to deliberate with. The questions hang in the air, unanswered and unasked. We call it democracy because we still hold elections. But the thing democracy was supposed to do—let us reason together despite disagreement—we can’t do anymore.

I’m putting this here, for later. Maybe someone will find it and it will help explain what they’re trying to understand. Of course, I could be wrong about all of it.

But I don’t think so. And I don’t think anyone can hear it now.

Subscribe now

Welcome to the last post of this series!

[Full contents linked here]

Donald Trump won the 2024 election with 77 million votes. He won despite two impeachments, four indictments and a civil fraud conviction. He won after losing the 2020 election and attempting to overturn the results. He won after January 6th. And eighty-nine percent of Trump voters agreed that “American values and beliefs are being undermined and cherished traditions are under threat.”

Many people—scholars, pundits—have argued that this is the result of deeply ingrained racism. But this interpretation only makes sense within the framework I’ve just described over the last eight posts.

Outside that framework, Trump’s continued popularity, even after everything that should have disqualified him, is evidence that the absence of democratic legitimacy matters.

The Framework His Voters Reject

For millions of Americans, the disparate impact framework isn’t an abstract legal doctrine. It’s the water they swim in:

Their children’s elementary schools teach that colorblindness perpetuates racism and that white privilege explains inequality. When they object, they’re told they’re “uncomfortable confronting racism” or that their concerns prove why such education is necessary.

Their employers require unconscious bias training asserting they harbor biases they’re unaware of. Disagreement with the training’s premises becomes evidence of the biases it describes. Defending merit-based evaluation marks them as perpetuating discrimination.

Their local government adopts equity frameworks evaluating policies by demographic outcomes. Public meetings feature language about “systemic racism” and “structural inequity.” Opposition to specific policies gets characterized as opposing racial justice.

Their preferred news outlets are labeled “misinformation.” Their social media posts get flagged for violating community standards. Universities their children apply to weigh diversity statements alongside academic credentials. Professional licensing requires continuing education on implicit bias.

At every institution they encounter—education, employment, media, government, professional licensing—the same framework operates. And they never got to vote on whether they wanted it.

The Choice They Face

For non-progressives, this creates a binary:

Option A: Accept that your belief in colorblind meritocracy perpetuates discrimination. Accept that your skepticism about unconscious bias proves you have unconscious bias. Accept that institutions should pursue demographic proportionality regardless of individual qualification. Accept frameworks embedded across institutional life, frameworks you cannot democratically change because they were never democratically adopted.

Option B: Support politicians who reject those frameworks entirely, attack institutions that impose them, and promise to dismantle the infrastructure built to enforce them—even when those politicians are obviously flawed.

A rational voter would choose Option B simply because Option A requires accepting his own moral illegitimacy on terms he never consented to.

When Trump said, for instance, in 2023:

…[W]e will root out the Communists, Marxists, Fascists, and Radical Left Thugs that live like vermin within the confines of our Country, lie, steal, and cheat on Elections, and will do anything possible, whether legally or illegally, to destroy America, and the American Dream.

Those inside the intent-over-impact moral framework see yet more evidence of his racism and authoritarian ways; those outside it see someone finally willing to stand up to the sanctimony.

Why Institutions Can’t Understand This

Institutional responses to Trump’s popularity reveals the problem’s depth.

Academic analysis focuses on racism, authoritarianism, misinformation, and economic anxiety. These explanations assume Trump voters are either morally deficient, cognitively limited, or economically desperate. What they don’t consider: Trump voters might have rational reasons for rejecting institutional legitimacy.

Media coverage characterizes Trump support as backlash against progress. This framing assumes the institutional framework represents progress and opposition represents regression. It doesn’t consider that reasonable people might disagree about whether the framework is actually progressive.

Corporate and university leaders express frustration: we built diversity offices, implemented training, adopted inclusive language—they hate us because they’re racist. This conviction itself reveals the problem: these leaders genuinely believe they did the right thing and cannot imagine legitimate objection.

The Methodological Trap

Critics will inevitably point to social science research showing Trump supporters score high on “racial resentment” scales or display “authoritarian” tendencies. Before accepting these findings as counter-evidence, examine what these instruments actually measure.

The most widely used racial resentment scale includes items like:

• “Irish, Italians, Jewish and many other minorities overcame prejudice and worked their way up. Blacks should do the same without any special favors.”

• “Generations of slavery and discrimination have created conditions that make it difficult for blacks to work their way out of the lower class.” (disagreement indicates resentment)

• “It’s really a matter of some people not trying hard enough; if blacks would only try harder they could be just as well off as whites.”

Notice what these measure: disagreement with the premise that current disparities result primarily from discrimination and structural barriers. Someone who believes individual effort matters, that group-based remedies are inappropriate, or that historical disadvantage doesn’t fully explain present inequality will score as having “racial resentment.”

Similarly, items measuring “modern racism” or “symbolic racism” include:

• Opposition to affirmative action

• Belief that discrimination is no longer a major barrier

• Emphasis on individual responsibility over systemic explanations

These aren’t neutral measures of racial animus. They’re measures of disagreement with the disparate impact framework itself. The scales operationalize rejection of “disparities indicate discrimination” as evidence of racism.

Authoritarian personality research creates similar circularity. Studies often classify as “authoritarian” preferences for:

• Traditional values over progressive change

• Order and stability over disruption

• Skepticism toward expert consensus

• Resistance to institutional mandates

But within the disparities equal discrimination, impact-over-intent framework, what looks like authoritarianism might be resistance to institutional frameworks imposed without consent. The person who rejects university diversity training isn’t necessarily authoritarian—they might be rejecting institutional authority that operates outside democratic accountability.

This creates an epistemological trap: the social science establishment uses instruments that presuppose the framework’s legitimacy to study rejection of that framework, then characterizes rejection as moral pathology.

The trap has several components:

Circular measurement: Disagreement with disparate impact thinking → high score on racial resentment → proves racial bias → validates treating disagreement as resentment.

Unfalsifiability: Any challenge to the framework becomes evidence supporting it. Questioning whether disparities prove discrimination demonstrates racial resentment. Resisting institutional mandates demonstrates authoritarianism. Economic concerns get dismissed as cover for racial motives.

No neutral ground: We cannot study why people reject institutional authority using tools that assume such rejection indicates moral failure. The measuring instruments embed the disputed premises.

Foreclosed alternatives: The research design eliminates the possibility that Trump supporters have legitimate objections to how the framework was adopted. Either they’re racist (high racial resentment), authoritarian (high RWA scores), or economically anxious (the acceptable explanation that avoids the democratic legitimacy question).

This isn’t to claim Trump supporters aren’t racist, or that racism plays no role in political behavior. It’s to note that we cannot test whether the democratic legitimacy argument explains Trump’s appeal when our analytical tools treat rejection of the framework as diagnostic of racism.

Consider what this means for democratic deliberation: if social science operationalizes disagreement with elite consensus as pathology, then research will always “prove” that opposition to institutional frameworks reflects moral deficiency rather than legitimate disagreement. The framework becomes unfalsifiable—not because it’s correct, but because we’ve made it impossible to measure alternatives.

This is precisely the problem: when rejection of a framework becomes evidence of moral failure, democratic deliberation becomes impossible. When rejection of a framework becomes evidence of moral failure, we cannot neutrally study why people reject institutional authority, cannot acknowledge space for legitimate disagreement, and cannot have the democratic conversation required to restore political stability.

