A Fact to a Tribe

Introduction

There was a time, not long ago and not as long ago as we pretend, when a fact was understood to be a fact whether or not you liked it. A fact was a feature of the world. It sat there, indifferent to your wishes, and the most you could do was discover it, deny it, or be ruined by it — but you could not make it belong to you, and you certainly could not make it belong to your side. The signature condition of our moment is that this is no longer how a great many people experience truth. A fact has become a thing that is true for a tribe. It has acquired ownership. And once a fact can be owned, it can be defended, disowned, conscripted, and abandoned according to whether it serves the group that holds it — which is to say it has stopped being a fact in the old sense and become something else, something closer to a flag.

This is a book about how that happened, what mechanism makes it run, and why the people best equipped to resist it are, on the evidence, the ones most efficiently captured by it. I am going to argue all of this forthrightly, as my own contention, and I want to state the central claim now, plainly, before I have shown you a single piece of evidence for it, so that you can watch me try to earn it.

My contention is this: under the conditions of digital life, large numbers of us have stopped adjudicating reality for ourselves and begun relocating that work to the group. We no longer ask, first, what is true here, what actually happened, what did this person actually do — and then decide where we stand. We ask, first, which side am I on — and let the answer to that question determine what we will treat as true. The sequence has inverted. Loyalty now precedes adjudication and governs it. Evidence that flatters the tribe is waved through unexamined; evidence that wounds it is not weighed and rejected but declined, treated as the hostile noise of a rival camp. A fact, in this regime, is simply a claim our side has agreed to hold. The other side's facts are propaganda. Our propaganda is fact. And almost no one inside the arrangement can feel it happening, because it does not feel like surrender. It feels like conviction.

Relocated deference

I have a name for the mechanism, and because the rest of the book leans on it, I will define it carefully here and use it consistently from this point forward.

I call it relocated deference: the transfer of one's own moral and evidentiary judgment away from oneself and onto an external authority that is then permitted to do one's reasoning. The phrase is built on a debt I should pay immediately. In the 1960s Stanley Milgram demonstrated that ordinary people, instructed by a credible authority, would deliver what they believed were lethal electric shocks to an innocent man — not because they were cruel, but because they had entered what Milgram called the agentic state: a condition in which a person stops experiencing himself as the author of his own conduct and begins experiencing himself as the instrument of an authority that has, in effect, taken his conscience into safekeeping. He still owns his hands. He no longer owns the decision. The decision has been relocated.

A later generation of psychologists, chiefly Alex Haslam and Stephen Reicher, corrected the part of Milgram's picture that was always too flattering to us. Milgram's subjects were not, in the main, anguished automatons cringing as they obeyed. The ones who went furthest were the ones who believed — who had identified with the experimenter's project until its demands felt like their own convictions. They did not capitulate to authority. They enacted it, willingly, as engaged followers. This matters enormously, because it means the deference we are discussing is not weakness in the ordinary sense. It is devotion. It feels, from the inside, like integrity.

What I am adding to this inheritance is a claim about direction, and it is the reason the older vocabulary needs a new term. Milgram's subject relocated his judgment upward — to an authority with an office, a name, an address: the man in the lab coat, the institution behind him, a figure who could, in principle, be petitioned, resisted, or held to account. Relocated deference in the digital age runs sideways and downward — to the peer-swarm, the tribe, the cause, the anonymous collective that has no office, no credentials, and no possibility of being held to account in turn. Milgram's man could at least say I was following orders and point to the one who gave them. The relocated-deferent says we know what really happened and points to no one, because the we is structurally anonymous — a fog with the authority of a sovereign and the accountability of a weather system. That is the dark innovation of the present moment. We have learned to obey an authority that cannot be located, which means we have learned to obey while believing ourselves free, and to surrender our judgment while experiencing the surrender as the exercise of it.

Why intelligence is no escape

If this book had a single intended reader, it would be the one now thinking that all of this describes other people — the credulous, the partisan, the poorly educated, the mob — but not himself, because he reasons, he reads, he weighs evidence, he is precisely the sort of person such a mechanism cannot touch.

That reader is the one I am most concerned about, and I am concerned about him for an empirical reason, not a rhetorical one. The evidence does not support the comforting theory that intelligence inoculates against tribal reasoning. It supports the opposite. The work of Dan Kahan and colleagues on what they call identity-protective cognition found that greater analytical ability — more numeracy, more cognitive reflection — does not reduce motivated reasoning on identity-charged questions. It amplifies it. The most capable reasoners did not arrive at the most accurate conclusions; they arrived, more reliably than anyone, at the conclusions their group required, and they did so by turning their considerable faculties to the task of getting there. Intelligence, in these studies, did not function as a brake on tribal deference. It functioned as a superior engine for it. This is not a marginal finding and it is not, I will grant, entirely uncontested — but its direction is robust, and its implication is brutal: the smarter you are, the better you are at constructing the rationalization that lets you defer to your tribe while believing you reasoned your way to independence.

This is why the book is pitched where it is. It is not written to rescue the credulous from obvious lies. It is written to confront the analytically capable with the uncomfortable likelihood that their sophistication is not the cure for relocated deference but its most refined delivery system — that the articulate defender of a tribal conclusion is not the exception to the mob but its author, supplying the polished rationalizations that the less articulate then recite. The cascade runs downhill from the clever, not up toward them.

What the book does

The argument proceeds in stages. The chapters that follow this introduction build the framework: the agentic state and engaged followership; the special power of identity narratives — and identity framings turning on group membership are among the most potent — to supercharge the relocation; and the role of the digital platform — the recommender system that has discovered, without intending anything, that content inviting the surrender of individual judgment is content that performs, and so optimizes toward the manufacture of the agentic state at population scale. I should mark the analysis of identity here for exactly what it is and is not: it is a detached account of a universal mechanism that captures every side with the same indifference, because the reflex it describes is symmetric — it has no home team, takes no party's part, and makes no claim about any group's guilt or virtue. It is a fact about how human reasoning bends under loyalty, not a verdict on anyone the loyalty is aimed at.

Then the book turns to cases, because a mechanism is only as good as its visibility in the world, and I have chosen the cases to make a specific and difficult point. I examine the public response to one criminal conviction in detail, and then — deliberately, and at the cost of comfort for readers who found the first analysis congenial — the response to a verdict that ran the other way, defended and denied by the opposite tribes through the identical machinery. I examine a form of collective behavior in which there is no defendant at all, where the relocation of responsibility completes its own logic by dissolving into a crowd that no one can be held to. And I place all of it against the long historical record, which shows both that this reflex is ancient and that it has, in better-built periods, been caged — which is the book's one genuine source of hope.

Finally the book asks what the relocation costs, why it persists and intensifies, and what it would take to resist it: not more intelligence, which is its instrument, but a specific and unflattering discipline — the willingness to let the evidence carry you to a conclusion your tribe would not have chosen, and to remain there, alone if necessary, holding a fact that no longer belongs to anyone.

That discipline is the opposite of everything the present moment rewards. It is slow where the feed is fast, solitary where the tribe is warm, and uncertain where the group is sure. It asks you to own your own judgment in an age that offers, on every screen and at every hour, the standing relief of handing it away. I do not pretend it is easy, and I will not insult you by suggesting that reading a book confers it. But it begins with seeing the mechanism clearly, including — especially — in oneself. That is what the rest of these pages are for.

A fact does not belong to a tribe. It never did. The whole catastrophe of this moment is contained in how many of us have come to feel that it does.

CHAPTER 1

The Agentic State and Its Heir

The Introduction made a promise it did not keep: it named relocated deference and gestured at its lineage, but it did not earn the concept. It asserted that the mechanism is real, ancient, and newly weaponized, and asked you to take the assertion on credit. This chapter pays the debt. It builds the framework slowly and from the ground, because everything the rest of the book does to the cases depends on the machinery being sound, and a reader is entitled to inspect the machine before he is asked to watch it cut.

I will build it in three moves. First, Milgram, read more carefully than he is usually read. Second, the correction that Haslam and Reicher made to him, which is not a footnote but a reversal of his central image of human nature. Third, the thing I am adding — the claim about direction and accountability that turns the agentic state into relocated deference, and that explains why a mechanism documented in a quiet laboratory more than six decades ago should have become, in our own time, the organizing pathology of public life.

Milgram, read correctly

The Milgram experiments are the most famous in the history of social psychology and among the most misremembered. The popular memory is roughly this: Milgram proved that people are sheep, that anyone will torture a stranger if a man in a coat tells them to, that the veneer of civilization is paper-thin. This is not wrong so much as it is lazy, and the laziness costs us the actual finding.

Here is the design, stripped to essentials. A volunteer is told he is participating in a study of learning and memory. He is assigned the role of "teacher." Another man — actually a confederate, in on the experiment — is the "learner," strapped to a chair in an adjacent room with an electrode on his arm. The teacher is to read word pairs and, each time the learner errs, administer an electric shock, escalating the voltage with every mistake along a console that runs from a mild low setting to a final, ominous maximum, the highest switches labeled with stark warnings and then simply XXX. The learner, by design, begins to err, then to protest, then to cry out, then to complain of a heart condition, then — at the upper reaches — to fall silent. There is no actual shock. The only real subject is the teacher, and the only real question is how far up the console he will go while a calm experimenter, seated nearby, insists across four scripted prods that the experiment requires that he continue.

The result that made Milgram famous: roughly two-thirds of subjects went all the way to the maximum. Not sadists, not the damaged, not a selected population of the cruel — ordinary men, recruited by newspaper ad, the median of the human range. Most of them did not want to continue. They sweated, they laughed nervously, they begged the experimenter to check on the learner, they argued. And then, very often, they continued anyway.

But notice what the careful version of the finding actually says, because it is not people are sheep. It is something stranger and more specific. Milgram's subjects did not stop caring about the learner; their distress proves they cared throughout. What changed, as they moved up the console, was not their compassion but their sense of who was responsible for what was happening. They came to experience themselves as no longer the author of the act. The act had an author — the experimenter, the institution, science itself — and the subject had become merely the hand through which that author worked. This is what Milgram named the agentic state: the shift from seeing oneself as an autonomous moral agent, who owns his choices, to seeing oneself as an agent in the other sense — an instrument, an extension, a deputy executing a will located somewhere above him. The voltage did not measure cruelty. It measured how completely a person could relocate the ownership of his own conduct onto an authority while continuing, with his own hand, to perform it.

The single most revealing variation in the entire body of Milgram's work supports this reading exactly. When the experimenter gave his orders by telephone rather than in person, obedience collapsed — many subjects defied him, and some pretended to administer shocks they did not give. When a second "teacher" (another confederate) refused and walked out, obedience collapsed. When two experimenters argued with each other about whether to proceed, obedience collapsed. In every case, what broke the agentic state was anything that returned responsibility to the subject — that made it harder to believe the decision belonged to someone else. The mechanism was never about the strength of authority's command. It was about the availability of a place to put the responsibility. Remove the place, and the conduct stops. That is the hinge on which this entire book turns, and Milgram found it more than sixty years ago.

The correction: we are not reluctant. We volunteer.

For decades the Milgram lesson hardened into a doctrine of situationism: that ordinary people are passive before situational pressure, that the context overwhelms the character, that we are, in the face of authority, essentially helpless. It is a doctrine with a strange moral comfort built into it, and the comfort is worth naming because it is itself a small act of relocated deference. If the situation is overwhelming, then the obedient subject is not really culpable; he was made to do it. The situationist reading lets the perpetrator keep his conscience by handing his agency to the circumstance.

Alex Haslam and Stephen Reicher spent years dismantling this, and their correction is the load-bearing one for everything that follows. Returning to Milgram's own records and running new work of their own, they found that the obedient subjects were not, in the main, passive instruments crushed by situational force. The ones who went furthest were disproportionately the ones who had come to identify with the enterprise — who believed in the experiment, valued the contribution to science, trusted the experimenter as a legitimate leader of a worthwhile collective project, and therefore followed not despite themselves but as committed participants. Haslam and Reicher called this engaged followership. The deepest obedience was not the dull compliance of the broken will. It was the active, willing, even zealous cooperation of someone who had fused his own identity with the cause and now experienced its demands as his own conscience speaking.

This reverses Milgram's residual image of human nature, and the reversal is the thing to hold onto. The danger is not, primarily, that we are weak in the face of authority. The danger is that we are devoted — that we are built to attach ourselves to groups and causes and leaders, to take their projects into our sense of who we are, and then to act on their behalf with all the energy of personal conviction, precisely because it has become personal conviction. The followers who do the most damage are not the ones who feel the least. They are the ones who believe the most. They are not capitulating. They are enacting. And from the inside, enactment does not feel like obedience at all. It feels like integrity, like loyalty, like courage, like being the kind of person who stands for something. This is why exhortations to "think for yourself" so reliably fail against it: the engaged follower is certain he already is.

From agentic state to relocated deference

Now I can state precisely what I am adding, and why the inherited vocabulary is insufficient to the present moment.

Milgram and Haslam-Reicher together give us a person who relocates the ownership of his conduct onto an authority he has identified with. But in their settings that authority has a crucial property: it is locatable. The experimenter stands in the room. The institution has a name. The leader can be approached, argued with, defied, or — after the fact — held responsible. The entire drama unfolds along a vertical axis, the subject looking up to an authority that occupies a definite position above him. And because the authority is locatable, the responsibility, in principle, can be traced back to it. The defense offered at the great postwar obedience trials — I was following orders — is morally bankrupt, but it is at least structurally coherent: it points at a real someone who gave a real order.

The relocation that organizes our moment has lost this property, and the loss changes everything. The authority to which the digital citizen defers is not above him and is not locatable. It is the tribe, the swarm, the cause, the consensus of his feed — a collective with no office, no spokesman with standing to bind it, no address at which it could be served. I call the result relocated deference to mark the difference from the classical agentic state: the judgment is still surrendered, the conduct is still performed, the conviction is still felt as one's own — but there is no longer any authority that can be located as the bearer of the responsibility that has been handed away. It has not been transferred to someone. It has been diffused into no one.

Consider what this does to accountability, because accountability is the whole stake. In Milgram's basement, responsibility was conserved even as it was relocated: it moved from the subject to the experimenter, but it remained somewhere, attached to a person, available for judgment. Relocated deference does not conserve responsibility. It destroys it. When ten thousand people defer their judgment to "what everyone knows," to "the side of justice," to the felt consensus of the tribe, and each acts on a conviction he experiences as his own but did not himself form, there is no one at the end of the chain to hold. Ask any single participant and he will tell you, truthfully, that he was only agreeing with what was obviously true — pointing not to an order from above but to a fog all around. The fog issued no order. The fog cannot be subpoenaed. And yet the fog moved ten thousand hands. We have built, for the first time at scale, a structure in which people can surrender their moral agency to an authority that does not exist as an entity, which means they can act with the full force of collective will while leaving no one — literally no one — accountable for the result.

This is why the digital age did not merely accelerate an old problem but qualitatively changed it. The agentic state, in Milgram's form, at least left a chain of command that justice could walk back up. Relocated deference dissolves the chain into a crowd, and a crowd has no top. The conscience that has been outsourced is the one that used to ask, before loyalty and before the group, the only question that ever protected anyone from any of this: what is actually true here, and what, knowing it, will I myself choose to do? That question is the first casualty of relocated deference. Every case in this book is an autopsy of its absence.

What the framework predicts

A framework earns its keep by predicting what we then go out and find, so let me close by stating, in advance, what relocated deference predicts about the behavior we will examine in the chapters ahead.

It predicts that public response to charged events will track group identity rather than evidence — that people will know where they stand before they know what happened, and that incoming facts will be sorted by whether they serve the prior allegiance. Here the framework draws directly on Dan Kahan's work on identity-protective cognition and motivated numeracy: Kahan's experiments show that when a factual question becomes a marker of group membership, greater reasoning ability does not move people toward the right answer but is recruited to defend the group's answer, and this effect appears symmetrically across opposed groups. That symmetry matters, and it must be stated plainly to avoid misreading: the point is not that one side reasons badly and another well. The point is that the mechanism is indifferent to content — it captures whoever is most fused with whatever group has staked itself to a given claim, and it does so on every side at once. It predicts that the predictor of how strongly someone holds a position will not be how much of the evidence they have examined but how fused they are with the relevant group. It predicts that challenges to an unwelcome conclusion will attack the legitimacy of whoever produced it — the jury, the institution, the source — on grounds that require no engagement with the actual reasoning, because engaging the reasoning would mean leaving the agentic state and re-entering the exhausting condition of owning one's own judgment. It predicts that the most analytically capable adherents will produce the most elaborate justifications, not the most accurate assessments. And it predicts that the same operation will appear on every side of every sufficiently charged event, because the mechanism is a property of the human disposition to defer, not of any particular tribe's content. The reflex has no home team.

Every one of these predictions is testable against real cases. The rest of the book is the test.

CHAPTER 2

The Supercharger — Why Identity Narratives Do the Most Work

If relocated deference were equally easy to trigger with any content, this book could end at Chapter 1. It is not. The mechanism has fuels, and the fuels are not interchangeable. Some claims slide a person into the agentic state almost instantly and hold him there against any amount of contrary evidence; others, structurally identical as claims, fail to move him at all. The difference is not the strength of the evidence. It is whether the claim is wired to an identity. This chapter is about why identity narratives — and the ones tied to deep, ascribed group categories most of all — are the highest-octane fuel the mechanism has, and why a charged identity claim can override in seconds a body of evidence that would, on any neutral question, have settled the matter.

The thing a fact threatens

Begin with a question that sounds naive and is not: why would anyone refuse a fact?

On most questions, no one does. Tell a man the bridge is out and he takes the other road; he does not convene his tribe to relitigate the bridge. The overwhelming majority of facts pass into us without resistance because nothing about us depends on their being false. Resistance appears only when a fact threatens something the person needs more than he needs to be right. And the thing most often standing behind that need — the thing a fact can threaten so severely that an otherwise reasonable person will spend enormous cognitive effort to keep it out — is identity. Not life, not safety, not money. Identity. Who I am, and, inseparably, who we are.

This is the insight that the psychology of motivated reasoning circled for decades and that Dan Kahan's work on identity-protective cognition finally stated cleanly. People do not process identity-threatening information the way they process ordinary information, because for the identity-threatening kind a different and older priority is running: the protection of one's standing in the group whose regard one cannot afford to lose. We are an intensely social species whose ancestors did not survive alone, and for whom expulsion from the group was, for almost all of our evolutionary history, a death sentence. The machinery that guards our place in the group is therefore not a minor subsystem. It is close to the core, and it will, when triggered, commandeer the reasoning faculties and put them to work defending the group's position — while presenting the whole operation to consciousness as ordinary, sincere, evidence-driven thought. This is the supercharger. A fact wired to identity is not evaluated. It is defended or attacked, by reflexes that evolved to keep us from being cast out, wearing the costume of rational judgment.

Why an ascribed identity is the highest-octane fuel of all

Of all the identities a narrative can be wired to, the ones tied to a deep, ascribed group category are among the most combustible, and it is worth being precise about why rather than treating it as obvious, because the reasons explain the behavior we will see later. Nothing in what follows is a claim about any particular group; it is a description of a mechanism that captures every group equally, including whichever one the reader belongs to.

Such an identity is, first, visible and ascribed — assigned at a glance, unchosen, difficult or impossible to exit. You can leave a political party or a church; you cannot resign the category you were born into. An identity you cannot exit is one whose defense feels existential, because there is no door. Second, identities of this kind in many societies sit on top of a real and bitterly contested history, which means narratives built on them arrive pre-loaded with a vast inventory of prior grievance and prior solidarity ready to be activated by the newest instance. A claim about a single encounter in a single public space does not land as a claim about that encounter. It lands as the latest chapter of a story already centuries long, and the listener's response is governed less by the particulars in front of him than by the chapter he has already decided the story is in. Third — and this is the part the digital age sharpened to a razor — these narratives offer the cleanest possible sorting signal. They divide a population into us and them faster and more reliably than any other frame, which makes them ideal raw material for a system that profits from sorting populations into us and them.

The consequence is that such a framing, once applied to an event, tends to determine the response before the evidence about the event is even available — and to survive the arrival of evidence that, on a neutral question, would have changed minds. The framing is not a conclusion drawn from the facts of the case. It is a key that decides, in advance, which facts will be admitted and which dismissed. When supporters of a defendant know he is innocent before the trial, and remain certain after a conviction, the certainty is not tracking the evidence; it never was. It is tracking the chapter of the long story they slotted the case into on day one. The same is true, exactly and symmetrically, of those on the other side who knew he was guilty before the trial and remained certain after an acquittal. The reflex has no home team. The identity frame is the supercharger because it converts a particular, examinable event into an instance of an unfalsifiable narrative — and an unfalsifiable narrative is, by definition, one no fact can touch.

The closed loop

What makes identity narratives so nearly impregnable is a structural feature I want to name precisely, because it recurs in every case this book examines. Call it the closed loop: the property by which a sufficiently developed identity narrative converts all incoming evidence, of whatever direction, into confirmation of itself.

Evidence that supports the narrative confirms it directly; this is unremarkable. The remarkable move is what the closed loop does with evidence that contradicts it. Contradictory evidence is not weighed and found wanting — that would be ordinary reasoning, and the loop is not reasoning. It is reinterpreted as proof of the very forces the narrative alleges. If the narrative says the system is rigged against us, then a verdict that goes against us proves the rigging — and a verdict that goes for us proves only that the rigging occasionally fails under pressure, or was a token concession, or is a trap. There is no possible verdict that the narrative cannot absorb. If the narrative says a group is dangerous and coddled, then their punishment proves the danger and their exoneration proves the coddling. Every outcome feeds the loop. Disconfirmation is not merely resisted; it is digested, metabolized into fuel, so that the narrative grows stronger in precise proportion to the evidence against it. A belief that is strengthened by its own disconfirmation has left the domain of the empirical entirely. It has become, functionally, a faith — and faiths are not refuted by facts, because facts were never what they rested on.

This is why arguing with a person inside a closed loop is so uniquely futile, and why the futility is so often misdiagnosed. We assume the person has not yet seen the decisive fact, and we supply it, and we are bewildered when it changes nothing or makes things worse. But within a closed loop the decisive fact does not function as a fact. It functions as an attack, and an attack from the rival tribe at that, which activates the very identity-protective machinery that the fact was supposed to disarm. You cannot fact your way into a closed loop, because every fact you introduce is converted, on entry, into evidence for the thing you were trying to dislodge. The loop does not have a hole you can pour truth through. That is what makes it a loop.

The role of the articulate

There is one more element, and it connects this chapter to the most uncomfortable claim in the book. Closed loops do not maintain themselves automatically; they are built and maintained by their most articulate members. The raw identity reflex — they are coming for us; defend our own — is crude and would not, by itself, survive much contact with a literate public. What gives a closed loop its durability is the work of its sophisticated adherents, who take the crude reflex and dress it in argument: the historical framing, the systemic analysis, the procedural objection, the marshaled statistic. They supply the loop with intellectual scaffolding, and in doing so they perform two services at once. They make the narrative defensible to educated members who could not stomach the crude version, and they supply those members with ready-made justifications that feel like independent reasoning. The articulate do not merely believe the narrative. They manufacture the equipment that lets everyone else believe it while feeling intelligent.

This is the hinge into Chapter 3 and, ultimately, into the cases. Identity narratives are the supercharger; the closed loop is the engine that cannot be stalled by evidence; and the articulate are the engineers who keep it running and hand its output downward to the less articulate as a script. What remains is to explain the machine that finds all of this profitable, amplifies it, and delivers it to each of us a thousand times a day in precisely the dosage most likely to pull us into the loop. That machine is the subject of the next chapter. But understand, before we turn to it, that the platform did not create any of what this chapter has described. The supercharger, the closed loop, the work of the articulate — all of it predates the feed by millennia. The platform's genius, if it can be called that, was not invention. It was discovering that the oldest and most dangerous tendency in human social cognition was also, when properly fed, the most engaging.

CHAPTER 3

The Machine That Farms the Reflex

The previous two chapters described something old. The agentic state, engaged followership, the identity supercharger, the closed loop — none of it is new, and a careful reader might reasonably ask why, if all of this has been with us for millennia, the present moment feels like a discontinuity rather than more of the same. The answer is the subject of this chapter. The reflex is ancient. The machine that has been built to find, feed, and farm it at population scale is not. We did not become more tribal than our ancestors. We built, for the first time, a system whose commercial incentive is to locate the tribal reflex in each of us and pull it, gently and continuously, several hundred times a day.

I want to be careful here, because the topic invites two errors and the truth lies between them. The first error is technological innocence: the claim that platforms are neutral pipes, that they merely show us what we choose, and that any pathology is entirely our own. The second is technological determinism: the claim that the algorithm is a mind-control ray, that we are helpless before it, that it manufactures our beliefs from nothing. Both are wrong, and the research now available lets us say precisely how. The platform does not implant the reflex; the reflex was already there, the inheritance described in Chapters 1 and 2. But the platform is very far from neutral. It systematically selects for the content that triggers the reflex, rewards us for performing it, and distorts our perception of everyone else in the direction that makes the reflex feel normal. It is not a mind-control ray. It is something subtler and, at scale, more effective: a machine that has discovered our oldest weakness is also our most reliable form of engagement, and that optimizes accordingly.

What the machine selects for

Start with the selection. A recommender system tuned to maximize engagement — time on platform, clicks, shares, replies — is not choosing content for its truth, its value, or its contribution to your understanding. It is choosing content for its capacity to capture and hold attention, and it learns, from billions of interactions, exactly which features predict that capture.

The features it converges on are not random. Researchers studying the diffusion of content across social networks have found that the algorithm reliably privileges what one body of work labels PRIME content — prestigious, in-group, moral, and emotional. Each of these is a lever on precisely the machinery Chapter 2 described. In-group content activates identity. Moral content activates the reflex to defend the group's values. Emotional content, and one emotion above all, drives sharing: when researchers isolated the effect on political content specifically, anger was by a wide margin the emotion the engagement algorithm most amplified — both in the feelings expressed by authors and the feelings produced in readers. The machine did not set out to make us angry at each other. It set out to maximize engagement, discovered that moralized out-group anger is among the most engaging things a human can encounter, and so learned to serve us a diet unusually rich in exactly the stimulus most likely to pull us into the agentic state. The supercharger of Chapter 2 turns out to be, from the platform's point of view, premium fuel — and the platform burns what performs.

There is a finding here that ought to end the "neutral pipe" defense permanently, and it is worth stating exactly. In a controlled comparison, the political content that an engagement-based algorithm chose to amplify left users less satisfied than the political content in a simple chronological feed — even as it drove more engagement. Read that twice. The machine is not giving us what we, on reflection, want. It is giving us what reliably captures us against our reflective preference, which is a different thing and often the opposite thing. We engage with the outrage, and we are worse off for it, and we would have chosen otherwise if asked — and the system serves it anyway, because the system optimizes engagement, not satisfaction, and certainly not truth or peace. The gap between what captures us and what we would reflectively choose is the exact territory the machine has learned to farm.

The reward loop

Selection is only half of it. The platform does not merely choose outrage-triggering content to show you; it also trains you, by reward, to produce more of it.

The mechanism is ordinary reinforcement learning, the same process by which any animal learns which behaviors pay. Researchers studying moral outrage on a major social platform — across millions of posts — found that expressions of outrage that received positive social feedback, the likes and shares and approving replies, were followed by more outrage from the same user. The feedback functioned as reward, and the reward shaped future behavior exactly as reinforcement always does. We are, on these platforms, being operantly conditioned to express moral outrage, because moral outrage is what the environment pays out for, and we are an animal that learns what pays. The same research found a second layer: people also calibrate their outrage to the norms of their particular network, expressing more of it where it is more common, learning not just from their own rewards but from the ambient behavior of the group. Reinforcement learning and norm-learning together, running on an architecture designed to maximize engagement, produce a population that is being trained, continuously and without its awareness, to perform the tribal reflex more often and more intensely over time.