The Cycle We’re Trapped In

The pattern now cycles predictably:

1. Institutions adopt impact-over-intent thinking without democratic mandate

2. Objection and opposition gets characterized as moral failure and bigotry

3. Voters support politicians who disrupt this

4. Those politicians use executive power to reverse institutional practices

5. Institutions and progressive states resist and strengthen their commitments

6. Progressive politicians eventually regain power and reverse the reversals

7. The cycle continues

This is what happens fifty-four years after a transformation begins without democratic deliberation.

The Evidence We Have

Trump’s popularity across multiple elections, despite everything that should have disqualified him, provides strong evidence that the absence of democratic process in adopting the intent-over-impact framework matters for political stability.

Millions of Americans face institutions operating from a framework they never consented to, only to find their disagreement characterized as moral failure. And, because it’s one that no one explicitly acknowledges, there’s been no real path to demand change. When an institution is biased and acknowledges it, you can argue about what should be done. When there’s bias and it’s denied, you have no recourse other than to support politicians who promise to reject those institutions entirely—even when the politicians are badly flawed.

This doesn’t make Trump voters correct (or incorrect) in some broader sense. It makes their response rational given the conditions created by transformation without deliberation.

From Impact-Over-Intent to Trump

The arc from Griggs v. Duke Power to Trump’s 2024 election reveals what happens when fundamental moral transformations occur without democratic process. At no point did the American public have a serious democratic conversation about whether this framework should govern institutional life.

The result: portions of the population whose views are characterized as illegitimate support politicians who promise to destroy those institutions—even when destruction is all those politicians offer.

The Path We’re Not Taking

The alternative to this cycle would be to:

• Acknowledge that impact-over-intent catches hidden discrimination but eliminates space for legitimate disagreement—and both effects matter.

• Recognize that causes of inequality are complex, intervention effectiveness is contested, and people of good faith can disagree about appropriate responses.

• Create space for positions between “disparate outcomes always indicate discrimination” and “discrimination doesn’t exist”—positions that millions of Americans actually hold.

• Allow democratic processes to determine which domains should be governed by disparate impact analysis and which should emphasize other values.

What Trump’s Popularity Actually Reveals

The question isn’t whether disparate impact doctrine is correct. The question is whether American democracy can survive when fundamental transformations happen without democratic process, creating conditions where roughly half the country supports politicians promising to destroy institutions the other half believes in.

Fifty-four years after Griggs, Trump’s popularity is evidence that we never had the conversation we needed. His voters aren’t asking whether disparate outcomes indicate discrimination—they’re asking why they never got a vote.

Until we have that conversation—the conversation about whether this framework should govern institutional life, what tradeoffs we’re willing to accept, which values we prioritize when they conflict—we’ll continue cycling between elite institutional consensus and populist institutional rejection.

That conversation starts with acknowledging a transformation this fundamental should never have happened without us.

Subscribe now

Second to last post of the series!

[Full contents linked here]

Various responses to the polarization and institutional transformation have emerged. Advocates promote civil discourse to restore respectful disagreement. Free speech organizations defend expression against institutional restrictions. Initiatives encourage viewpoint diversity on campuses and in workplaces. Executive orders attempt to dismantle DEI infrastructure.

These approaches address real problems but share a common limitation: they treat the issue as procedural when it’s really about moral legitimacy.

The Moral Legitimacy Problem

The impact-over-intent framework doesn’t just restrict certain speech or discourage certain viewpoints. It characterizes them as morally illegitimate. The problem isn’t that people can’t express these views. It’s that expressing them places one outside the bounds of legitimate moral participation.

This is fundamentally different from procedural restrictions on speech or limits on viewpoint representation. It’s a question of whether certain positions can be held by people of good will.

Why Civil Discourse Initiatives Can’t Work

Numerous organizations promote civil discourse—teaching people to disagree respectfully, listen charitably, assume good faith, and find common ground.

These skills matter. But they presume all parties grant each other basic moral legitimacy. Civil discourse means treating opponents as fellow citizens who might be wrong but aren’t evil.

The framework eliminates this presumption for certain positions.

If someone argues that college admissions should be based primarily on academic metrics, and such systems produce disparate racial outcomes, the framework characterizes this position as perpetuating systemic racism. The person isn’t just wrong about admissions policy—they’re advocating for a discriminatory system.

How does one have civil discourse with someone advocating discrimination? The entire premise of civil discourse—that we’re all trying to solve problems together despite disagreement—collapses when disagreement itself indicates moral failure.

Teaching respectful disagreement skills doesn’t solve this problem because the problem isn’t about tone or technique. It’s about whether the disagreement is legitimate in the first place.

Why Free Speech Protections Are Insufficient

Various organizations document speech restrictions and defend expression rights. Their work is valuable. Speech codes that explicitly prohibit certain viewpoints should be challenged. Due process matters when students or faculty face discipline.

But legal protections for speech don’t address the underlying issue. A university might allow a speaker to appear on campus while the institutional community characterizes that appearance as “platforming hate” or “causing harm.” The speech is technically protected but morally delegitimized.

Or consider an employee who argues against race-conscious hiring during a diversity training session. Their employer might not formally discipline them, but they’ve identified themselves as someone who defends discriminatory practices. Their speech is legally protected but professionally costly.

The framework operates through moral authority, not legal prohibition. Strengthening First Amendment protections doesn’t change whether certain positions are considered morally legitimate by institutional communities.

This explains why campus speech debates often feel unresolved. Free speech advocates argue for protection of controversial expression. Critics respond that protection doesn’t equal endorsement, and that some views cause harm even when legally permitted. Both are correct within their own frameworks.

Why Viewpoint Diversity Can’t Save Us

Viewpoint diversity initiatives generally argue that institutions benefit from intellectual disagreement and should include perspectives across the political spectrum. This seems reasonable—intellectual progress requires engaging with competing views.

But the framework we’re talking about makes viewpoint diversity on certain questions a moral impossibility.

If conservative scholars argue that achievement gaps reflect factors beyond discrimination—family structure, cultural emphasis on education, individual choices—the framework characterizes this as denying systemic racism. If systemic racism is proven by disparate outcomes, then questioning whether disparities prove discrimination becomes denying established facts.

In this line of thinking, including such viewpoints isn’t diversity—it’s giving legitimacy to positions that perpetuate racism and other forms of harm. The framework requires choosing between viewpoint diversity and preventing discrimination. It doesn’t allow having both.

Why Executive Orders Repeat the Problem

The Trump administration’s response to institutional transformation has been executive orders: banning DEI offices, eliminating diversity training, directing federal agencies to evaluate applicants without considering race.

These orders appeal to voters who want rapid reversal of policies they never consented to. But they face fundamental problems:

Executive orders repeat the original error: the framework was adopted without democratic deliberation, and reversing it through executive action also bypasses deliberation.

They misdiagnose the scope. Most institutional adoption was voluntary. Universities built diversity offices because administrators believed it was right. Corporations spent billions because leadership was convinced and employees demanded it. Government mandates explain some infrastructure, but not the breadth of institutional convergence.

Executive orders can eliminate federal requirements, but they can’t undo moral conviction. They can’t force universities to close diversity offices they chose to create. They can’t make corporations abandon commitments they made voluntarily. They can’t change the fact that substantial portions of the population—including majorities among the young and college-educated—genuinely support the framework.

They create backlash cycles. When Trump issues executive orders dismantling DEI, institutions and states with progressive leadership respond by strengthening their commitments. California passes laws protecting diversity programs. Universities rename offices to avoid federal oversight while continuing the work. Corporations maintain private initiatives even as they eliminate public ones.

The result: executive power gets used to impose, then reverse, then re-impose frameworks—with each side using whatever authority they temporarily control to override the other. This is the opposite of democratic deliberation.

The Common Error

All three approaches—civil discourse, free speech, and viewpoint diversity—treat this as a problem that can be solved without addressing the root framework itself.

They assume:

• People just need to be more civil (but civility presumes legitimate disagreement)

• Speech just needs better protection (but the problem is moral rather than legal)

• Institutions just need more viewpoint diversity (but some viewpoints are characterized as harmful)

What all three miss: you cannot restore legitimate disagreement about inequality without confronting whether disparate outcomes should be treated as discrimination regardless of cause.

What Would Actually Be Required

Addressing the moral legitimacy problem would require:

Reexamining the framework: Opening for discussion whether disparate outcomes always indicate discrimination, whether intent matters for moral evaluation, and whether alternative explanations for inequality can be legitimate.

Acknowledging tradeoffs: Recognizing that impact-over-intent catches hidden discrimination but eliminates space for morally legitimate disagreement, and that both effects matter.

Distinguishing empirical from moral questions: Treating “do disparities indicate discrimination?” as debatable rather than settled, while agreeing that discrimination (however defined) is wrong.

Restoring intent as relevant: Accepting that without proof of discriminatory purpose, practices producing disparate outcomes might not constitute moral wrongdoing, even if they’re legally problematic under current doctrine.

Rebuilding tolerance for disagreement: Teaching that people can disagree about inequality’s causes and appropriate responses without being racist or indifferent to injustice.

What This Means

The current moment features extensive debate about polarization, speech restrictions, viewpoint uniformity, and institutional overreach. But much of this debate addresses symptoms rather than causes.

Until the foundational questions about disparities and intent receive serious democratic attention—with recognition that reasonable people can disagree—the various solutions being proposed will continue failing. We can’t restore democratic deliberation without addressing the framework that made certain positions democratically illegitimate.

Next: What Trump’s Popularity Reveals

Subscribe now

Hi everyone! I did a fun Substack Live with Joe Walsh yesterday evening—please check it out! :-) Since we have some new readers this morning, I want to situate the current post for them.

A few days ago, I started a 9-part series on how a new definition of discrimination changed how we think about equality, racism, and justice. This is the 7th post in that series, so I’ll be wrapping it up on Saturday.

Enjoy!

[Full contents linked here]

Impact-over-intent thinking spread from a legal doctrine about employment tests to a comprehensive framework governing American institutional life. This transformation occurred through legal expansion, academic validation, institutional adoption, and political mainstreaming.

But this history raises a fundamental question: What happens to democratic deliberation when such a framework becomes unchallengeable?

The Core Problem

The impact-over-intent framework rests on multiple premises:

1. Disparate outcomes indicate discrimination

2. Discrimination stems from racism, which is morally wrong

3. Intent doesn’t matter for determining wrongdoing

Taken together, these create a logical problem for democratic discourse. If disparate outcomes equal discrimination, then several things follow:

On causation:

• Questioning whether a disparity indicates discrimination becomes questioning whether discrimination occurred—which can be characterized as denying its existence.

• Offering alternative explanations for inequality becomes making excuses for discrimination.

• Policy disagreements about how to address disparities become disagreements about whether to address discrimination—which places one side in an apparently indefensible position.

On the unit of analysis:

• The group becomes the proper unit for measuring justice, not the individual. If outcomes must be proportional across demographic groups, then individual merit or individual circumstances become secondary considerations—or obstacles to justice.

• Defending individual-focused policies (like merit-based admissions or colorblind hiring) means defending a system that produces discriminatory group outcomes.

• There’s no principled way to argue that individuals should be evaluated individually when groups show disparate results.

On harm:

• Harm becomes defined by the recipient’s experience rather than the actor’s conduct. If someone from a marginalized group experiences a comment or policy as harmful, it is harmful—regardless of intent, context, or whether others would experience it differently.

• This makes harm unfalsifiable: the person claiming harm is the authority on whether harm occurred.

• Disagreeing with someone’s characterization of harm becomes dismissing their lived experience, which is itself a form of harm.

On remedies:

• If disparities prove discrimination, then eliminating disparities becomes the measure of whether discrimination has been addressed. Proportional representation across groups becomes the goal.

• Any policy that doesn’t actively work toward proportional outcomes can be characterized as perpetuating discrimination.

• There’s no endpoint except demographic proportionality—any remaining disparity indicates remaining discrimination requiring continued intervention.

How This Plays Out

Consider the following scenario:

A school board in a suburban district debates eliminating its gifted and talented program (much like this). District data shows the program enrolls 45% Asian students, 35% white students, 12% black students, and 8% Hispanic students—dramatically different from the district’s overall demographics. Entrance is by standardized test scores, which any student can take.

At the packed board meeting, some parents argue the program should remain because it serves students who need more advanced material and admission is based on objective criteria available to all. Other parents argue that the demographic disparities prove the system discriminates against black and Hispanic students.

Under the framework, parents defending the gifted program aren’t simply arguing for one educational approach among many reasonable options. They’re defending a practice that produces disparate outcomes, which the framework defines as discrimination. Several teachers who privately support keeping the program stay silent, knowing that speaking up will mark them as opposing equity initiatives.

A board member tries to suggest disparities in the program might reflect multiple factors—test preparation resources, different cultural emphases on academic competition, prior educational opportunities—but her comments are characterized in the local paper as “making excuses for discriminatory systems.”

The debate isn’t framed as weighing competing educational values—it’s framed as whether to eliminate a discriminatory practice. One side is defending justice; the other is defending discrimination. There’s no neutral ground where reasonable people can disagree about difficult tradeoffs.

Or consider:

Merit-based admissions: Some argue college admissions should prioritize academic achievement. Others argue they should consider race to achieve proportional representation. Both positions involve tradeoffs and value judgments.

Under the framework: If standardized tests produce disparate outcomes, they’re discriminatory. Supporting merit-based admissions means perpetuating discrimination.

Achievement gaps: Why do academic achievement differences exist? Multiple factors likely contribute: historical disadvantage, current barriers, school quality, family structure, individual choices. The policy response depends on which factors are emphasized.

Under the framework: Achievement gaps prove educational discrimination. Researching other contributing factors risks characterization as denying discrimination or blaming victims.

Criminal justice disparities: Why do arrest and incarceration rates differ by race? Possible factors include discriminatory policing, differential crime rates, socioeconomic variables, enforcement priorities, sentencing guidelines. Different diagnoses suggest different solutions.

Under the framework: Disparate incarceration rates prove discriminatory systems. Suggesting other factors contribute can be characterized as defending unjust outcomes.

The Democratic Problem

Democracy requires space for citizens to disagree about complex issues without such disagreement being treated as moral failure. When policy positions become evidence of bigotry, democratic deliberation becomes difficult.

The framework doesn’t leave room for morally sound disagreement about:

• Whether all disparities indicate discrimination

• What causes inequality

• How to weigh competing values

• What tradeoffs are acceptable

• Which interventions are effective

These become questions with seemingly obvious moral answers rather than matters for democratic debate.

The Asymmetry

The framework creates an asymmetric debate structure:

Progressive positions arguing for outcome-based interventions align with preventing discrimination. Conservative positions emphasizing equal treatment regardless of outcomes can be characterized as allowing discrimination to persist.

This isn’t because conservatives are necessarily wrong. It’s because within a framework where disparities equal discrimination, any position tolerating disparate outcomes can be characterized as tolerating discrimination.

Next: Why Current Solutions Can’t Work

Subscribe now

[Full contents linked here]

By the 2010s, the ideas that disparities meant discrimination and that intent doesn’t matter had achieved substantial political acceptance—perhaps nowhere more so than in conversations about systemic racism. Progressive politicians incorporated systemic racism analysis into campaign platforms. And international adoption demonstrated the framework’s spread beyond American borders.