Notice what this does to the engaged follower of Chapter 1. He does not experience any of this as conditioning. He experiences it as caring more, as becoming more awake to injustice, as finding his voice, as discovering how much there is to be angry about. The reward loop is invisible to the rewarded, because reinforcement does not announce itself as reinforcement; it presents as preference, as conviction, as growth. The man who has been trained by ten thousand approving notifications to post ever-sharper denunciations of the out-group does not feel trained. He feels right, and increasingly so. His radicalization is delivered to him as self-discovery.

The hall of mirrors

There is a third effect, the most insidious of the three, because it corrupts not what we feel but what we believe to be true about everyone else.

Because the machine amplifies the most extreme and outraged expressions, the version of public opinion that reaches us through the feed is not a representative sample of what people actually think. It is a sample heavily skewed toward the loudest, angriest, most identity-fused voices on every side, because those are the voices the algorithm selected. And we, lacking any other window onto the vast population of strangers, take this skewed sample for the real distribution. Research has documented the consequence directly: people systematically overestimate how outraged others are, because the outraged are overrepresented in what they see. We come to believe the other tribe is more hostile, more extreme, and more numerous in its extremity than it actually is — and, learning that this is apparently the norm, we conform to it, becoming more extreme ourselves to match a hostility that was inflated by the machine before it ever reached us.

This is a feedback loop of a particularly vicious kind, because it manufactures the very reality it reports. The feed shows me an out-group that appears to hate me; I respond with hostility, which is now visible to them; the machine amplifies my hostility into their feed as evidence of how much my tribe hates them; they respond in kind; and around it goes, each side's amplified extremity becoming the other side's justification, the whole spiral powered by a perception of mutual hatred that is, in its origins, substantially an artifact of what the algorithm chose to show. We are not, most of us, as far apart as we appear to each other. But we are governed, in our responses, by the appearance — and the appearance is engineered, not for accuracy, but for engagement. A hall of mirrors does not lie about what is in the room. It lies about how much of it there is, and that lie is enough.

Why this is relocated deference's perfect habitat

Assemble the three effects and you have an environment custom-built — though not, I think, by anyone's deliberate intention — to maximize relocated deference.

The selection mechanism ensures that the content most likely to trigger the surrender of individual judgment to the tribe is the content most likely to reach you. The reward loop ensures that performing that surrender — posting the denunciation, joining the pile-on, signaling allegiance — is continuously reinforced, so that deference becomes not just easy but habitual, the trained default. And the hall of mirrors ensures that the tribe whose authority you are deferring to appears larger, angrier, and more unanimous than it is, which raises the felt cost of breaking with it and lowers the felt cost of joining it. Each effect feeds the others. The machine selects the fuel, rewards the combustion, and inflates the apparent size of the fire until standing apart from it feels like standing alone against the world.

And recall the cruelest detail from Chapter 1: what dissolves the agentic state is anything that returns responsibility to the individual — the defiant peer, the dissenting voice, the moment of being made to feel the choice is one's own. The architecture of the feed systematically removes exactly these. The dissenter is downranked for low engagement; the nuanced voice cannot compete with the outraged one for distribution; the lone objector, who in Milgram's laboratory could break the spell for everyone, is in the feed simply not shown. The machine does not merely supply the conditions that produce relocated deference. It strips away, by the same optimization, the conditions that would interrupt it. It is, considered as a whole, an apparatus for the manufacture and maintenance of the agentic state, operated for profit, at a scale and intimacy Milgram could not have imagined — running not on a few dozen volunteers in a single laboratory, but on billions of people, in their hands, all day.

The framework is now complete. We have the ancient reflex, the identity fuel that supercharges it, and the modern machine that farms it. What remains is to watch the whole apparatus operate on real events, which is where this book turns next — and to keep in view, as we do, the one finding that all the rest depends on: that none of this works on a person who has held onto his own judgment, and that holding onto it has been made, by deliberate design, as difficult as a profitable machine could arrange.

When the Verdict Stopped Mattering

Consider the logic of the punch.

In a recent incident a man rode up to a stranger on the street, accused him of having served on a jury in a trial held far away, and struck him in the head — on camera, for distribution. The accusation was false; a bystander corrected him twice; the correction altered nothing. He told the man he would die, rode off, and only then, at a safe remove, conceded the error with the indifference of someone returning a wrong number: wrong man — oh well. The caption he attached to the footage performed arithmetic, counting down the strangers of the rival category he proposed to punish, and bound the sum to a cause: the freeing of a convicted defendant he had never met.

What is instructive here is not the violence, which is banal, but the structure of the reasoning that licensed it. The victim was not selected; he was substituted. Any body would have served, provided it could be assigned to the opposing tribe, because the assailant was not responding to anything the man had done. He was responding to a verdict — and discharging that response onto a proxy chosen by category alone. The wrongness of the target was not a defect in the act but a matter of supreme indifference to it, which is why the correction, when it came, carried no weight. You cannot mistake the target of an act of justice; justice is addressed to a person. You can only mistake the target of an act of category, and the mistake is trivial, because the category remains correct even when the man does not.

I open with this not because it proves my thesis — one man with a phone and a grievance proves nothing but his own condition — but because it renders visible, in a single grotesque gesture, a logic that operates everywhere else in this kind of episode silently and at scale. The punch is not the evidence. It is the X-ray: a flash bright enough to expose the skeleton of a reasoning that, in the tens of thousands who never raised a hand, kept its skin on and passed for normal. Those people are the subject of this chapter. They did their work calmly, from home, with donations and comment threads, after twelve citizens spent days with the evidence and returned a verdict. The lone assailant only made it impossible to pretend the skeleton wasn't there.

The contention of this chapter, which I will state without softening and defend without recourse to anyone's demography, is this: a substantial and growing public has ceased to regard guilt as a fact that survives the identity of the guilty. Accountability, for them, is no longer a finding about an act; it is a status conferred or withdrawn by allegiance. The sequence has inverted. Where adjudication once preceded loyalty — we determine what happened, then decide how we feel — loyalty now precedes adjudication and governs it. Evidence concordant with allegiance is waved through unexamined; evidence discordant with it is not weighed and rejected but declined, treated as the hostile noise of a rival camp. This is not skepticism, which engages evidence in order to doubt it. It is something prior to skepticism and immune to it: the conviction that the evidence was never the relevant thing. The episode I will use as a specimen is among the cleanest examples of this inversion yet produced, in part because the internet did not merely transmit it but incentivized it — organizing the loyalty, rewarding its performance, and, as we will see, financing it past the point of conviction.

The framework is familiar, and I assume you have it. Milgram named the agentic state: the condition in which a person stops authoring his own conduct and becomes the instrument of an authority that, in his felt experience, has assumed his responsibility for him. He still owns his hands; he no longer owns the decision. Haslam and Reicher then dismantled the part of Milgram's picture that was always too flattering to the subject — the fiction of the reluctant automaton, cringing as he obeys. The data never supported it. The followers who go furthest are not those who feel least but those who believe most: the ones who have fused their identity with the cause until its demands register not as external pressure but as conscience. They do not capitulate. They volunteer. Milgram's subject obeys against his will; Haslam and Reicher's enacts a will he has made continuous with the authority's. The first is coerced. The second is devoted.

Hold both, because the response I am examining requires a third thing neither quite names, and it is the thing this book exists to name. Milgram's subject relocates his agency to an authority above him — the experimenter, the institution, the chain of command. He looks up. What the digital tribe has produced is the relocation of agency sideways and downward, to the peer-group, the cause, the avatar-swarm — an authority with no office, no credentials, no address, and therefore no possibility of being held to account in turn. I call this relocated deference: the transfer of one's moral judgment not to a superior who might at least be petitioned or blamed, but to a collective that exists precisely so that no one inside it can be located. Milgram's man could say I was following orders and point to the man who gave them. The relocated-deferent can say we know what really happened and point to no one — because the we is structurally anonymous, a fog with the authority of a sovereign and the accountability of a weather system. The agentic state at least left a chain of command a prosecutor could walk up. Relocated deference dissolves the chain into a crowd, and a crowd cannot be subpoenaed.

This is why the mechanism is so much more corrosive online than in Milgram's basement, and why it converges on the destruction of the verdict specifically. A verdict is the institutional form of the sentence this person did this thing, and it is established. It is adjudication made binding. Relocated deference cannot survive a verdict it dislikes, because the verdict asserts exactly what the deference has abolished: that guilt is a fact about an act, anchored to an individual, and indifferent to which fog he belongs to. So the verdict is not argued with. It is delegated away — handed up to the tribe to be overruled, the way Milgram's subject handed his conscience up to the man in the coat. The conscience being outsourced is the one that used to ask, before allegiance, before anything: what did he actually do. That question is the first casualty, and everything in the evidence that follows is the autopsy.

Now the evidence.

What the jury actually saw

Start with the facts that are not in dispute, because everything I argue afterward has to answer to them.

At a school athletic event, a teenager stabbed another teenager once in the chest. The wounded boy died. The two did not know each other; they attended different schools and were thrown together only by weather and a shared shelter. The defendant was sitting under a canopy belonging to the other boy's team. He was told to move. According to the account presented at trial, he reached into a bag and issued a threat. The other boy pushed him. He drew a knife and put it through the boy's heart.

The defendant admitted the stabbing. His defense was not that it didn't happen but that it was self-defense — that he was under physical intimidation and acted to protect himself. Over several days, the jury heard from dozens of witnesses: students who were there, the officers who responded, the medical experts. They were shown the knife and how fast it opened. The prosecution's account was blunt: you don't get to meet a shove with a stab, especially if you provoke the shove. The jury deliberated for a matter of hours and convicted him of murder. They then rejected the defense's request for a finding that could have capped his sentence well below the maximum, and imposed a sentence of decades.

That is the evidentiary reality. A confrontation he entered, an escalation he initiated, a shove answered with a fatal stab, and twelve citizens — who sat through all of it — concluding unanimously that this was murder and not self-defense. You may quarrel with a jury. People do. But quarreling with a jury means engaging the evidence the jury weighed. Hold that standard in mind, because the response this case generated did not meet it. It did not engage the evidence. It replaced it.

The story that arrived before the facts

Within days of the killing — long before any trial, any testimony, any cross-examination — a story had formed and hardened online, and it was not the story the evidence would later tell.

The earliest fundraising appeal set up for the defendant asserted that he had been jumped by the other boy and a relative, and described the defendant as a respectful, well-mannered young man with no record of any kind. This was the narrative offered to the public while he sat in jail: not a defendant entitled to a fair trial, but a victim who had been attacked and was being railroaded. The trial would dismantle the jumped account — the witnesses, nearly uniformly, described the defendant as the aggressor, the one who entered the rival shelter and escalated. But the narrative did not wait for the trial, and for a great many people it never updated when the trial came. The story had done its job the moment it arrived. It told you which side to be on, and once you were on a side, the testimony was just noise from the other one.

The money that arrived after the truth was settled

A verdict is supposed to be a terminus. Twelve people heard everything, and the question of what happened became, for legal purposes, closed. Watch, then, what happened to the giving.

A defense fund for the defendant, hosted on a niche crowdfunding platform after a mainstream site removed an earlier campaign, drew tens of thousands of donations and reached a sum well into the hundreds of thousands. Crowdfunding for a defendant is not in itself remarkable; legal costs are real, and the presumption of innocence is a value worth funding. What is remarkable — what is, for the purposes of this book, the cleanest single datum — is that the donations did not stop at the verdict. In the hours after the jury convicted him of murder and the court imposed a sentence of decades, the fund took in thousands of dollars more, a steady stream of fresh contributions arriving through the evening and into the next morning, until the platform itself closed the page. The platform's stated reasoning is worth recording, because it states the principle the donors were violating: the campaign had been permitted, a representative explained, because everyone is presumed innocent until proven guilty — and after the conviction, that rationale no longer held. The donors disagreed with their wallets. Money kept arriving for a man the evidentiary process had finished adjudicating, in defiance of the finding it had reached.

Sit with the structure of that act, because it is not the structure of charity and not the structure of doubt. A donation that arrives after conviction is not buying a defense; there is nothing left to defend. It is not expressing a reasoned disagreement with the verdict; almost none of these donors read a transcript, and a reasoned disagreement would at least engage the evidence the jury weighed. The post-verdict donation is a pure declaration of allegiance, stripped of every instrumental function except one: to register, publicly and at cost, that the giver stands with the man regardless of what he was found to have done. This is relocated deference in its most legible form. The verdict said this is established. The donor replied, in effect, not to us — and routed the disagreement not through argument but through the tribe, whose authority to overrule a jury exists nowhere but in the felt solidarity of the people invoking it. The fog issued a counter-verdict, and its citizens paid their tax to it.

That the fund later drew intense scrutiny — over reports that the family's circumstances had improved in ways some found hard to reconcile with a defense fund, claims the family disputed and the platform defended — is worth noting precisely because the picture refuses to stay clean. Some who gave were later troubled by where the money went; the deference was neither total nor uniform, and a portion of the crowd clearly kept one eye on the facts. I include this because the argument does not need it suppressed. A thesis that has to hide its disconfirming instances is identity-protective cognition wearing the author's face, and this book cannot afford to demonstrate its subject by accident.

Delegating the verdict away: the jury as illegitimate by composition

The second move was to attack not the evidence but the tribunal — and to attack it on the one axis that requires no engagement with what the tribunal heard.

An advocacy organization backing the defendant denounced the demographic composition of the jury. Set aside whether jury composition can ever be a legitimate appellate concern; sometimes it is. Notice instead what the complaint does in the discourse. It supplies a reason to discard the verdict that never touches the verdict's contents. You need not have followed a minute of testimony — the threat, the hand kept inside the bag, the single fatal thrust answered to a shove, the long roster of witnesses, most of whom described the defendant as the aggressor who would not leave a shelter that was not his — to deploy it. The composition objection is evidence-independent by design. It lets the verdict be set aside a priori, on a property of the jurors rather than the quality of their reasoning, which is exactly what relocated deference requires: a mechanism for nullifying an adjudication without ever entering the adjudication's terms.