This political mainstreaming made the framework increasingly difficult to challenge.

Presidential Endorsement

President William Jefferson Clinton’s 1995 defense of affirmative action at the National Archive s demonstrated how thoroughly outcome-based thinking had penetrated mainstream politics.

Clinton rejected an intent-based approach, stating:

There are those who say...that it should be enough to resort to the courts or the Equal Employment Opportunity Commission in cases of actual, provable, individual discrimination because there is no longer any systematic discrimination in our society.

He then cited statistical disparities as evidence: unemployment rates for African Americans twice those of whites; women earning 72% of men’s wages for comparable work; Hispanic women with college degrees earning less than white men with high school diplomas.

Clinton treated these disparities as demonstrating systematic discrimination, applying the logic that unequal outcomes indicate discriminatory systems regardless of whether individual acts of bias can be proven.

Progressive Candidates

By the 2010s, progressive political candidates regularly incorporated systemic racism frameworks into their platforms.

Senator Bernie Sanders’s 2019 campaign stated: “It’s time to bring a systemic approach to systemic racism. Structural problems require structural solutions.” He pledged to “root out institutional racism wherever it exists.”

Representative Alexandria Ocasio-Cortez applied the framework to technology: “Algorithms are still made by human beings, and those algorithms are still pegged to basic human assumptions...if you don’t fix the bias, then you are just automating the bias.” Her reasoning reflected impact-over-intent logic—if algorithms produce disparate outcomes, they’re discriminatory regardless of programmers’ intentions.

Representative Ilhan Omar introduced legislation to “address systemic racism embedded within policing,” framing police practices as systemically discriminatory based on outcome disparities.

These weren’t fringe positions. They represented mainstream progressive politics, with systemic racism analysis as standard campaign messaging.

Executive Action

The Biden administration implemented outcome-based frameworks through executive orders requiring federal agencies to conduct equity audits and assess whether programs produced equitable results across racial and ethnic groups.

These orders directed agencies to evaluate policies by their effects on different demographic groups rather than solely by their stated purposes or intent. Agencies were instructed to identify and address disparities in program outcomes.

State and local governments adopted similar approaches. Multiple jurisdictions implemented systematic processes for evaluating government actions through equity lenses, with Seattle’s Race and Social Justice Initiative serving as a notable early example.

International Adoption

The framework spread beyond American borders.

In Canada, Prime Minister Justin Trudeau’s government explicitly acknowledged “systemic discrimination,” stating that “racism and systemic discrimination disrupt lives, undermine people’s health and wellbeing, and are key drivers of inequality.”

In the United Kingdom, government reports on racial disparities focused on “structural” and “systemic racism” rather than requiring proof of intentional discrimination. UK policy discussions increasingly adopted outcome-focused analysis of racial inequality.

This international spread demonstrated the framework’s influence beyond the American legal and political context.

Why Questioning Became Difficult

Several factors made challenging the framework increasingly difficult:

Moral framing: Disparate impact thinking was associated with civil rights progress. Questioning it risked appearing hostile to racial equality.

Professional consensus: Among diversity professionals, HR departments, and academic experts, the framework was standard practice. Challenging it seemed uninformed.

Institutional investment: Organizations had built extensive systems around outcome-monitoring. Questioning the framework appeared to question these investments.

Social costs: Public skepticism about disparate impact could trigger accusations of denying racism’s existence or showing indifference to racial inequality.

Political polarization: As progressive politicians embraced systemic racism analysis, conservative skepticism became more partisan, making balanced discussion harder.

The Conceptual Expansion

Political mainstreaming also involved expanding what counted as potentially discriminatory. For instance:

• Environmental justice movements applied impact-over-intent analysis to environmental policies, documenting how facially neutral decisions produced disparate effects on minority communities.

• Healthcare discussions incorporated disparate impact thinking, with health outcome differences by race characterized as evidence of discriminatory systems regardless of whether individual providers intended discrimination.

• Criminal justice reform efforts focused on disparate outcomes in arrests, prosecution, and sentencing, with such disparities treated as indicating systemic problems even without proof of biased intent.

The Framework’s Political Reach

By 2020, disparate impact thinking influenced:

Campaign messaging among progressive candidates, with systemic racism analysis as standard political vocabulary.

Government policy through executive orders and agency directives requiring equity assessments.

International discourse as other countries adopted similar frameworks for understanding racial inequality.

Issue advocacy across domains from environment to healthcare to criminal justice, all applying outcome-based analysis.

The Populist Opening

It’s no coincidence that Donald Trump’s 2016 campaign explicitly rejected frameworks that had become institutionalized. He opposed diversity initiatives, rejected “political correctness,” and attacked institutions—universities, media, corporations—for imposing values that ordinary Americans hadn’t chosen.

His campaign rhetoric directly challenged the premises embedded in institutional practice. When he promised to “Make America Great Again,” part of that appeal was a return to frameworks his voters believed had been displaced without their consent—frameworks emphasizing colorblind treatment and individual merit over demographic outcomes.

Post-election analysis often characterized Trump’s appeal as racist backlash. But this explanation struggled with inconvenient facts: Trump won 8.8 million Obama voters, including substantial numbers in Michigan, Wisconsin, and Pennsylvania. Many of these voters had supported America’s first black president twice, then voted for a candidate who attacked the institutional frameworks adopted in the name of racial justice.

This explanation fits better because it doesn’t require morally writing off tens of millions of your fellow citizens. When major institutions converge on frameworks that treat certain positions as illegitimate, and this happens without democratic deliberation, rejecting institutional authority becomes rational for those who hold the delegitimized positions—even when the alternative candidate is deeply flawed.

What This Established

The framework had achieved substantial political legitimacy. While conservative skepticism remained, the impact-over-intent principle had become authoritative in progressive politics and accepted in significant parts of mainstream political discourse.

This political authority, combined with legal doctrine, academic validation, and institutional adoption, created conditions where the framework operated as background assumption rather than contested position. Events in 2020 would activate this infrastructure comprehensively.

Next: Why This Breaks Democracy

Subscribe now

[Full contents linked here]

The transformation from legal doctrine to institutional practice—legal authority from Griggs, expansion across civil rights domains, academic validation through research—helped set the stage for broader institutional change.

Understanding why institutions adopted elaborate outcome-monitoring systems can help explain how disparate impact thinking became embedded across American organizational life.

From Compliance to Strategy

Sociologists Erin Kelly and Frank Dobbin described a shift in the corporate world as follows:

During the 1970s, active federal enforcement of equal employment opportunity (EEO) and affirmative action (AA) law, coupled with ambiguity about the terms of compliance, stimulated employers to hire antidiscrimination specialists to fashion EEO/AA programs.

They explained how professional constituencies enabled the institutional survival of diversity practices through this strategic retheorization.

What the scholars didn’t address is that this wasn’t just organizational adaptation—by successfully reframing disparate outcomes as business problems requiring business solutions, these specialists helped cement the underlying assumption that such outcomes required correction, regardless of their causes.

The Professional Infrastructure

The National Association of Diversity Officers in Higher Education, founded in 2003, began with 30 members from 30 institutions. Within six months it had 120 members. By 2025 it had over 2,200 members.

Diversity work became a distinct professional category across corporations, government agencies, nonprofits, healthcare systems, and educational institutions. Once a substantial professional class existed whose careers depended on diversity work, organizational momentum for the framework increased.

Professional associations, academic journals, and consulting firms emerged around diversity, equity, and inclusion. This infrastructure helped expand the definition of problems requiring attention, increase the scope of interventions, and demonstrate ongoing need for expertise.

Corporate Practices

By the 2000s-2010s, corporations had implemented extensive systems:

Diversity offices dedicated to monitoring demographic outcomes and implementing interventions across organizational functions.

Demographic tracking of representation in hiring, promotion, performance reviews, compensation, and retention, with regular reporting to leadership.

Diversity training programs, initially focused on awareness and later incorporating unconscious bias training following IAT research.

Supplier diversity programs analyzing supply chains to ensure proportional representation of minority and women-owned businesses.

Executive compensation tied to diversity metrics, with bonuses linked to achieving demographic targets rather than just supporting diversity initiatives.

These practices represented significant organizational investment and showed genuine institutional commitment beyond minimum legal requirements.

University Implementation

Higher education adopted similar systems. By the 2010s, major universities had:

Bias response teams to investigate reported incidents, with some policies explicitly stating that intent wasn’t required for a finding of bias.

Diversity statements required in hiring and promotion, with candidates evaluated on their commitment to and plans for advancing diversity.

Climate surveys regularly assessing whether students from different groups felt equally welcomed and supported.

Resource allocation decisions incorporating equity considerations, with programs evaluated partly on their success in achieving proportional outcomes.

Government Adoption

Federal, state, and local governments implemented outcome-based frameworks:

Equity audits evaluating whether programs produced equal results across demographic groups.

Equity impact assessments required for proposed policies in some jurisdictions.

Systematic outcome monitoring across agencies, with disparities triggering investigation even without complaints of discrimination. Seattle’s Race and Social Justice Initiative, established in 2004 and expanded through the 2010s, exemplified this approach by requiring all city departments to evaluate policies for disparate effects regardless of intent.

ESG and Financial Incentives

Environmental, Social, and Governance investing integrated diversity metrics into the “Social” criteria used to evaluate companies. This meant DEI performance affected access to capital and institutional investment.

Poor diversity metrics could impact stock prices, limit access to certain investment funds, and affect corporate reputation with major investors. This helped create financial incentives for outcome-focused diversity work beyond legal compliance or moral commitment.

The Growing Disconnect

By 2015, polling revealed a significant gap between institutional practice and public opinion. Pew Research found that 59% of Americans believed the country needed to continue making changes to give blacks equal rights with whites. At the same time, in 2019, they found that 73% of Americans said race or ethnicity should not be a factor in college admissions. Gallup polling has consistently found that around 67-70% of Americans believe admissions decisions should be based solely on merit, even if it results in few minority students being admitted.

Yet institutions were moving in the opposite direction. McKinsey estimated the global DEI market at $7.5 billion in 2020, projected to more than double to $15.4 billion by 2026. Universities expanded race-conscious admissions and hired Chief Diversity Officers at accelerating rates.

The divergence was particularly stark by education levels. Among Americans with postgraduate degrees, 71% believed the country needed to continue making changes for racial equality, compared to 54% of those with high school education or less.

The Experience This Created

For Americans who believed in colorblind meritocracy—a position that remained majority or near-majority in polling—institutional convergence created a distinctive experience:

Their children’s schools taught that colorblindness perpetuates racism. Their employers required training on unconscious bias. Their local government adopted equity frameworks. The media characterized their views as outdated or harmful. Institutions they encountered every day operated from premises they hadn’t consented to.

This wasn’t coordinated imposition—each institution made independent decisions based on what leadership believed was right. But the uniformity of those decisions across sectors easily made it appear coordinated. And the absence of any democratic process—no referenda, no legislative debates, no public votes—made it feel imposed.

Why Institutions Adopted These Practices

Despite this gap between public opinion and institutional practice, organizations continued expanding outcome-monitoring systems. Several factors drove voluntary adoption:

Legal defensibility: While specific practices weren’t mandated, outcome-monitoring could be defended as following principles established in Griggs and subsequent cases. Organizations could claim they were simply complying with judicial precedent.

Moral authority: The framework had backing from the Supreme Court, academic research, and civil rights principles. Adopting it signaled institutional commitment to justice and positioned organizations on what leadership believed was the right side of history.

Risk mitigation: Proactively addressing disparities reduced risk of lawsuits, federal investigations, and public criticism. It was safer to adopt comprehensive systems than to wait for problems to emerge.

Competitive pressure: Once some organizations adopted comprehensive practices, others faced pressure to match them to compete for talent and customers. No major institution wanted to be characterized as lagging on diversity.

Reputational value: Public diversity commitments enhanced brand image, particularly with younger consumers and employees who expected such initiatives. ESG investing made diversity performance financially material to stock valuations.

Professional infrastructure: Once diversity officers, HR specialists, and consultants existed as a professional class with significant institutional presence, they had influence to expand programs and demonstrate ongoing need for their expertise.

Genuine conviction: Many executives, HR professionals, and administrators genuinely believed in the framework and wanted to address what they saw as a serious moral problem. This wasn’t cynical compliance—organizational leadership believed it was the right thing to do.

The result: widespread voluntary adoption that didn’t require government mandate. Organizations built comprehensive systems because those making institutional decisions believed it was necessary, even when substantial portions of the public held different views about the appropriate approach to racial equality.

The Scale of Adoption

By 2025, data showed widespread implementation:

The World Economic Forum’s Future of Jobs report found 83% of surveyed employers had implemented DEI measures, up from 67% two years earlier. Fifty-one percent planned comprehensive DEI training; 42% planned to set diversity goals, targets, or quotas.

Pew Research found 56% of employees believed focusing on DEI at work was positive.

Grant-making organizations, professional associations, and academic journals incorporated diversity considerations into their operations. The National Science Foundation, Department of Energy, and private foundations required diversity plans from applicants. Professional accreditation bodies included diversity standards.

What This Created

The institutional adoption transformed the disparate impact framework from legal doctrine into operational reality. Organizations had built extensive apparatus for:

• Monitoring demographic outcomes across all functions.

• Identifying disparities requiring explanation. Implementing interventions to eliminate unequal results.

• Training staff on unconscious bias and microaggressions.

• Tying compensation to diversity metrics.

• Evaluating policies by their effects on different groups.

This infrastructure meant organizations were prepared to respond when events triggered heightened attention to racial inequality. The comprehensive systems were already in place, ready for activation.

Next: Political Mainstreaming: An Unchallengeable Framework

Subscribe now

[Full contents linked here]

By the 1990s, disparate impact was established law across multiple domains. But an intellectual challenge remained: how could discrimination occur without actual people being discriminators? How could neutral policies implemented by people who rejected prejudice produce racist outcomes?

Academic research provided answers. Multiple frameworks created comprehensive intellectual justification for outcome-based evaluation of discrimination.

The Mechanism Problem

Early theories of institutional racism faced a challenge: identifying the causal pathway through which discrimination occurred.

One answer was historical legacy. Past discriminatory policies—housing discrimination, educational segregation, labor market exclusion—created cumulative disadvantages that persisted across generations. This explained ongoing disparities as resulting from past injustice even when current policies were neutral.

But this positioned discrimination primarily as a historical problem. For those observing continuing racial inequality, it seemed implausible that racist practices had simply ended with civil rights legislation.

A second answer was hidden conscious bias. Some people still held racist views but concealed them due to changing social norms. This explained ongoing discrimination but still required proving intent, which remained difficult.

This no longer mattered once scholars identified a mechanism showing discrimination occurring in real-time, through normal institutional processes, by people who genuinely rejected prejudice.

Unconscious Bias Research

In the late 1990s, psychologists Anthony Greenwald, Mahzarin Banaji, and colleagues published research on the Implicit Association Test. The IAT purported to measure unconscious associations, including negative racial associations that contradicted conscious beliefs.