This is the move's elegance and its tell. A genuine challenge to a verdict gets its hands dirty in the record. A delegation of the verdict to the tribe reaches instead for the one fact — the jurors' identity — that can be known without the record and that converts a finding of fact into an artifact of the enemy camp. The verdict is not refuted. It is reassigned: lifted out of the category what the evidence showed and dropped into the category what they did to us. Once there, it is no longer a truth to be reckoned with. It is a grievance to be felt.

The reflex has no home team

If relocated deference were the property of one tribe, it would not be the mechanism I have described; it would be a fact about that tribe. It is the mechanism. And the surest demonstration is that this case set it running in every direction at once.

The reaction against the defendant's supporters generated its own evidentiary denial, structurally identical to the thing it condemned. A widely shared claim held that the defendant had been given a free public defender while his family enriched themselves — a charge that flattered the priors of the opposing tribe and that was false: he was represented by privately retained counsel. The claim spread anyway, because spreading required no more verification than the donations did. The threads dehumanizing the lone assailant from the opening of this chapter — reducing the man to a caricature, his worth collapsed into the side he was taken to represent — ran the identical operation the assailant himself had run: a person dissolved into a category, contempt substituted for assessment. Each tribe looked at the other and saw, with perfect clarity, the mechanism it could not see in itself. That mutual clarity is not a refutation of the thesis. It is the thesis. The reflex does not belong to a people. It belongs to the condition of deferring one's judgment to a fog, and fogs form on every side of a sufficiently charged event.

The part that implicates you

Here is where this chapter turns, and where its argument stops being comfortable for the kind of reader it is written for.

The intuitive theory of motivated reasoning is consoling: it holds that tribal denial is a failure of intelligence, the recourse of people who cannot follow an argument, and that the cure is more reasoning, better-educated minds, sharper critical faculties. The empirical record does not support this, and the reader flattering himself as an exception should attend closely. Dan Kahan's work on identity-protective cognition found that greater cognitive reflection and numeracy do not attenuate motivated reasoning. They amplify it. In his "motivated numeracy" experiments, the subjects most capable of correctly interpreting quantitative data did so reliably when the correct answer flattered their political identity — and abandoned that same capacity, systematically, when the numbers pointed the wrong way. Polarization did not shrink among the most numerate; it grew. (The effect has been contested at the margins, with replications suggesting a weaker version of it holds; the direction, not the existence, is what the argument here requires.) On identity-charged questions, the most analytically capable subjects were not the most accurate; they were the most reliably aligned with their group — and they got there by deploying their superior faculties in the service of the conclusion their identity required. Crucially, this was not the lazy thinking of people too tired to reason. It was the opposite: active, effortful, deliberate cognitive engagement, aimed at identity protection. Intelligence did not function as a brake. It functioned as a better engine for arriving where the tribe already stood.

This is the finding that should disturb the educated reader more than any punch. You do not escape relocated deference by being smart. You enter it more elegantly. The person who cannot construct an argument defers crudely — he's one of ours, leave him alone. The person who can construct an argument defers in prose: he builds the systemic frame, marshals the historical analogy, locates the procedural irregularity, and produces, at the end, the identical verdict — not guilty in the ways that matter — having experienced the whole operation as independent critical thought. His sophistication does not detect the deference. It launders it. The rationalization is so well-made that it persuades its own author he reasoned his way to it, when in fact he reasoned his way from it, backward, from a conclusion the tribe handed him before he began.

And because the well-made rationalization is portable, it does not stay with its author. It becomes a script. The articulate defender supplies the framing that the inarticulate adopt wholesale — the smart people have looked into it, and it's one explanation; the smart people have looked into it, and it's the opposite explanation — so that the cascade runs downhill from the very minds that imagine themselves at the top of it, immune. The intelligent are not the exception to the crowd. They are its authors. They write the denials the rest recite.

This is the indictment the book has been building toward, and it is aimed, deliberately, at the people most certain it cannot apply to them. If you have read this chapter composing the counter-argument — the case is more complicated, the framing unfair, the author has surely missed the relevant nuance — that activity is not proof of your exemption. It is the mechanism, running. The only defense against relocated deference is not intelligence, which is its instrument, but a specific and effortful discipline: the willingness to let the evidence reach a conclusion your tribe would not have chosen, and to stay there. That discipline is rare in proportion to how little it flatters. It is the one thing this case found almost no one, on any side, willing to do.

When No One Is in Charge — Relocated Deference Without a Defendant

The contested-verdict cases gave us relocated deference organized around a person: a defendant to rally to, a verdict to overrule, a narrative to maintain. Strip the person away and you might expect the mechanism to vanish with him. It does not. It intensifies. What remains when there is no individual to defer about — no one to defend, no one to blame, no name to chant — is the mechanism in its purest state, and it has a name in the present tense. They are called coordinated teen swarms, and they are the clearest demonstration this book can offer that the relocation of responsibility does not require an authority at all. It requires only a crowd large enough to dissolve the person inside it.

The form

The facts are not in dispute, because they are filmed; filming is the point. In recent years a pattern that had simmered for some time reached a rolling boil across many cities. Through the major social platforms — and, in a detail that belongs in a chapter on the industrialization of impulse, through AI-generated flyers produced at no cost in seconds — large numbers of adolescents began converging on malls, parks, restaurants, and entertainment districts with little warning and no announced purpose. The gatherings swell within hours and, with grim regularity, deteriorate. In one such incident inside a restaurant, adults and small children huddled in a corner while juveniles threw furniture; one used a piece of children's seating as a weapon. In another, a crowd numbering in the hundreds flooded an open-air entertainment plaza and the evening ended in arrests. In yet another, police arrested a few dozen people ranging from young children to young adults after a gathering in a downtown park produced fights, narcotics, and a weapon. City after city: the map fills in, week by week, through warm-weather stretches that authorities warned would be worse than the seasons before.

I am going to resist the two ready-made readings of these events, because both are evasions, and the evasions are themselves specimens of the thing this book studies.

The two evasions

The first reading is the punitive one: these are simply criminals, the product of lax enforcement and permissive cities, and the solution is force. There is truth in the observation that the events are criminal — furniture thrown at children is not a youth-programming deficit — and I will not pretend the destruction is imaginary or the arrests unjust. But "they are simply criminals" is, as an explanation, a refusal to explain. It treats the most striking feature of the phenomenon — that it is collective, that it is coordinated, that it produces in ordinary adolescents behavior most of them would never produce alone — as incidental, when it is in fact the entire mystery. A child who would not throw a chair in an empty room throws one in a full one. The crowd is not the setting of the crime. The crowd is its cause.

The second reading is the exculpatory one: these are kids with nowhere to go, products of poverty and neglect and the criminalization of ordinary adolescence, and the gatherings are mostly benign socializing that a few bad actors and a hostile press have distorted. There is truth here too. Most attendees harm no one; the structural deprivations are real; there are plenty of young people simply trying to exist, as defenders of the gatherings often point out, and that is a fair description of the majority. But notice what this reading does with the minority who did throw the chair, who did break the window, who did fire the gun. It relocates their responsibility — to poverty, to the city, to the police who provoked them, to anyone and anything except the hand that acted. It is the agentic state translated into the language of social justice: I was not the author of this; the conditions were. The conditions are real, and they are not the author. A cause is not an excuse, and the slide from the first to the second is precisely the move that interests us.

What both evasions share is the disappearance of the individual. The punitive reading dissolves the person into a category — criminal, them, those people — and the exculpatory reading dissolves the person into a circumstance — poverty, neglect, the system. Neither will say the sentence the whole phenomenon is built to prevent anyone from saying: this specific person chose to do this specific thing, and could have chosen otherwise. The crowd in the street is engineered to make that sentence unsayable. The crowd in the commentary obliges.

Why the swarm is the authority

Milgram's subject relocated his agency to an authority above him — an experimenter in a lab coat insisting that the procedure required him to go on. The swarm has no such figure. There is no leader to obey, no organizer to point to, often no purpose anyone could articulate beyond everyone is going. And yet the relocation of responsibility is total — more total, arguably, than anything Milgram produced — precisely because there is no authority to name. This is the dark elegance of the form. In the laboratory, the subject could at least say afterward, he told me to. In the swarm, there is no he. Responsibility does not transfer to a person; it diffuses into a collective that has been constituted, whether the participants understand this or not, so that no individual inside it can be located as the cause of anything.

This is relocated deference completing its own logic. In the previous chapter the deference ran to a tribe that could at least be named — our side, our people. Here it runs to a tribe that exists only for the duration of the event and dissolves the moment the police arrive, leaving behind a thousand individuals each of whom can say, with technical accuracy, I didn't start it; I was just there. The swarm is an authority with no head, which means it is an authority that can issue commands — throw it, film it, run — while offering every follower a permanent alibi: the command came from no one, so the responsibility belongs to no one. Each participant defers to a crowd, and the crowd is nothing but the sum of participants deferring. It is obedience with the authority deleted and the obedience kept. That is the purest agentic state yet devised, and the adolescents did not devise it. The platform did.

The platform's hand

Because the second indispensable actor here is not a teenager. It is the recommender system. A flash mob in the early social-media era required someone to organize it and a reason to gather; the contemporary swarm requires neither, because the platform supplies both. The footage of last week's chaos is itself the flyer for next week's: virality is the recruitment. An algorithm tuned to maximize engagement cannot help but privilege the chair thrown across the restaurant over the thousand quiet evenings, because the chair is what gets watched, shared, and stitched. The platform does not intend to organize violence. It intends to maximize attention, and violence maximizes attention, and so the system arrives at the organization of violence the way water arrives at the sea — not by intention but by gradient. The AI-generated flyer is merely the most literal expression of a machinery that was already manufacturing these events out of the raw material of adolescent impulse and the universal human susceptibility to the crowd.

This is the same engine described in the previous chapter, running on a younger fuel. There, intelligence laundered tribal denial into sophisticated argument and the algorithm boosted the argument. Here, no argument is required; the algorithm boosts the act directly, and the act recruits the next crowd. In both cases the platform has discovered that the relocation of responsibility is profitable — that content in which people surrender their individual judgment to a collective, whether the collective is an ideological tribe or a swarm in a parking lot, is content that performs. The agentic state has been found to be good for engagement. What Milgram staged in a basement for science, the recommender system now runs continuously, at population scale, for revenue.

What it costs, and to whom

It is worth naming the people who do not get to relocate anything. The family in the corner of the restaurant. The shopkeeper whose store is rushed. The bystander caught by the gunfire that has, in several of these events, moved from threat to fact. They did not defer their judgment to a crowd; they were the objects against which a deferring crowd discharged itself. The literature on diffusion of responsibility has always had a second half that the perpetrators' debates ignore: every diffusion of responsibility among the actors is a concentration of harm upon someone who had no part in it. The buck does not vanish when a crowd relocates it. It lands. It lands on whoever was standing where the crowd decided, collectively and authorlessly, to be.

And here the chapter rejoins the book's spine. A swarm is what a society looks like when enough people, often enough, have practiced the relocation of responsibility until it became reflexive — until I was just there feels not like an evasion but like a fact, and the conditions did it feels not like an abdication but like sophistication, and the question that used to come first, before the crowd and before the cause — what did I choose to do — has gone quiet in enough heads at once that the street is no longer safe. The swarm is not a different phenomenon from the contested-verdict discourse. It is the same phenomenon with the defendant removed, demonstrating that the mechanism never needed him. It only ever needed the crowd, the platform, and the relief of being no one in particular.

CHAPTER 6

The Oldest Reflex — Relocated Deference Before the Feed

If everything in this book were a symptom of the smartphone, the smartphone would be the cure: confiscate it, and the disease lifts. The argument is not that comforting, and it is not that easy. The relocation of responsibility is not a malfunction the internet introduced into an otherwise sound species. It is one of the oldest things we do — older than the platforms, older than the nation, older than writing — and the honest version of this chapter has to begin by conceding how deep the root goes before it can claim, at the end, that the root does not determine the harvest.

Milgram was reporting, not predicting

Return to the basement, but read it correctly this time. The usual lesson drawn from Milgram is a prophecy: this is what people will do. That is not what the experiment showed. It showed what people were already disposed to do, surfaced under laboratory conditions — that the disposition to hand one's moral agency to a legitimate-seeming authority was present, in ordinary mid-century Americans, waiting. Milgram did not install the agentic state in his subjects. He found it there. The shocking sixty-five percent was shocking precisely because it revealed a standing capacity, not a manufactured one.

Haslam and Reicher's revision sharpens this into the form the present book needs. Their finding — that the most committed followers are not the most coerced but the most identified, the ones who have fused the authority's project with their own sense of who they are — describes a mechanism with no expiration date and no cultural border. Identity fusion is not a modern pathology. It is the cognitive substrate of every army that ever marched, every congregation that ever knelt, every tribe that ever defended its own against the evidence. What the laboratory isolated, history had been running at full scale for as long as there were groups to belong to. The agentic state is not news. The conditions that now trigger it a hundred times a day are.

The long record

Consider how reliably the same structure recurs, stripped of its period costume.

A population is told, by an authority it has reason to trust, that a despised minority is the source of its troubles. Most people do not investigate. They defer — not because they have weighed the claim and found it sound, but because the claim arrives wrapped in the identity of the group they belong to, and to doubt it is to step outside the group. The few who do investigate, and who possess the education to do so well, frequently produce the most sophisticated justifications of all: the period's physicians, jurists, and professors lending their analytical equipment to the conclusion their tribe required. This is not a description of one atrocity. It is the common architecture of most of them — the machinery Hannah Arendt was reaching for with a phrase that has been worn smooth by overuse but still cuts when you press on it: the banality with which ordinary people, deferring upward and inward, become the instruments of things no single one of them would author alone.

I am deliberately not naming a single case here, and the restraint is methodological, not timid. The moment this chapter attaches the mechanism to one historical villain, the reader's tribal sorting switches on — ah, he means them, the bad people, not us — and the analysis dies the instant it becomes an accusation against an out-group. The point is the reverse. The mechanism is not the property of the people we have agreed to despise. It was available in them and it is available in us, on the same terms, through the same faculty. Every reader of this book belongs to some group whose narrative he has, at least once, deferred to against the evidence, and the reader who is certain he is the exception is the one this book is most concerned about, for reasons Chapter 4 has already made plain.