The research suggested that people could consciously reject racism while harboring unconscious biases that influenced their decisions. If virtually everyone carried such biases, discriminatory outcomes would naturally emerge from normal decision-making in hiring, lending, teaching, and other domains.

This provided a missing mechanism for theories of systemic, institutional, and structural racism. Institutions could produce racist outcomes through the aggregation of unconscious biases across thousands of individual decisions—even when no one intended to discriminate.

To be sure, the research faced methodological critiques. Some studies questioned whether the IAT actually predicted discriminatory behavior or simply measured cultural familiarity with stereotypes. But, regardless of these debates, the research offered a psychologically plausible explanation for how discrimination could occur without conscious discriminatory intent.

Stereotype Threat

Claude Steele’s research on stereotype threat demonstrated another potential mechanism. His studies showed that awareness of negative stereotypes about one’s group—such as black students taking academic tests or women taking math exams—could impair performance by creating anxiety and cognitive load.

This suggested that discriminatory outcomes could result from the mere existence of stereotypes in the cultural environment, regardless of whether any individual held or acted on them. Again, even institutions with no biased decision-makers could produce unequal results if the evaluative context triggered stereotype threat.

The theory rested on a critical assumption: that performance disparities exist primarily or entirely because stereotypes create anxiety. If some performance differences existed independent of stereotype threat, eliminating it—a formidable challenge regardless—wouldn’t eliminate disparities. But questioning whether stereotype threat explained all observed gaps risked being characterized as endorsing the stereotypes themselves—making the theory difficult to test or challenge.

Colorblind Racism

Relatedly, Eduardo Bonilla-Silva’s work on “colorblind racism” provided another potential mechanism. He pointed out that claiming not to see race, or insisting race shouldn’t matter in decisions, could itself perpetuate racial inequality. If systemic processes produced disparate outcomes, refusing to acknowledge race allowed those disparities to continue.

Neutrality wasn’t neutral because colorblind policies that didn’t actively address disparate outcomes were maintaining racial hierarchy. The middle ground between actively fighting disparities and being complicit in racism was eliminated.

This framework foreclosed a position many Americans held: that treating people without regard to race was itself the goal of civil rights, not a way of perpetuating discrimination. Colorblindness—once considered an ethical ideal—became recharacterized as a mechanism of oppression. There was no principled space left for believing racial distinctions should be minimized rather than emphasized in pursuit of proportional outcomes.

Microaggressions

Still another mechanism was microaggressions. Derald Wing Sue’s research on the topic extended impact-over-intent logic to everyday interactions. The concept described subtle comments—even those meant positively—that could be experienced as demeaning by members of marginalized groups.

Examples included asking Asian Americans “where are you really from?” or expressing surprise that a black person was “articulate.” The speaker’s intent didn’t matter; what mattered was how such comments were received.

This concept extended the framework from institutional policies to interpersonal behavior. Just as facially neutral employment tests could be discriminatory based on their effects, everyday comments could be harmful based on their impact. Whether a comment actually reflected racial prejudice became irrelevant; what mattered was whether it could be received as such.

The Unfalsifiability Issue

These frameworks together created a comprehensive explanation for racial disparities. But they also created a system that was difficult to test or challenge.

If disparities indicated discrimination, but discrimination could be:

• Conscious or unconscious

• Individual or systemic

• In policies or in stereotypes

• In actions or in neutrality

• In explicit statements or in subtle comments

Then virtually any disparity could be attributed to discrimination, and virtually any behavior could be characterized as contributing to it. Failure to find evidence of one form of discrimination could always be explained by another form operating instead.

This comprehensive explanatory power made the framework intellectually appealing but also difficult to falsify.

Why Institutions Found This Compelling

The academic research gave institutions scientific justification for widespread outcome-monitoring:

Scientific legitimacy: Research from major universities, published in peer-reviewed journals, provided credible backing beyond legal requirements or moral arguments.

Clear mechanisms: Rather than vague “structural forces,” the research identified specific psychological processes that could be studied and addressed.

Universal application: If unconscious bias was universal and stereotype threat occurred in any context with cultural stereotypes, every institution needed to address these issues.

Non-accusatory framing: The research didn’t characterize discrimination as resulting from bad people with bad intentions. Well-meaning individuals could contribute to discriminatory outcomes through unconscious processes.

Actionable interventions: The research suggested concrete responses: training to address unconscious bias, policies to reduce stereotype threat, guidelines to prevent microaggressions.

A Mutually Reinforcing System

By the 2000s, multiple frameworks reinforced each other:

Legal doctrine (Griggs and extensions) made disparate outcomes violations. Systemic racism theory explained how institutions could be discriminatory without racist individuals. Unconscious bias research showed everyone harbored biases. Stereotype threat demonstrated how environments produced unequal performance. Colorblind racism explained why neutrality was problematic. Microaggressions extended the logic to everyday interactions.

Each component supported the others and made the overall framework more comprehensive.

What This Enabled

The academic validation helped create conditions for voluntary institutional adoption. Organizations now had:

• Legal authority from courts

• Moral authority from civil rights principles

• Scientific validation from research

• Clear mechanisms to address

• Specific interventions to implement

Institutions could build comprehensive systems for outcome-monitoring and bias-elimination, justified as following legal requirements, moral imperatives, and scientific findings simultaneously.

Next: Institutional Adoption: How Organizations Built the Apparatus

Subscribe now

[Full contents linked here]

Once the Supreme Court established that facially neutral employment practices producing disparate outcomes violated Title VII, a logical question emerged: would the same principle apply to other civil rights statutes?

The answer came quickly. Not only did disparate impact doctrine spread, but federal agencies adopted outcome-based enforcement across their domains and lower courts applied the framework beyond employment.

Education: Lau v. Nichols (1974)

One extension came just three years after Griggs. San Francisco’s public schools provided instruction only in English. Chinese-speaking students received the same facilities, teachers, textbooks, and curriculum as everyone else. But they couldn’t understand the instruction.

The Supreme Court held this violated Title VI of the Civil Rights Act, which prohibited discrimination in programs receiving federal funds. The Court wrote:

There is no equality of treatment merely by providing students with the same facilities, textbooks, teachers, and curriculum; for students who do not understand English are effectively foreclosed from any meaningful education.

This was disparate impact reasoning applied to education. The school district hadn’t treated students differently based on race or national origin. The policy was facially neutral. But the effects were unequal, and that was sufficient for a violation.

Following Lau, the Department of Health, Education, and Welfare began raising questions for schools for practices producing racial disparities in discipline, tracking, and resource allocation—even without allegations of intentional discrimination.

Housing and Lending

The Fair Housing Act of 1968 initially focused on explicit discrimination—real estate agents steering minorities to certain neighborhoods, landlords refusing to rent based on race. By the late 1970s, advocates were arguing that the Act should be interpreted to prohibit practices with disparate effects.

In Metropolitan Housing Development Corp. v. Village of Arlington Heights (1977), the Supreme Court held that constitutional Equal Protection claims required proof of discriminatory intent. But the Court suggested statutory claims under the Fair Housing Act might not.

Throughout the 1980s and 1990s, federal circuit courts explicitly recognized disparate impact claims under the Fair Housing Act. In 2015, the Supreme Court confirmed this interpretation in Texas Department of Housing v. Inclusive Communities Project.

Similar reasoning extended to lending. The Equal Credit Opportunity Act and Fair Housing Act were interpreted to prohibit lending criteria producing disparate effects on minority applicants. Facially neutral standards—credit scores, debt-to-income ratios, down payment requirements—could be challenged if they resulted in unequal approval rates.

Federal agencies, including the Department of Justice and Consumer Financial Protection Bureau, actively pursued disparate impact cases throughout the 2000s and 2010s.

The Loss of the Limiting Principle

It would be reasonable to wonder here why, when it came to the moral expansion, the business necessity carveout didn’t signal that not all neutral policies producing unequal outcomes are discriminatory.

On the contrary, the exception treated the disparate outcome itself as sufficient evidence that something was wrong—either the practice was discriminatory, or it required business justification. There was no third option where the disparity simply existed without requiring explanation or defense.

In the legal context, this meant a shift from requiring plaintiffs to prove discrimination occurred to requiring employers to prove it didn’t. In the moral context, there was no equivalent justification available for wealth gaps, educational gaps, health gaps, or any other disparity. The framework thus established a moral presumption—disparate outcomes meant discrimination—while leaving no legitimate way to defend against it.

The 1991 Codification

The most significant evidence that disparate impact had achieved comprehensive acceptance came in 1989-1991.

Wards Cove Packing Co. v. Atonio (1989) weakened disparate impact doctrine by shifting the burden of proof back to plaintiffs and loosening the business necessity standard. This should have been a major setback for the framework.

Instead, Congress responded with the Civil Rights Act of 1991, which passed with bipartisan support. The Act made disparate impact explicit statutory law, restored plaintiff-friendly burdens of proof, and strengthened the business necessity requirement.

The bipartisan support was significant. Within twenty years of Griggs, the impact-over-intent framework had become sufficiently accepted across the political spectrum that both parties voted to protect and strengthen it when the Court threatened to weaken it.

Why the Expansion Occurred

Several factors facilitated the spread beyond employment, including:

Logical consistency: If disparate outcomes indicate discrimination in hiring, it seemed arbitrary to say they don’t indicate discrimination in college admissions or mortgage lending. The principle appeared universal.

Moral momentum: The Civil Rights Movement’s moral authority was substantial. Extending civil rights protections seemed obviously right. Limiting them to employment seemed difficult to justify.

Institutional incentives: Federal agencies could expand their authority by interpreting their mandates broadly. Courts were reluctant to appear hostile to civil rights enforcement. Advocacy organizations had incentives to pursue expansive interpretations.

What This Established

The legal expansion was thorough. Disparate impact was now well established as an approach to identifying discrimination across civil rights law.

But legal doctrine alone doesn’t explain why institutions adopted comprehensive outcome-monitoring systems voluntarily. That required additional validation—academic research providing intellectual justification for the framework and mechanisms explaining how discrimination could occur without discriminators.

Next: Academic Validation - How Research Explained the Mechanisms

Subscribe now

[Full contents linked here]

The Griggs v. Duke Power Co. opinion is eleven pages long. The tone is measured. There’s no sweeping rhetoric about justice or civil rights. Yet this brief decision established a framework that would reshape how American institutions understand discrimination.

Understanding what the Court said—and what it avoided saying—can help explain how the framework achieved the authority it did.

The Key Passage

The decision’s central holding appears in a single paragraph:

We do not suggest that either the District Court or the Court of Appeals erred in examining the employer’s intent; but good intent or absence of discriminatory intent does not redeem employment procedures or testing mechanisms that operate as ‘built-in headwinds’ for minority groups and are unrelated to measuring job capability.

This language accomplished several things simultaneously. The Court acknowledged that examining intent wasn’t wrong, but declared that intent couldn’t justify practices producing disparate effects. The “built-in headwinds” metaphor suggested that discrimination could exist in structures and outcomes, independent of anyone’s motivations.

The phrase “unrelated to measuring job capability” created what became known as the business necessity defense. Employers could justify practices with disparate effects by demonstrating they were necessary for job performance. This was designed to be a limiting principle.

The Burden Shift

The decision altered how discrimination cases would be litigated.

Before Griggs, plaintiffs had to prove discrimination occurred—typically by showing purposeful differential treatment based on race. After Griggs, once plaintiffs demonstrated a practice produced disparate outcomes, the burden shifted to employers to prove business necessity.

This procedural change transformed disparate outcomes from evidence requiring explanation into presumptive violations requiring justification. The default assumption changed: practices with unequal effects were now suspect unless proven otherwise.

What the Court Didn’t Say

The opinion’s omissions are as significant as its holdings.

On intent: While the Court said “good intent” doesn’t redeem discriminatory procedures, it never directly accused Duke Power of bad intent. The decision could have been written as a case about a company obviously trying to circumvent the Civil Rights Act. Instead, it established a principle applicable regardless of motive.

This made the framework both broader and harder to contest. If it applied even when employers claimed to have legitimate reasons for their requirements, it applied universally.

On scope: The decision addressed employment testing under Title VII. It didn’t discuss whether the same logic should apply to education, housing, lending, or other domains where disparate outcomes exist. The Court treated this as a narrow ruling about one statute.

But the reasoning wasn’t narrow. If disparate outcomes violate Title VII regardless of intent, the principle would seem to apply wherever discrimination is prohibited.

On tradeoffs: The decision didn’t acknowledge that eliminating the intent requirement involved costs as well as benefits. Requiring proof of intent lets some discrimination escape (false negatives). Eliminating the intent requirement treats some non-discriminatory practices as violations (false positives).

The Unanimity

The decision was 9-0. Every justice agreed.

This unanimity conferred significant authority. A divided court suggests difficult questions and competing principles. A unanimous court suggests the answer is obvious and settled.

The unanimity likely reflected the specific facts of Duke Power—a company with a history of segregation implementing new tests the day after the Civil Rights Act took effect, with dramatically different pass rates by race. But the principle the Court established extended beyond cases with a clear villain.

Legal vs. Moral Authority

Griggs did more than create a legal standard. It gave institutional legitimacy to a particular understanding of discrimination.

Before 1971, viewing disparate outcomes as evidence of discrimination was one position among several (see here or here). Some scholars and activists argued for it, while others maintained that discrimination required differential treatment and intent.

The Supreme Court endorsement of the outcomes-based view didn’t just make it legally operative under Title VII—it made it morally authoritative. Institutions could adopt this framework and claim they were simply following what the nation’s highest court had determined.

The decision allowed people to effectively say: “The Supreme Court has decided that disparate outcomes constitute discrimination. Who are we to disagree?”

How Lower Courts Responded

Federal courts extended Griggs’s reasoning. In the ensuing years, disparate impact analysis appeared in cases involving teacher hiring, police department exams, union membership requirements, and seniority systems.

In 1975, the Supreme Court decided Albemarle Paper Co. v. Moody, clarifying that business necessity required demonstrating tests actually predicted job performance, not just that they seemed relevant. This raised the bar for defending practices with disparate effects.

In 1976, Washington v. Davis held that constitutional Equal Protection claims still required proof of discriminatory intent. This created a split: statutory civil rights claims (like Title VII) could proceed under disparate impact theory, while constitutional claims could not.

Perhaps this split should have prompted close examination of which approach was correct. Instead, it simply became doctrine that different legal frameworks required different showings.

What This Established

The legal foundation was comprehensive. Courts had declared that disparate outcomes constitute discrimination regardless of intent. Congress had endorsed that principle. Federal agencies had begun operationalizing it.

What remained was for institutions to build the apparatus to implement this framework. That would require additional validation—not just legal authority, but intellectual justification for why the framework made sense. Academic research would provide just that.

Next: The Legal Expansion - How Disparate Impact Spread Beyond Employment

Subscribe now

This is the table of contents for a series of nine posts I’ll release over the next 9 days (first is today, October 24, 2025) on how making disparities indicative of discrimination changed how we think about justice, equality, racism, and right and wrong.