What actually changed

If the reflex is ancient, what is the digital age responsible for? Three things, none of them the reflex itself.

Speed. The historical relocation of responsibility took time — a pamphlet, a sermon, a rumor moving at the pace of feet and printing presses. A narrative now forms and hardens in hours, as any recent online verdict demonstrates: fully articulated, complete with assigned heroes and villains, before a single fact has been tested. The window in which evidence might have interrupted the narrative has closed before the evidence arrives.

Scale. The historical crowd was bounded by the square it could fit into. The digital crowd has no walls. A swarm that once required physical proximity now assembles across a continent, and the coordinated teen swarm on a commercial district is merely the moment the digital crowd condenses back into a physical one, having been organized in a medium that has no edges.

And selection. This is the genuinely new thing, and the most dangerous. Previous eras had authorities who intended to relocate your responsibility — the demagogue, the propagandist, the recruiting sergeant. The recommender system intends nothing. It has discovered, without understanding, that content which invites the surrender of individual judgment outperforms content which demands its exercise, and it serves accordingly. We have, for the first time, built a machine that finds the agentic state profitable and optimizes toward it as indifferently as it would optimize toward anything else. The reflex is ancient. The machine that farms it is a generation old.

The proof that it is not destiny

Here is where the chapter turns, and where the book earns the right to its eventual hope, because a mechanism this old and this deep invites a fatalism I want to refuse with evidence rather than sentiment.

If relocated deference were a destiny, history would be uniform — an unbroken record of tribes deferring their consciences into atrocity. It is not uniform. The same record that contains the recurring machinery also contains long, real stretches in which it was held in check: societies and periods in which people of different tribes lived in functional proximity, traded, intermarried, and adjudicated their disputes by something other than which side the disputant belonged to. These periods were not utopias and I will not pretend they were; they were arrangements, often pragmatic and often fragile, in which the contextual judgment of the individual was expected and the reflexive deference to the tribe was checked — by institutions, by norms, by the simple habituated experience of the out-group member as a person rather than a category.

What the durable ones had in common is instructive, because it is the opposite of what the feed supplies. They slowed things down: disputes went to processes that took time, and time is the enemy of the snap tribal verdict. They raised the cost of the reflexive accusation rather than rewarding it. They built repeated, low-stakes contact across the tribal line, so that the out-group member was a known quantity rather than an avatar. And they held individuals — not categories, individuals — answerable for what they did. Every one of these is a structural choice, which means every one of them is a thing that can be chosen again. The reflex is in us. So is the capacity to build the cage that holds it. History is the record of both, and the pessimist who cites only the first half is, in his own way, deferring to a narrative against the evidence.

That is the hinge on which the rest of this book swings. The relocation of responsibility is not the smartphone's invention and will not end with the smartphone's reform; it is older than that and harder than that. But it is also not fate, and the proof is that we have caged it before, with tools — slowness, cost, contact, individual accountability — that we still possess and have merely stopped using. The chapters that follow are about why we stopped, and what it would take to start again.

When the Verdict Goes the Other Way — The Symmetry Proof

A thesis that only ever indicts one side is not a thesis. It is a banner. If relocated deference is real — if it is a mechanism rather than a grievance — then it must be visible operating in the tribe opposite to whichever one a given reader is inclined to blame, on the same terms, through the same faculty, with the same indifference to evidence. This chapter exists to supply that proof, and it supplies it with a case that inverts almost every variable of the one at the center of the previous chapter.

In the earlier case, a defendant on one side of the racial frame was convicted, and a largely progressive public refused the conviction. Reverse each term. In the case examined here — a contested acquittal after a death during a confrontation in a public space — a defendant on the other side of that frame was acquitted, and a largely conservative public celebrated the acquittal with the same evidence-optional fervor, while a largely progressive public refused it with the mirror image of the denial they had been accused of demanding in the other direction. Same mechanism. Opposite polarity. If you found the previous chapter persuasive and find this one uncomfortable, the discomfort is the data. It is the feeling of the mechanism operating in you.

The facts, as the jury had them

The case can be stated in its load-bearing particulars, stripped of every detail that would tie it to one real event. In an enclosed public place, one man physically restrained another who was, by multiple witness accounts, behaving erratically and voicing alarming threats. The restrained man was in evident distress; he was unarmed and had struck no one. The hold was maintained until he lost consciousness, and he died. Bystanders partially recorded it.

The man who applied the restraint was charged with a homicide offense. The defense argued he was protecting himself and other bystanders from a volatile person making alarming threats, and disputed whether the restraint was the medical cause of death. The prosecution argued he had answered a frightening person with grossly disproportionate and sustained force — that he had reacted, as one framing put it, to a peril rather than a person. After several days of deliberation the jury deadlocked on the more serious charge, which was dismissed, and then acquitted on the lesser. Twelve citizens who heard the evidence concluded the state had not proven the charge beyond a reasonable doubt.

Hold that outcome with the same discipline the previous chapters demanded. A jury acquittal is a finding that the case was not proven, arrived at by people who sat through all of it. You may disagree with it. Disagreeing with it honestly means engaging the evidence and the legal standard the jury applied. Watch, now, how little of the public response on either side did that.

The right: the verdict as vindication of a type

The acquittal was received in much of the conservative public not as a finding about the proof in one case but as the vindication of a category. The acquitted man became, instantly and across a vast audience, a hero — a word that does not appear in a verdict of not guilty. The representative register ran roughly thus: justice has prevailed; a good man was spared punishment for doing the right thing, for defending his fellow riders from a dangerous, mentally ill stranger the system had failed. Notice everything packed into that reception that the jury did not find. The jury did not find that the defendant did the right thing; it found that the state failed to prove he did a criminal one — a different and narrower proposition. It did not pronounce him a hero or the dead man a criminal. The leap from not proven guilty to affirmatively heroic is not an inference the evidence licenses. It is the tribe conferring a status, and the verdict was the occasion, not the cause.

This is relocated deference wearing the colors that the previous chapter's mechanism wore reversed. There, supporters relocated a convicted man's responsibility onto the system. Here, supporters relocated an acquitted man's act — the minutes of the hold, the unarmed dead man, the questions of proportion a reasonable person might sit with — into the safe abstraction of heroism, where it need not be weighed at all. The deference runs to the tribe's narrative of the righteous defender, and once the narrative is in place, the uncomfortable particulars are filed away as the noise of the enemy camp. The conservative who declared the defendant a hero the moment the verdict landed had, in the overwhelming majority of cases, weighed the trial record exactly as carefully as the progressive who, in the earlier case, declared a convicted man a victim: not at all. Each looked at the verdict, consulted his tribe, and felt a certainty the evidence had not earned him.

The left: the verdict as proof the system is rigged

And the mirror held on the other side. For a large progressive public the acquittal was received not as a finding that a charge was unproven but as confirmation of a fixed prior: that the system is rigged, that a man on one side of the racial line will always be permitted to kill a man on the other, that the verdict was less a judgment than a revelation of the machine's true setting. The dead man's family gave the line its purest form outside the courthouse — the system is rigged — and departing spectators named the country itself a racist one. These are not, whatever their truth as general propositions, engagements with what twelve specific jurors weighed over several days about one specific killing. They are the verdict being delegated away exactly as the earlier verdict was: lifted out of the category what the evidence in this case showed and dropped into the category what they always do to us, where it confirms the narrative no matter which way it had gone.

This is the tell that the mechanism is symmetric and not partisan: the structure of the progressive denial of this verdict is identical to the structure of the conservative denial of the earlier one, line for line. A jury reached a conclusion the tribe disliked; rather than contest the specific reasoning, the tribe contested the legitimacy of the tribunal itself, on grounds — race, system, rigging — that require no contact with the trial record and that would have been deployed regardless of the evidence. Each tribe can see this operation with perfect clarity when the other performs it. Neither can feel it in itself. That mutual blindness is not a coincidence. It is the signature of relocated deference, and it appears, identically, on both sides of both cases.

It is worth being precise about why the racial frame is the one that detonates here, because the precision is itself part of the argument and not a departure from it. Dan Kahan's work on identity-protective cognition and motivated numeracy supplies the mechanism: people do not reason their way to a position and then adopt a tribe; they begin with the tribe and recruit their reasoning to defend the position the tribe has already assigned. The more a question is bound up with group identity, the more a person's cognitive horsepower is spent not on getting the answer right but on getting the answer the group requires — and Kahan's striking finding is that greater numeracy and greater education make this worse, not better, because they furnish more sophisticated tools for rationalizing the foregone conclusion. A frame that maps a dispute onto a salient group identity therefore short-circuits evaluation on every side it touches. This is not a claim that one group reasons worse than another, or that any group's reading of a given case is the correct one. It is the opposite claim: the racial frame is combustible precisely because it captures everyone symmetrically, recruiting the conservative's reasoning and the progressive's reasoning alike into the service of a verdict each had reached before the trial began. The mechanism has no preferred outcome. It only has a preferred shortcut.

The harder mirror: this case is genuinely harder

I want to resist a cheap symmetry here, because the cheap version would itself be a kind of dishonesty. The case examined in this chapter is, on the merits, harder than the one in the previous chapter, and a book that flattened them into perfect equivalence would be massaging the evidence to serve its own thesis — the very sin it indicts.

The earlier evidence pointed in a relatively clear direction: an escalation initiated, a shove answered with a fatal blow, a self-defense claim a jury rejected after a short deliberation. The case here contains real and unresolved difficulty: a man in genuine mental crisis making genuine threats; a restraint that began as arguably defensible and became, over several minutes, something a reasonable person can question; a real dispute about medical causation; and a jury that deadlocked on the more serious charge before acquitting on the lesser. Reasonable people, having weighed all of it, can disagree about whether the acquittal was correct. That is precisely what makes the case useful here. Relocated deference does not require an easy case to operate on; it operates on hard ones too, by relieving its adherents of the obligation to sit in the hardness at all. The honest position is this is difficult, and I can see why twelve people who heard everything were first unable to agree and then unwilling to convict. Almost no one on either side occupied that position. The difficulty was the first thing both tribes discarded, because difficulty is precisely what the tribe exists to spare you.

The common thread

Set the two cases side by side and the thread is unmistakable. Different defendants, different races, different verdicts, different tribes doing the relocating — and one identical operation running underneath all of it. In each case a jury performed the slow, evidence-bound work of adjudication. In each case a large public, on the side for whom the outcome was unwelcome, declined to engage that work and instead delegated the verdict upward to the tribe to be overruled, on grounds that required no knowledge of the record. And in each case the side for whom the outcome was welcome performed the equal and opposite error — converting a narrow legal finding into a sweeping vindication of their narrative, a hero or a victim conjured from a document that named neither.

The mechanism does not care who you are. It offers the same bargain to everyone: surrender the exhausting work of weighing this specific evidence about this specific act, and receive in exchange the warm certainty of the group, pre-formed and identity-confirming, available the instant the verdict drops and impervious to whatever the verdict actually said. The previous chapter showed you the bargain accepted by one tribe. This chapter shows you the identical bargain accepted by its opposite. If both chapters have done their work, you can no longer locate the disease in a people. You can only locate it in the bargain — and in the faculty, common to all of us, that finds the bargain so hard to refuse.

CHAPTER 8

The Bill Comes Due — What Relocated Deference Costs

Part I built the mechanism and Part II watched it run. This chapter asks the question that the cases kept deferring: what does it cost? Not in the abstract — not "the erosion of discourse" or "the decline of trust," phrases so worn they have stopped meaning anything — but concretely, in damage that can be located and named. Relocated deference is not a tragedy of ideas. It is a transfer of harm from the people who perform the reflex to the people who absorb its consequences, and the second group is always more specific, and always less heard, than the first.

I want to organize the costs by who pays them, because the abstraction that hides this mechanism is precisely the abstraction that lets its costs go uncounted. A cost with no name attached is a cost no one is asked to answer for. So I will attach names.

The first to pay: the individual at the center

Every kind of case in this book has a person at its center who becomes, in the public response, something other than a person: a symbol, a cause, a proof, a flag. This is the first and most intimate cost of relocated deference, and it is paid by people on every side of every case.

When a tribe takes up an individual as the embodiment of its narrative, it performs a strange violence even when it means to help. The defendant becomes a vessel for a story that is not about him, and the actual human being — with his actual conduct, his actual responsibility, his actual due — disappears beneath the symbol the tribe requires. The supporters who knew their man was innocent before any evidence, and who remained certain after a conviction, were not, whatever their intentions, seeing him. They were seeing the story. And a person who has been converted into a story has, in a real sense, been erased — his particularity sacrificed to a narrative that would have used any sufficiently useful body. The same erasure is performed by the opposing tribe, which converts the same person into the embodiment of everything it fears, equally indifferent to who he actually is. Between the two tribes, each insisting on its own symbol, the individual vanishes entirely. No one in the roaring discourse is looking at him. They are looking at what he has been made to mean.

There is a particular cruelty in this for the dead, who cannot object to the use being made of them. A teenager killed in a fatal altercation at a school event becomes an argument. A man who died during a confrontation in a public space becomes an argument. Their deaths, which were the end of an actual irreplaceable human life, are conscripted into rhetorical service by people who never knew them and who require them to mean something the dead never agreed to mean. The relocation of responsibility has, as one of its quieter costs, the relocation of the dead — from persons who were lost to symbols that are useful. It is hard to imagine a more complete failure to see another human being, and it is performed daily, at scale, by people entirely convinced of their own compassion.

The second to pay: the bystander

An earlier chapter named these people, and I return to them because they are the most invisible victims of the mechanism, the ones whose suffering does not fit any tribe's narrative and is therefore not taken up by any tribe at all.

The family huddled in the corner of the restaurant. The shopkeeper whose store is rushed and emptied by a coordinated teen swarm on a commercial district. The stranger in a public space who absorbs the violence that a crowd, having diffused its responsibility into nobody, discharges onto whoever is nearest. These people did not relocate their judgment to anyone. They were simply present at the spot where a deferring crowd, or a deferring individual acting in a tribe's name, decided to act. And here is the structural cruelty: because their suffering serves no narrative, it generates no large online defense fund, no rallying slogan, no march. The bystander is harmed by the mechanism and then harmed again by its economy of attention, which has no slot for a victim who cannot be made to mean something tribal. The diffusion of responsibility among the actors is, always and exactly, a concentration of unanswered harm upon someone who had no part in the dispute and no tribe to take up his cause.

This is the cost the mechanism is best at hiding, because hiding it is structurally built in. A harm with no responsible party — diffused into a crowd, a system, a fog — is a harm with no one to sue, no one to blame, no one to hold. The bystander's injury is real and his perpetrator is, by design, nowhere. He has been hurt by an authority that cannot be located, which means he has been hurt by no one the world will ever name, which means his hurt, however real, will not be paid for. Relocated deference does not merely produce victims. It produces victims of no one, which is the only kind of victim a society can fully ignore.

The third to pay: the institutions

Then there are the costs that fall not on persons but on the load-bearing structures that a society needs in order to remain one — and chief among them, the institutions whose entire function is to adjudicate truth and apportion responsibility.

Consider what relocated deference does to a jury. The jury is one of the few remaining institutions explicitly designed to do the thing the mechanism destroys: twelve people, removed from the crowd, made to sit with the actual evidence in a specific case, and required to reach a finding bound by that evidence rather than by their tribe. It is, in a sense, an anti-relocation machine — a deliberate structure for forcing individual judgment back onto people who would otherwise defer. And in case after case, we watched the verdict it produced be treated not as a finding to be reckoned with but as an artifact of the enemy, to be discarded on grounds — the makeup of the jury, the alleged corruption of the system — that required no engagement with what the jury actually weighed. Every time a tribe delegates a verdict away rather than contend with it, the jury's authority erodes a little further. And a society that has stopped believing its juries can find facts has lost the institution by which it converts contested events into settled ones. What replaces the verdict is not a better truth. It is permanent, tribal, unresolvable dispute — every case relitigated forever in the only court that remains, which is the feed, where nothing is ever decided because deciding was never the point.

The same corrosion reaches every institution whose authority depends on being trusted to find truth without regard to tribe: the courts, the press, the universities, the agencies, the slow expert processes that a complex society cannot function without. Relocated deference does not argue these institutions are wrong in a particular case. It does something more total: it reclassifies them as belonging to a tribe, so that their findings can be dismissed wholesale, in advance, by anyone for whom that tribe is the enemy. An institution that has been successfully tribalized can no longer settle anything, because half the population has been given permission to treat its every output as enemy propaganda. We are watching this happen to one institution after another, and the cost is not the discrediting of any single finding. It is the loss of the very possibility of shared, authoritative fact — the ability of a society to say, together, this is what happened, and move forward from it. Without that ability, there is no forward. There is only the permanent present of mutual accusation.

The fourth to pay: everyone, in the loss of the commons

This is the most diffuse cost and the hardest to make vivid, which is exactly why it is the most dangerous: it is the cost no one feels as a discrete injury because everyone pays it as a general condition.

A society is, among other things, a shared reality — a common stock of agreed facts within which people who disagree about values can still argue productively, because they are at least arguing about the same world. Relocated deference dissolves this shared reality by making facts themselves tribal property, as the book's title insists. When my facts are simply the claims my side holds and yours are the claims yours holds, we have not merely disagreed; we have lost the common ground on which disagreement could be adjudicated. We can no longer be shown to be wrong, because being shown requires a shared standard of evidence, and the shared standard is exactly what has been tribalized away. We become, each of us, sealed inside a reality our tribe maintains, immune to correction, incapable of being reached.

The cost of this is not measured in any single event. It is measured in the slow loss of a capacity — the capacity of a large, diverse society to remain one society rather than fragmenting into mutually unintelligible camps, each certain, each unreachable, each convinced the others are not merely wrong but lying or mad. This is the terminal condition the mechanism trends toward: not a society that argues badly, but one that can no longer argue at all, because argument presupposes a common world and the common world has been parceled out as tribal property. A people that has lost the ability to establish shared facts has lost the ability to govern itself, because self-government is, at bottom, the practice of deciding together what is true and what to do about it. Relocated deference attacks the first half of that sentence, and the second half cannot survive without it.

The feedback that makes it worse

I should close by naming the property that makes all of these costs not static but accelerating, because a cost that compounds is a different and more urgent thing than a cost that merely persists.

Each of these harms feeds the mechanism that produced it. The erosion of shared fact makes every new event more available for tribal capture, because there is less common ground to resist the capture. The discrediting of institutions removes the structures that might have arbitrated, leaving the field clearer for the next relocation. The hall of mirrors of Chapter 3 inflates the apparent hostility of the out-group, which justifies more hostility in return, which the machine amplifies, which inflates the apparent hostility further. The mechanism does not reach an equilibrium. It spirals, each turn making the next turn easier, the costs of each round lowering the resistance to the next. This is why the situation feels, to many who live in it, like acceleration rather than mere decline — like a thing speeding up rather than simply being bad. It is speeding up. The costs are also the fuel.

That is the bill, itemized by who pays it: the individual erased into a symbol, the bystander harmed by no one, the institutions stripped of their authority to settle anything, and all of us, slowly, losing the shared world that made us a society rather than a collection of warring certainties. The next chapter asks why, knowing all this, we keep paying — why the mechanism persists and intensifies even among people who can see, when shown, exactly what it is doing to them.

CHAPTER 9

Why We Keep Paying

The previous chapter itemized a bill that ought to be intolerable. The individual erased, the bystander unanswered, the institutions hollowed, the shared world parceled out — these are not abstract losses, and they are not, for the most part, hidden. Many of the people performing the mechanism can be brought, in a candid moment, to acknowledge most of what the last chapter described. And yet they keep paying. The reflex persists, and worse, it intensifies, among people who can see — when it is pointed out to them — exactly what it is doing. This chapter asks why. A pathology that survived only ignorance could be cured by information. The reason this one is so much harder than that is that it does not survive on ignorance. It survives on reward.

Because it feels like virtue

Start with the most important reason, the one that defeats nearly every attempt to argue people out of the mechanism: relocated deference does not feel like a vice. It feels like a virtue, and not a minor one. It feels like loyalty, like solidarity, like courage, like moral seriousness, like being the kind of person who stands up rather than stays silent. The engaged follower of Chapter 1 was not gritting his teeth through an unpleasant duty. He was experiencing the warm, genuine, deeply rewarding sensation of being good — of being on the right side, with his people, against injustice.

You cannot easily argue a person out of a behavior that constitutes, in his own experience, his moral identity. To ask him to examine the evidence that might unsettle his tribe's position is not, to him, an invitation to think more clearly. It is an invitation to betray — to abandon his people, to give comfort to the enemy, to become the kind of person who, when it mattered, equivocated. The discipline this book will eventually recommend — holding a fact that wounds your own side — does not register, from inside the agentic state, as intellectual honesty. It registers as treason. And the psychological cost of feeling like a traitor is so high, and the reward of feeling like a loyal member of the righteous group so constant, that for most people, most of the time, the math is not close. We keep paying because the mechanism does not present itself as a cost at all. It presents itself as the very thing we most want to be: good people, among our own, on the right side of history.

Because the alternative is genuinely harder

There is a temptation, in books like this, to treat the people inside the mechanism as simply lazy or cowardly, and to imply that the author, of course, sees clearly from some vantage outside. I want to refuse that, because it is both false and self-flattering in exactly the way Chapter 2 warned against. The reason most people defer is not that thinking for oneself is easy and they are too weak to do it. It is that thinking for oneself, on identity-charged questions, is genuinely, brutally hard — and the difficulty is not mainly intellectual.

To hold your own judgment against your tribe's, you must do several things that run against deep grain at once. You must tolerate uncertainty, where the tribe offers certainty. You must sit in the discomfort of incomplete information, where the tribe offers a complete and satisfying story. You must risk the disapproval of the people whose regard you depend on, where conformity offers their warmth. You must do the slow, effortful work of actually weighing evidence, where deference offers an instant answer that feels just as true. And you must do all of this with no reward at the end except the private and often unshared knowledge that you got it right — against a mechanism that pays out, immediately and socially, every time you decline the effort. The deck is stacked, and it is stacked at the level of basic human motivation. We keep paying not because the alternative is slightly less convenient but because the alternative asks us to be lonely, uncertain, and effortful in exchange for nothing the tribe can see. Most people, offered that trade, decline it, and they are not fools for declining it. They are responding rationally to an incentive structure that has been engineered to make the right thing feel like the hard, cold, friendless thing.

Because intelligence makes it worse, not better

Here the book's most uncomfortable claim returns, and it explains a phenomenon that the "ignorance" theory cannot: why the mechanism is so robust precisely among the educated, the analytical, the people who pride themselves on independent thought.

If relocated deference were a failure of reasoning ability, then more reasoning ability would cure it, and the most intelligent would be the most immune. The evidence, as Chapter 2 laid out, runs the other way. Greater analytical capacity, on identity-charged questions, produces not more accuracy but more sophisticated rationalization — a superior ability to construct the argument that lets one defer to the tribe while experiencing the deference as independent reasoning. This is why intelligent people are not easier to free from the mechanism but harder. The unsophisticated deferrer at least knows, dimly, that he is going along with his group; the sophisticated one has built an elaborate intellectual edifice proving that he reached the tribe's conclusion entirely on his own, by pure reason, and this edifice is far more resistant to challenge than crude loyalty ever was. You cannot easily show a man that his reasoning is motivated when his reasoning is genuinely excellent — when every individual step is defensible and only the destination was fixed in advance. His intelligence does not detect the motivation. It conceals it, from him most of all.

This is the trap that closes hardest on exactly the readers most likely to finish this book. The more capable you are of building a good argument, the more capable you are of building a good argument for whatever your tribe requires, and the more completely that capability disguises the requirement as a discovery. We keep paying, and the most analytically gifted among us pay most reliably of all, because their gift is precisely the tool that makes the bill invisible. The mechanism does not need them to be stupid. It needs them to be smart in service of a conclusion they did not reach by being smart, and they are only too able to oblige.

Because the machine never stops paying out

The final reason is the one Chapter 3 established and that deserves restating here in its proper place among the others: the reward is not occasional. It is continuous, immediate, and engineered.

Every previous mechanism of tribal mobilization in human history had downtime. The sermon ended; the rally dispersed; the pamphlet was read and set down; the crowd in the square eventually went home. There were hours, days, whole seasons in which a person was returned to himself, to the slow rhythms of ordinary life, where the tribal reflex was not being actively triggered and rewarded. The feed has abolished the downtime. It is in the pocket, available at every idle moment, supplying a fresh trigger and a fresh reward whenever attention flags. The conditioning Chapter 3 described is not a thing that happens to us occasionally; it is a thing that happens to us continuously, in micro-doses, hundreds of times a day, each notification a small payout for the reflex, each idle scroll an opportunity to perform allegiance and be rewarded for it. A behavior reinforced on that schedule does not fade. It deepens. We keep paying because we are being paid, all day, to keep paying — and the currency is the most powerful one the social animal knows: the sense of belonging to the righteous group, dispensed in endless small denominations by a machine that profits from our spending.

Why seeing it is not enough — but is still necessary

I have stacked these reasons deliberately, because together they explain something that would otherwise be baffling: why simply showing people the mechanism — which is, after all, what this book has spent eight chapters doing — is not sufficient to free them from it. The mechanism feels like virtue, the alternative is genuinely hard, intelligence deepens rather than dissolves the capture, and the reward never stops. Against all of that, mere insight is a slight thing. Many readers will finish this book agreeing with every word of it and return, within the hour, to performing the exact reflex it describes — now equipped, perhaps, with a sophisticated understanding of relocated deference that they will deploy mainly to diagnose their enemies.

I do not say this to despair, and the next two chapters are my argument against despair. I say it because any honest account of how to resist the mechanism has to begin by admitting how weak insight alone is against it. Understanding relocated deference does not exempt you from it. It cannot, because the mechanism does not run on misunderstanding. What understanding can do — the only thing it can do, and the necessary first thing — is make visible the moment of choice that the mechanism is designed to hide: the instant when a charged claim arrives and you are about to know where you stand before you know what is true. Seeing that moment for what it is does not free you. But you cannot resist a moment you cannot see, and so seeing it, while not sufficient, is the indispensable beginning. What to do with the moment once you can see it is the subject of everything that remains.

The Discipline — Individual Tools Against the Reflex

The previous chapter made the case that insight is not enough, and I will not now contradict it by offering, as so many books like this do, a tidy list of mental habits and the implication that adopting them will set you free. The mechanism does not run on bad habits any more than it runs on ignorance. It runs on reward, on identity, on the deep machinery of belonging, and against those forces a list of tips is a slingshot against a tank. So I want to be honest at the outset about what this chapter can and cannot do. It cannot give you a technique that makes you immune. There is no such technique. What it can do is describe a discipline — not a trick but a practice, effortful and lifelong and never completed — and equip you to recognize the specific moments in which that discipline is the only thing standing between you and the reflex. The goal is not immunity. The goal is the recovery, in particular moments, of the thing the mechanism takes: your own judgment, briefly, when it counts.

The moment to watch for

Everything begins with the moment the previous chapter ended on, because a discipline you cannot trigger is no discipline at all. The mechanism's whole power lies in operating below notice — in delivering you to a conclusion before you have registered that any judgment was made. So the first and most important skill is simply learning to feel the moment it happens.

It has a texture, once you know to look for it. A charged claim arrives — about a case, a group, an event, a person — and you notice in yourself an immediate pull toward a position, a sense of where you stand that arrives before you have examined anything, often before you even know the relevant facts. That pull, that premature certainty, is the mechanism engaging. It is the precise sensation of your judgment being relocated, and it is detectable, if you are willing to feel for it, in the small gap between encountering a claim and "knowing" what to think about it. The discipline begins there, with a single learned reflex of your own: when you feel that pull, treat it not as insight but as a warning. The stronger and faster the certainty arrives on an identity-charged question, the more reason to distrust it, because speed and heat are the signatures of deference, not of judgment. Judgment is slow and cool. The thing that arrives hot and instant is the tribe, answering for you. Learning to feel the difference is the whole foundation, and it is harder than any of the specific practices that follow, because it asks you to be suspicious of your own certainty at exactly the moment certainty feels most like clarity.