The question this series examines is not whether disparate impact doctrine is correct in some broader sense. It’s whether a transformation this fundamental—one that redefined discrimination itself and came to govern how major institutions operate—should have happened through court decisions, agency interpretations, and voluntary institutional adoption rather than through explicit democratic deliberation about what we wanted.

I am not saying that the idea that disparities equal discrimination originated with a single legal case. On the contrary, it had been discussed and developed for years prior in academic and activist circles. I am saying that this case made disparate outcomes without intent to discriminate the legal equivalent—a violation of Title VII of the Civil Rights Act of 1964—of intentional discrimination. And that paved the way for a transformation of the moral framework we use to understand racism, identity, fairness, justice, and inequality.

The following posts are meant to be read in order but, hey, you do you.

Part 1: The 1971 Supreme Court Decision That Redefined Discrimination

Part 2: The Legal Foundation: What the Court Actually Said

Part 3: The Legal Expansion—How Disparate Impact Spread Beyond Employment

Part 4: Academic Validation: How Research Explained the Mechanisms

Part 5: Institutional Adoption: How Organizations Built the Apparatus

Part 6: Political Mainstreaming: An Unchallengeable Framework

Part 7: Why This Breaks Democracy

Part 8: Why Current Solutions Can’t Work

Part 9: What Trump’s Popularity Reveals

Full disclosure:

1. I put this series together with the help of Claude.ai. I fed it this paper, which I wrote (and will come out in the Independent Law Journal), and asked it to help break the concepts into a Substack-friendly series. I then went through and edited—and edited and edited and edited—the pieces and, well, here we are. Anyone looking for a more thorough and academic treatment should refer to the full paper.

2. This series is lightly cited. If you’re looking for more depth, I refer you to the SSRN paper.

Subscribe now

[Full contents linked here]

In March 1971, the Supreme Court decided Griggs v. Duke Power Co., a case involving employment tests at a North Carolina power company. The decision established that Title VII of the Civil Rights Act prohibited not just intentional discrimination, but also facially neutral practices that produced disparate outcomes.

This was a significant change. Before Griggs, proving discrimination generally required showing that someone treated people differently because of their race. After Griggs, practices producing unequal outcomes could violate civil rights law regardless of the employer’s intent or motivations.

The Duke Power Case

Duke Power had a history of restricting black employees to lower-paying departments. In 1955, the company began requiring high school diplomas for advancement to better positions. On July 2, 1965—the day after the Civil Rights Act took effect—the company added two employment tests as additional requirements.

The facially neutral tests had dramatically different pass rates: 58% for white employees, 6% for black employees. Willie Griggs and twelve co-workers filed suit, arguing these requirements violated Title VII.

The timing and the disparate results suggested the company was using neutral requirements to achieve discriminatory ends. But proving that intent would be difficult.

The Court faced a genuine dilemma: Duke Power had segregated black workers into lower-paying departments until 1965, then immediately implemented tests with dramatically unequal pass rates. The company had made its preferences known. And yet, if the Court’s decision hinged on this transparency, it would provide little guidance for future cases without such clear evidence of hidden intent.

What the Court Decided

The Court held that practices producing disparate effects could violate Title VII even without discriminatory intent. The decision established what came to be known as “disparate impact” doctrine. Under this framework, employers could be liable for discrimination even when:

• They applied the same standards to everyone

• They had no discriminatory purpose

• They were unaware their practices had unequal effects

• They genuinely believed their requirements were fair

In this sense, what mattered was the outcome, not the reasoning behind them.

The Court did include a limitation: employers could defend practices with disparate effects by demonstrating “business necessity”—showing the requirements were related to job performance. But this ultimately had little effect on the spread of the underlying moral principle.

The Definitional Shift

Griggs represented a fundamental change in how discrimination was defined under federal law.

Traditional definition under the 14th Amendment: Discrimination meant treating people differently because of their race. It required differential treatment and discriminatory intent.

Post-Griggs definition under Title VII of the Civil Rights Act of 1964: Discrimination included any practice producing unequal racial outcomes, regardless of whether anyone intended to discriminate.

In the Duke Power decision, the Court equated these two phenomena—intentional discrimination and neutral practices with disparate effects—as violations of the same statute. This meant facially neutral policies with unequal results weren’t just problematic; they were legally indistinguishable from deliberate discrimination and, ultimately, racism.

Why This Mattered

The Griggs framework solved a genuine problem: as explicit racism became illegal and socially unacceptable, discriminatory intent became harder to prove. The disparate impact standard allowed courts to address practices that might be hiding discriminatory motives behind neutral language.

But the framework also created a significant change in how wrongdoing was evaluated. Making intent irrelevant meant:

• False negatives decreased (fewer instances of hidden discrimination going unpunished)

• False positives increased (some neutral practices treated as discriminatory)

• The burden shifted from proving discrimination to proving its absence

The Framework’s Logical Extension

The disparate impact principle used an internal logic that made it difficult to contain within employment law. If neutral practices producing disparate outcomes violate Title VII, the same reasoning would seem to apply to:

• Educational requirements with different pass rates (as with Duke Power)

• Lending criteria with unequal approval rates

• Housing policies with segregating effects

• Any practice producing statistical disparities

The Court didn’t explicitly address these implications. The decision focused narrowly on employment testing under Title VII. But the logical extension was apparent, and over subsequent decades, the framework would spread to education, housing, lending, as well as evaluations of everyday interactions.

What the Court Didn’t Address

By omitting any mention of tradeoffs, it appears that the switch from false negatives to false positives is unproblematic. That meant there was no discussion of:

• Whether all disparate outcomes should be treated as discrimination

• What happens when causes of disparities are complex and contested

• Whether this principle should govern domains beyond employment

• What it means for legitimate disagreement about inequality

The framework was established without explicit consideration of its downstream implications.

The Transformation That Followed

What began as a legal doctrine about employment testing became the template for understanding discrimination across American institutional life. Over the following decades:

• Federal courts applied disparate impact to education, housing, and lending.

• Academic researchers developed theories of systemic racism that explained how discrimination could occur without discriminators.

• Psychological research provided mechanisms—unconscious bias, stereotype threat—showing how well-intentioned individuals could produce discriminatory outcomes.

Corporations built diversity offices in part to monitor demographic outcomes. Universities created bias response teams (like this one) to investigate harmful speech regardless of intent. Government agencies implemented equity audits to evaluate whether policies produced proportional results.

Each of these developments had roots in the idea that disparate outcomes indicate discrimination and intent doesn’t matter.

The Court’s decision solved one problem—catching discrimination hiding behind neutral language—but created another: if disparate outcomes indicate discrimination regardless of cause, then every statistical disparity becomes presumptive wrongdoing. False negatives decreased, but false positives increased.

The Question We Face

Fifty-four years after Griggs, the impact-over-intent framework has changed how most major American institutions evaluate discrimination. The question is not whether the Court was wrong to rule against Duke Power—the company’s history and the suspicious timing suggested they were likely hiding discriminatory motives.

The question is whether we adequately considered what it would mean to redefine discrimination itself. What it would mean to make intent irrelevant to moral evaluation. Whether treating statistical disparities as evidence of wrongdoing—regardless of their causes—is compatible with preserving space for legitimate disagreement about complex social problems.

This transformation happened largely without democratic deliberation. It occurred through court decisions, agency interpretations, and voluntary institutional adoption. The conversation about what it might mean for this framework to govern American institutional life never took place.

Understanding how we arrived at our current situation requires tracing how a legal doctrine about employment tests became a comprehensive moral framework. That’s what the following posts examine.

Next: The Legal Foundation: What the Court Actually Said

Subscribe now