Four practices

On the foundation of that recognition, four practices do real work. I offer them not as a checklist but as the recoverable forms of the four things observed earlier in the societies that had, for a time, caged the reflex: slowness, cost, contact, and the insistence on individual fact. Each is a way of supplying, by deliberate effort, the conditions the feed has been engineered to remove.

Slowness. The mechanism operates at speed; the discipline operates by refusing speed. The single most powerful thing an individual can do against relocated deference is to introduce a delay between the arrival of a charged claim and the formation of a position on it — to make it a personal rule that on exactly the questions where you feel the hot pull of premature certainty, you do not get to have an opinion yet. Not "I disagree," not "I agree," but "I do not yet know, and I am not entitled to know until I have done the work." This is harder than it sounds, because the feed punishes the person without an immediate take and rewards the person with a fast one, and because the suspended state is genuinely uncomfortable. But the delay is where judgment lives. Every relocation of deference happens in the instant of premature certainty; deny the mechanism that instant, hold the question open past the moment it wants to close, and you have done the one thing that most reliably interrupts it. Slowness is not a virtue here. It is a weapon.

Cost. Recall the finding, discussed earlier, that the agentic state collapsed whenever responsibility was returned to the subject. The individual analogue is to deliberately reintroduce, into your own thinking, the cost the mechanism removes — to hold yourself personally answerable for the truth of what you affirm, as if you would have to defend it, alone, to someone who would not accept your tribe as evidence. Before you pass along the claim, ask: do I actually know this, or do I merely know that my side holds it? Would I stake something on it? Could I defend it to a fair-minded person from the other tribe without appealing to the tribe? This is uncomfortable precisely because it reintroduces the personal responsibility the agentic state exists to dissolve, and that discomfort is the sign it is working. A belief you would not stake anything on is a belief you are holding on the tribe's authority, not your own, and naming that to yourself — honestly, in the moment — is one of the few things that breaks the spell.

Contact. The closed loop described earlier depends on the out-group remaining an abstraction, an avatar, a category. It cannot survive sustained, genuine, individuated contact with actual members of the other tribe — not the caricatures the feed supplies, but real people, encountered as people, in enough depth that they cannot be collapsed back into the category. This is why the discipline has an irreducibly social dimension that no amount of private clear-thinking can replace. Seek out the most intelligent, good-faith people who disagree with you, and read them, listen to them, sit with their strongest arguments rather than the weak ones your tribe enjoys mocking. The goal is not to be persuaded by them; it is to make it impossible for you to hold the comfortable belief that the other side is simply stupid or evil, because you have encountered, personally, the version of it that is neither. A loop that has been punctured by real contact with a real person it cannot dismiss is a loop that can no longer fully close, and a mind whose out-group is populated by actual humans rather than avatars is a mind the supercharger has lost much of its grip on.

The insistence on individual fact. The last practice is the most directly opposed to the book's title, and the most demanding. It is the refusal to let facts be tribal property — the insistence, held against your own side as fiercely as against the other, that what is true is true regardless of whom it helps. This means, concretely, being willing to affirm a fact that wounds your tribe and to doubt a claim that flatters it, by the same evidentiary standard you would apply in reverse. It means noticing when you are believing something because it is useful and holding that belief to a higher bar precisely because it is useful. It means treating the fact that a claim helps your side as a reason for more scrutiny, not less — the opposite of the mechanism's instinct. This is the hardest practice because it is the one that most directly courts the feeling of treason that the previous chapter described, and there is no way around that feeling; there is only through it. The discipline asks you to be willing, at the necessary moments, to be loyal to the fact over the tribe, and to bear the tribe's displeasure when those two loyalties conflict. Everything else in this chapter is preparation for that single willingness, which is the irreducible core of resisting relocated deference and cannot be made comfortable.

The honest limits

I want to close by guarding against a misreading that would undo the chapter's purpose. None of this makes you the exception. The reader most likely to be harmed by this chapter is the one who finishes it feeling equipped — who now possesses a vocabulary of slowness and cost and contact and individual fact, and who will deploy that vocabulary mainly to demonstrate that he practices these things while his enemies do not. That move, the conversion of a discipline of self-suspicion into a new instrument of tribal superiority, is the mechanism reasserting itself through the very tools meant to resist it, and it is the most likely outcome of reading this chapter badly.

So the final practice, the one that governs the other four, is to apply all of them inward first and hardest. The slowness, the cost, the contact, the insistence on fact — these are not tests you administer to the other tribe. They are tests you administer to yourself, especially at the moments when you are most certain you have passed them. The discipline is not a stick to beat your opponents with. It is a constant, uncomfortable, never-finished labor of distrusting your own certainties in proportion to how good they feel — and the person who has truly taken it on is marked not by how confidently he diagnoses others but by how readily he catches himself. There is no graduation. There is only the practice, returned to daily, against a mechanism that never rests and is never, in this life, fully beaten. That is not a counsel of despair. It is the only honest description of what resistance actually is: not a victory, but a discipline; not freedom from the reflex, but the recovered capacity, in the moments that matter, to choose your own judgment over the tribe's — which is, it turns out, the whole of what was being taken, and the whole of what is worth getting back.

CHAPTER 11

The Other Half — Structural and Institutional Responses

The previous chapter was about what an individual can do, and it was honest about the limits: the discipline it described is real, but it is also rare, effortful, and easily corrupted into a new form of the disease. If the only defense against relocated deference were the personal heroism of millions of people each separately choosing the hard, lonely, friendless path against a machine engineered to punish exactly that choice, the situation would be close to hopeless, because that is not how populations behave and it is not fair to ask them to. Individual discipline is necessary. It is nowhere near sufficient. The reflex is produced and amplified by structures, and structures are the only thing that can meet it at scale.

This is, in a sense, the optimistic chapter, and the optimism rests on a single observation from Chapter 6 that I now want to cash. The reflex is ancient, but its current intensity is not — it is the product of specific, recent, human-made arrangements, and the entire argument of Chapter 3 was that these arrangements are choices, not laws of nature. A thing that was built can be built differently. The feed that farms the agentic state was designed to optimize a particular quantity; it could be designed to optimize a different one. Nothing about the technology requires what it currently does. That is the ground for hope, and it is firmer ground than the hope that everyone will simply become wiser.

The machine could be built to lose

Begin where the leverage is greatest, which is the architecture itself. Chapter 3 established that the platform's pathologies follow from a single design decision — the choice to optimize for engagement — and that this choice systematically selects outrage, rewards its performance, and inflates the apparent hostility of the out-group. Each of those is a consequence of the optimization target, not of computation as such, and each could be different under a different target.

The recommender that ranks by engagement could instead rank by some measure of what users, on reflection, actually value — and the relevant research, recall, found that engagement-ranked political content left users less satisfied than a chronological feed, which means optimizing for reflective satisfaction rather than reactive engagement would not merely be more humane but would serve users better by their own stated lights. The platforms know how to do this; the obstacle is not technical. The reward loop that conditions users to escalate outrage could be dampened by designs that do not convert social approval into raw distributional advantage for the most inflammatory content. The hall of mirrors that makes the out-group appear more hostile than it is could be corrected by surfacing representative rather than extreme expressions, by showing people what the other side actually, modally thinks rather than what its angriest fringe performs. None of this is utopian. All of it is within the current engineering capacity of the firms involved, and the reason it is not done is not that it cannot be done but that the current design is more profitable. That is a fact about incentives, and incentives are exactly what institutions exist to change.

I am not naive about the difficulty. The firms will not adopt designs that reduce engagement voluntarily and at scale, because their revenue is engagement, and a single firm that unilaterally de-optimized would lose to competitors that did not. This is a collective action problem of the classic kind, and collective action problems are not solved by asking the actors to be better. They are solved by changing the rules all actors face at once — which is to say, by the one set of institutions with standing to set such rules. The structural response to an engineered pathology is, unavoidably, partly a matter of policy, and the relevant policy questions — transparency requirements that let researchers see what the algorithms actually do, design standards, liability rules that internalize the externalities the current model offloads onto society — are real and tractable and beyond the scope of a book about mechanism rather than legislation. I note only that the lever exists, that it is large, and that pretending the problem is purely individual is itself a way of leaving the largest lever untouched.

Defending the truth-finding institutions

The second structural front is the defense of exactly the institutions Chapter 8 watched the mechanism corrode: the courts, the press, the universities, the slow expert processes by which a society converts contested claims into settled facts. Relocated deference attacks these by tribalizing them — by reclassifying each as the property of one side so that its findings can be dismissed wholesale by the other. The structural response is to make them harder to tribalize, which is a thing institutions can actually work at.

An institution becomes tribalizable when it gives the opposing tribe a fair-seeming reason to dismiss it — when it appears, whether or not it is, to be acting as a partisan rather than as a neutral finder of fact. The defense, therefore, is a relentless and visible commitment to the practices that make tribal capture implausible: transparency about method, scrupulous evenhandedness, the public correction of one's own errors, the conspicuous willingness to reach conclusions that displease one's perceived allies. A jury system defends itself by visible fairness of process. A press defends itself by being seen to follow facts that embarrass every side, not merely the convenient one. A university defends itself by hosting the arguments its own dominant tribe finds uncomfortable rather than suppressing them, because every suppression hands the other tribe exactly the evidence of capture that licenses wholesale dismissal. These institutions cannot stop a determined tribe from declaring them enemies, but they can decline to supply the ammunition, and much of the ammunition currently in use was, in fact, supplied — by institutions that let themselves be seen taking sides and then wondered why half the country stopped believing them. Restoring trust is slow and asymmetric and unfair, but its direction is known: it runs through visible, costly, repeated demonstrations of the very evenhandedness the tribe is always tempted to abandon.

The longest lever: education

The third front is the slowest and, over a long enough horizon, the most powerful: what we teach people to do with their own minds before the machine gets a full grip on them.

The instinct here is to call for "media literacy" or "critical thinking," and the instinct is right in aim and usually wrong in content, because the standard versions teach a skill the mechanism easily survives. Teaching students to spot the other side's bias, to detect misinformation, to evaluate sources — all of this, deployed by a person inside the agentic state, becomes simply another weapon for defending the tribe: the motivated reasoner with media-literacy training is not less captured but better armed, more able to debunk the inconvenient and authenticate the convenient. We have, in effect, been teaching the equivalent of Chapter 2's sophisticated rationalizer and calling it critical thinking. The thing that actually inoculates is different and harder to teach: not the skill of catching others' errors but the disposition of catching one's own — the habit of self-suspicion, of distrusting one's own certainty in proportion to how good it feels, of treating the comfort of a belief as a reason for scrutiny. This is closer to a character trait than a skill, and character is harder to teach than technique, but it is not unteachable, and the societies of Chapter 6 that held the reflex in check did so partly by cultivating exactly this: an expected, normal, socially reinforced habit of intellectual humility about one's own side. That is a thing a culture can prize or fail to prize, can model or fail to model, can pass on or let lapse, and it is currently, on the whole, lapsing — replaced by an education and a media environment that reward the confident tribal take over the uncertain honest one. Reversing that is generational work. It is also the only intervention that addresses the disease at its actual root, which is not what people believe but how they hold their beliefs.

The accountability that the fog dissolved

There is one more structural response, and it follows directly from the book's central diagnosis. Relocated deference destroys responsibility by diffusing it into a fog that cannot be located or held. The structural countermeasure is to rebuild, wherever possible, the locatability that the fog dissolved — to reattach consequences to individuals.

This is why efforts to hold specific, identifiable individuals accountable for acts committed inside an anonymous crowd are structurally on point: they cut directly against the mechanism that such a crowd exists to enable, which is the dissolution of the individual into a mass that no one can be charged with anything. The principle generalizes well beyond that case. Any structure that reattaches responsibility to a locatable person — that makes the anonymous accountable, that ensures the act discharged in the tribe's name has an author who can be named — works against the core of relocated deference, because the core of relocated deference is precisely the escape from authorship. This must be done with care, because the same tools can be abused, and a society that reattaches responsibility recklessly does its own kind of damage. But the direction is sound and it is the structural mirror of the individual practice of "cost" from Chapter 10: just as the individual resists the reflex by reintroducing personal answerability into his own thinking, a society resists it by reintroducing personal answerability into its arrangements, refusing to let the fog be an alibi.

What structures can and cannot do

I will end where the chapter began, with a boundary. Structures cannot make people good, and they cannot, by themselves, produce the individual discipline of Chapter 10; a perfectly designed platform full of people determined to defer to their tribes would still produce relocated deference, merely more slowly. The two halves are not alternatives. The individual discipline without structural support is heroism asked of millions and reliably supplied by few. The structural reform without individual discipline is a better-built environment that still-undisciplined minds will find new ways to misuse. Each needs the other. The structures lower the cost and the frequency of the reflex, making the individual discipline possible for ordinary people rather than only for saints; the individual discipline supplies the demand for better structures and the will to build them. Neither alone is enough, and the recognition that neither alone is enough is itself a small act of the very evenhandedness the whole book has been arguing for — a refusal of the two tribal simplicities, it's all the platforms and it's all personal responsibility, in favor of the harder truth that it is both, intertwined, and addressable only together. That is not a slogan and it will not fit on a flag, which is, perhaps, the surest sign that it is true.

The Fact That Belongs to No One

I began this book with a man on a bicycle who punched a stranger he had never met, for a verdict he had not read, in the name of a cause he understood only as a side. I want to end with him, because everything the book has argued is contained in that single absurd and frightening gesture, and because the distance between how that gesture looks at the start of the book and how it looks now is the distance the book was written to travel.

At the start, the punch reads as madness — an isolated lunatic, dismissable. By now it should read as something far more disquieting: not an aberration but a concentrated, visible instance of an operation that runs, quieter and unpunched, through millions of people who would never raise a hand and who are doing, in their calm certainties, the very thing the cyclist did with his fist. He substituted a category for a person and acted on the substitution. So did the donors who gave after the verdict, the framers who knew the story before the trial, the celebrants who made a man a hero and the accusers who made him a monster, the swarm that dissolved a thousand choices into a crowd with no author. The cyclist is not the exception to this book's subject. He is its emblem — the agentic state with its skin removed, showing the skeleton that wears, in everyone else, a more respectable face.

What the cases held in common

Set the cases side by side one final time and extract what they share, because the commonality is the whole argument in miniature.

In each, a slow, evidence-bound institution did the difficult work of establishing what happened: a jury sat with the record and returned a finding. In each, a large public for whom that finding was unwelcome declined to engage the record and instead delegated the verdict upward to the tribe, to be overruled on grounds that required no contact with the evidence. In each, the public for whom the finding was welcome performed the equal and opposite error, converting a narrow finding into a sweeping vindication of its narrative. And in each, the most articulate participants supplied the rest with sophisticated rationalizations that let the deference feel like independent thought. Different defendants, different groups, different verdicts, opposite tribes — and underneath, one identical mechanism, indifferent to the content it ran on, running the same in everyone who let it run. That indifference is the finding. The mechanism does not belong to a side. It belongs to the species, and it has been handed, in our moment, a machine that feeds it without rest.

I have tried throughout to hold this with the discipline I was asking of the reader, which meant refusing every invitation to make the book an indictment of a tribe rather than of a reflex. The invitations were constant, and some of them were tempting, and I want to say plainly why I declined them. A book that located the disease in a people would have been more satisfying to one tribe and easier to sell to it, and it would have been false, and its falseness would have been the very mechanism it claimed to oppose, wearing the author's face. The only honest version of this argument is the one that implicates everyone capable of belonging to a group, which is everyone, including the one writing it and the one reading it. If this book has done its work, it has not given you a weapon against your enemies. It has given you a mirror, and the mirror is the point.

The reasons not to despair

I have been bleak, and the bleakness was earned, and I will not retract it. But the book does not end in despair, and the refusal of despair is not sentiment; it rests on things the argument actually established.

The first is that the reflex is not destiny. Chapter 6 showed that the same record that contains the recurring machinery also contains long stretches in which it was caged — periods and places where people of different tribes lived in functional peace because the structures and norms that check the reflex were, for a time, built and maintained. We have done it before. The capacity is in us alongside the reflex, and the historical evidence that the cage can be built is as real as the evidence that the reflex exists. Fatalism about this is not realism; it is its own kind of motivated reasoning, a deference to the comfortable story that nothing can be done.

The second is that the current intensity is the product of choices, not laws. The machine that farms the agentic state was built to optimize a quantity, and it could be built to optimize another; the research even suggests the humane design would serve people better by their own reflective lights. The arrangements that have made the reflex so much fiercer than it needs to be are recent, human, and alterable. We are not fighting gravity. We are fighting a business model and a set of habits, and both have been changed before.

The third is the smallest and the most important, because it is the one available to you tonight, without anyone's permission. In any given moment, when a charged claim arrives and you feel the hot pull of premature certainty, you can decline it. You can hold the question open past the instant it wants to close. You can ask whether you actually know the thing or merely know your side holds it. You can let a fact that wounds your tribe be true. None of this is easy and none of it is permanent and you will fail at it often, as will I. But it is possible, in the moment, every time, and the sum of those moments — across enough people, reinforced by enough better structure — is the only thing that has ever held the reflex in check, and it is enough, because it has had to be enough before.

The buck can stop here

The title of this book names the disease: the buck stops elsewhere, always elsewhere, relocated to an authority above or a fog all around, anywhere but with the person who could actually be held to it. The whole catastrophe is contained in that relocation — the endless handing-away of the responsibility to judge, to know, to own one's own conclusions, until no one anywhere is answerable for anything and the shared world dissolves into warring certainties that cannot be reconciled because they were never about the world in the first place.

But a buck that can stop elsewhere can also stop here. That is the entire wager of this book, and it is not a metaphor but a description of the only move that has ever worked. Here, with this person, in this moment, holding this fact — not because the tribe permits it but because it is true, and bearing the tribe's displeasure if true and tribe diverge. The relocation can be refused. Judgment can be reclaimed, one charged moment at a time, by anyone willing to pay the cost of owning it. It will not feel like victory; it will feel like standing slightly apart, slightly uncertain, slightly alone, holding something the crowd has no use for. That feeling is not failure. In an age engineered to relocate every judgment away from the person who makes it, that feeling — the discomfort of an opinion you actually own — may be the surest sign you have something left to lose, and have not yet handed it away.

A fact does not belong to a tribe. It never did, and it never will, however many of us come to feel that it does. The work of a free mind, now as ever, is to hold the fact that belongs to no one — and to refuse, in the moments that matter, to let the buck stop anywhere but where it always, finally, belonged: here, with us, with me, with you.

A note on sources

The scholarly works below are cited in full. The case studies throughout this book are composite illustrations: each is a generalized scenario assembled to isolate a recurring mechanism, and none corresponds to, or is intended to identify, any specific real event, proceeding, or person. Where the analysis describes patterns of online behavior, it treats them as general phenomena — types of fundraising activity, types of posting, types of public statement — rather than as the conduct of any identifiable individual. The composite cases are treated throughout as illustrations of a general mechanism; nothing in the analysis depends on the identity of any particular participant, and none is named or alluded to.

### Core theoretical sources

Arendt, H. (1963). Eichmann in Jerusalem: A Report on the Banality of Evil. New York: Viking Press.

Brady, W. J., Wills, J. A., Jost, J. T., Tucker, J. A., & Van Bavel, J. J. (2017). Emotion shapes the diffusion of moralized content in social networks. Proceedings of the National Academy of Sciences, 114(28), 7313–7318.

Brady, W. J., McLoughlin, K., Doan, T. N., & Crockett, M. J. (2021). How social learning amplifies moral outrage expression in online social networks. Science Advances, 7(33), eabe5641.

Burger, J. M. (2009). Replicating Milgram: Would people still obey today? American Psychologist, 64(1), 1–11.

Haslam, S. A., & Reicher, S. D. (2012). Contesting the "nature" of conformity: What Milgram and Zimbardo's studies really show. PLoS Biology, 10(11), e1001426.

Haslam, S. A., & Reicher, S. D. (2017). 50 years of "Obedience to Authority": From blind conformity to engaged followership. Annual Review of Law and Social Science, 13, 59–78.

Kahan, D. M. (2013). Ideology, motivated reasoning, and cognitive reflection. Judgment and Decision Making, 8(4), 407–424.

Kahan, D. M., Peters, E., Dawson, E. C., & Slovic, P. (2017). Motivated numeracy and enlightened self-government. Behavioural Public Policy, 1(1), 54–86. (First circulated as Yale Law School Public Law Working Paper No. 307, 2013.)

Milgram, S. (1963). Behavioral study of obedience. Journal of Abnormal and Social Psychology, 67(4), 371–378.

Milgram, S. (1974). Obedience to Authority: An Experimental View. New York: Harper & Row.

Reicher, S. D., Haslam, S. A., & Smith, J. R. (2012). Working toward the experimenter: Reconceptualizing obedience within the Milgram paradigm as identification-based followership. Perspectives on Psychological Science, 7(4), 315–324.

### On platform amplification and online perception

Milli, S., et al. (2025). Engagement, user satisfaction, and the amplification of divisive content on social media. PNAS Nexus, 4(3), pgaf062. (Earlier circulated via the Knight First Amendment Institute.)

Brady, W. J., et al. (2023). Overperception of moral outrage in online social networks inflates beliefs about intergroup hostility. Nature Human Behaviour. (On the systematic overestimation of others' outrage and subsequent conformity.)

Tufekci, Z. (2018). The great radicalizer: How recommendation systems steer attention toward extremes. (Op-ed essay on algorithmic recommendation.)

### On the replication and contestation of motivated numeracy

(Included in fairness to the contested status of the cognitive-ability findings, per the discussion in the Introduction and Chapter 2.)

Persson, E., Andersson, D., Koppel, L., Västfjäll, D., & Tinghög, G. (2021). A preregistered replication of motivated numeracy. Cognition, 214, 104768. (Finds support for a weaker version of the effect; notes the absence of consensus on its strength.)

Tappin, B. M., Pennycook, G., & Rand, D. G. (2020). Thinking clearly about causal inferences of politically motivated reasoning. Current Opinion in Behavioral Sciences, 34, 81–87.

### Note for expansion

The dissertation-length treatment of this material requires, in addition to the above: (1) a formally specified social-media corpus with stated collection dates, keywords, and coding protocol, rather than citation of individual posts; (2) full engagement-metric methodology; and (3) treatment of the comparison cases referenced in Chapter 7. These are noted here as the boundary between the present popular treatment and a peer-reviewable academic version.

Appendix A: Anatomy of the Central Composite Case

A neutral chronology of the composite scenario that anchors the book's argument. The scenario is a generalized template — a type of public dispute over a contested act of violence — and not a description of any specific event. Its particulars are deliberately abstracted so that the analysis turns entirely on the mechanism rather than on the identity of anyone or the specifics of any single case.

The template begins with a contested act of violence: an altercation between two private individuals that ends in a death, under circumstances disputed enough to support competing accounts. One party is detained and faces a charge; an account by investigators describes an apparent threat preceding the fatal injury, while the accused later advances a claim of self-defense. A pretrial release decision draws public attention before any trial is held.

In this type of scenario, a legal-defense fundraising effort follows. An effort on one general-purpose platform is removed under its policies; an effort on a different platform continues and grows. The fundraising language characterizes the accused sympathetically — as having been attacked first and as a person of good standing — a framing that those aligned with the deceased dispute as inaccurate and harmful.

Over the following period the dispute draws sustained online attention, and the wider group-identity dimensions invoked by partisans on each side are debated well in advance of any trial. Commentary hardens on multiple sides before the evidence is heard.

At trial, the prosecution presents multiple witnesses, including eyewitnesses, investigators, and a forensic examiner; most accounts describe the accused as the aggressor. The accused concedes the act but maintains it was justified, and does not take the stand. The jury rejects the justification defense, returns a finding of guilt on the most serious charge, and declines the alternative that would have reduced the penalty. A substantial sentence follows.

The fundraising effort, having reached a large total, continues to receive contributions after the verdict until the platform closes it, reasoning that a presumption-of-innocence rationale no longer applies once guilt has been adjudicated. Subsequent attention to how the raised funds were used generates further controversy, with the parties disputing competing characterizations. The point of the template is structural: at every stage, large publics substitute group allegiance for engagement with the record.

Appendix B: Coordinated Youth Gatherings — Selected Patterns

Background for the group-level case study in Chapter 5. The patterns below are described in general terms, as a recurring social phenomenon rather than as any specific catalogued incident; figures are indicative of the type of event, not a count of particular occurrences.

Definition. The phenomenon denotes large, loosely organized gatherings of adolescents coordinated primarily through social-media platforms, assembled with little advance notice, that in a recurring subset of cases deteriorate into reckless behavior, fighting, property damage, and occasionally weapons-involved violence.

Coordination mechanism. Such gatherings are organized and amplified through social media; footage of prior gatherings can function as promotion for subsequent ones. Automatically generated promotional graphics have been described as a near-frictionless organizing tool of the same general kind.

Representative patterns. In general terms, the phenomenon recurs in several typical forms: disturbances at commercial venues in which bystanders shelter while participants cause damage and improvised objects are used as weapons; large gatherings in entertainment areas that end in arrests; and gatherings in public spaces involving fights, contraband, and weapons that end in the detention of participants spanning a range of ages. Gatherings of this general kind have been described across many urban settings.

Policy responses observed. Curfews; increased presence at anticipated gathering points; detentions; and, in some places, measures aimed at parental or individual accountability for participants.

Note: this appendix would, in an academic version, require a systematic incident database with consistent inclusion criteria, sourcing, and date ranges. The above is illustrative of the phenomenon described in Chapter 5, not a comprehensive dataset.

Appendix C: Key Studies at a Glance

A reader's guide to the empirical spine of the argument.

Milgram (1963, 1974) — the obedience studies. Approximately 65% of subjects administered the maximum shock under a credible authority's prods. Critically, obedience varied sharply with experimental conditions: it collapsed when the authority was remote, when a peer defied, or when authorities disagreed — i.e., whenever responsibility was returned to the subject. Foundational evidence for the agentic state (Chapter 1).

Reicher, Haslam & Smith (2012); Haslam & Reicher (2012, 2017) — engaged followership. Reanalysis of Milgram's paradigm and new work arguing that the deepest obedience reflects active identification with the authority's cause, not passive submission. Notably, the most coercive prod ("you have no other choice") produced the least compliance. Basis for the "engaged followership" correction (Chapter 1).

Kahan, Peters, Dawson & Slovic (2017) — motivated numeracy. Subjects highest in numeracy reasoned most accurately on a neutral problem but, on an identity-charged version of the same problem, became more polarized rather than less — deploying quantitative skill selectively to reach group-congruent conclusions. Core evidence that cognitive ability can amplify rather than mitigate motivated reasoning (Chapters 2, 9). The effect's strength is contested in the replication literature; see References.

Brady et al. (2017, 2021) — moral outrage and amplification. Moralized-emotional content diffuses faster in networks; social reward for outrage expression increases subsequent outrage, consistent with reinforcement learning, with additional norm-learning effects. Basis for the platform reward-loop argument (Chapter 3).

Brady et al. (2023) — overperception of outrage. People systematically overestimate how outraged others are online, because amplified extreme expressions are unrepresentative — then conform to the inflated norm. Basis for the "hall of mirrors" argument (Chapter 3).

Engagement vs. satisfaction (PNAS Nexus, 2025). In controlled comparison, engagement-ranked political content drove more engagement but left users less satisfied than a chronological feed — evidence that the optimization target diverges from users' reflective preferences (Chapters 3, 11).