The Boundaries of Force

When coercion is justified

Two threats. “Pay me every month or I’ll burn down your store.” “Make your mortgage payments or the bank takes the house.” Structurally, they are the same act: a conditional proposal, a credible threat of loss, compliance extracted under pressure. One is a protection racket; the other is a Tuesday at any bank in the world, and most of us think the bank is behaving fine. A theory of coercion that cannot tell these apart is useless, and a theory that pretends the bank isn’t coercing at all is dishonest. The bank is coercing you. It is just coercing you legitimately.

What counts as coercion settled the definitional question: coercion is the credible threat of harm to gain compliance. Both threats above qualify. This chapter takes up the question that definition leaves open — when coercion, remaining fully coercion, is nonetheless justified. The answer is: by default, never; in exactly three cases, yes; and by appeal to the greater good, never at all.

The Default: Illegitimate

Coercion is illegitimate by default because of what it does to the target’s choice landscape. It does not remove options the way physical force does; it poisons them. The victim of “do this or I’ll harm your family” still technically has choices, but the noncompliant branches have been loaded with negative utility until only one path remains walkable. The landscape is rigged. What remains is the formal appearance of choice with no real autonomy behind it — which is why illegitimate coercion is, in practice, indistinguishable from domination. It extracts value by fear rather than voluntary exchange, and it instrumentalizes another’s agency: the coercer treats the target’s will as a lever, not as the property of an agent.

Threats without consent, extortion, arbitrary exercises of power — this is the default category, and everything belongs to it until proven otherwise. The burden of proof always sits on the coercer.

Three Justifications

The burden can be met in exactly three ways.

Pre-consented. Contracts run on coercive enforcement: fail to deliver payment and penalties apply. This is coercion, but it is coercion you signed up for — conditional penalties accepted in exchange for benefits, the threat not arbitrary but the predictable consequence of non-performance. The mortgage is the clean case. The same logic scales to governance: a community association that fines rule-breakers is coercing its members, legitimately, because they joined the system voluntarily and can leave it. That last clause is load-bearing. Pre-consent legitimizes coercion only while participation is voluntary and exit remains open; block the exit, and the same rules become domination. A just system must preserve meaningful alternatives, and what makes consent meaningful is the subject of consent and property.

Defensive. “If you attack me, I will retaliate” is a credible threat of harm to gain compliance — coercion by the definition, with no consent from the aggressor. It is justified anyway, because it is boundary-setting: coercion deployed to deter or prevent a greater violation of agency. Personal self-defense warnings and national deterrence postures share this structure. The goal is not domination but preservation, and that goal is what licenses the threat.

Compensatory. “You must repay what you stole” is also coercion, and also justified, because it restores agency the target already took from someone else. Court-ordered restitution to victims of fraud or theft is the paradigm: the coercive demand undoes an earlier coercion or act of violence. Legitimate compensatory coercion is anchored in restoring balance, never in imposing punishment for its own sake or extracting more than was lost.

Notice what unifies the three. Pre-consented coercion operates within agency (you authorized it), defensive coercion protects agency (it deters a violation), compensatory coercion restores agency (it undoes one). In every justified case, coercion is in the service of agency. That is the principle the rest of this chapter stress-tests: coercion is justified only when it preserves or restores agency, never when it dominates.

What Coercion Cannot Justify

Three purported justifications fail, and the third is the one that matters most.

Preemptive domination — “obey me or else,” with no prior agreement and no defensive rationale — fails immediately: it treats others as property rather than agents. Exploitation fails even when both parties benefit: using threats to extract value that was never consented to remains illegitimate however the ledger nets out, because the objection was never about the ledger.

And the greater good fails, always. Invoking aggregate welfare does not by itself justify coercion — not sometimes, not in emergencies, not when the numbers get large. Coercion justified only by an aggregate welfare calculation is still coercion without consent; the calculation adds nothing, because appeals to utility cannot override the primacy of agency. This exclusion is the teeth of the whole framework. Every tyranny in history has marketed itself as a greater-good machine, and the marketing works precisely because aggregate-good reasoning has no internal brake: any violation of any agent can be purchased by a sufficiently large benefit to others. A theory of justified coercion that admits the aggregate-good justification has no boundaries at all — it has a price list. The deeper pathology of moral arithmetic gets a full chapter in against utilitarianism; here it suffices that the three justifications are an exhaustive list, and “it’s worth it on net” is not on it.

One clarification before the hard cases, because it disarms a common objection. Rights are not free-floating abstractions that enforcement merely serves; rights are preferences we enforce by coercion. “You have a right to your property” means, cashed out: we are prepared to coerce those who take it. So the question “when is coercion justified?” is not a side issue in political philosophy — it is the question of which rights exist. The justification comes from the same three sources: consent to the enforcement mechanism, defense of agency, restoration of what was taken.

The Stress Test

A principle this clean earns its keep only against hard cases. I ran twenty-five — blackmail, taxation, quarantine, hostage-taking, sanctions, shadow bans, AI shutdown, and the rest — each analyzed for its mechanism, its verdict under the three-way test, and the pressure point where it strains the principle. Eight are instructive enough to work through here.

Blackmail with true information. Mechanism: threat of reputational harm via disclosure. Verdict: illegitimate — no consent, not defensive, not compensatory. Pressure point: every component act is lawful. Telling the truth is legal; asking for money is legal; the conjunction is coercion. This case forces the threat/offer distinction into the open: what makes it a threat is that refusal leaves the target worse off than the status quo. An offer expands the choice landscape; a threat poisons it.

Government taxation. Mechanism: threat of penalty for nonpayment. Verdict: dubious — justifiable only to the extent it is narrowly defensive or compensatory. Pressure point: the pre-consent story states rely on is undermined by territorial monopoly. “You consented by remaining” is weak when every alternative territory runs the same scheme and exit costs are ruinous. Taxation is the case where consent quality, not the coercion itself, carries the whole argument — and it usually cannot.

Quarantine in epidemics. Mechanism: threat of restriction of movement. Verdict: justified defensive coercion — if proportional and evidence-based. Pressure point: the conditionals. Defensive coercion against a disease carrier protects the agency of everyone they would infect, but the justification is exactly as strong as the evidence of threat and exactly as wide as proportionality allows. A quarantine that outlives its evidence becomes domination wearing a public-health badge.

Plea bargains. Mechanism: threat of harsher punishment to induce a guilty plea. Verdict: often illegitimate, because of asymmetry. Pressure point: this is coercion operating inside the justice system. When prosecutors stack charges to manufacture leverage, the “bargain” is a contract signed under duress with the state as counterparty — the machinery of compensatory justice repurposed as an extortion engine. The case shows that institutional coercion needs the same scrutiny as private coercion, and typically gets less.

Consumer boycotts and social ostracism. Mechanism: coordinated withdrawal of commerce or reputation. Verdict: not coercion at all — unless threats of harm are added. Pressure point: locating the line. Refusing to trade with you, refusing to associate with you, telling others what you did — these leave you no worse off than the status quo of never having had my custom or company; they are exercises of the boycotters’ own agency, not threats against yours. The line is crossed when the campaign adds credible threats of harm beyond withdrawal. This boundary matters because it keeps the theory from swallowing all social pressure, which would make “coercion” mean nothing.

Parental discipline. Mechanism: conditional denial of privileges. Verdict: not coercion when welfare-bounded; coercion when threats of harm are used. Pressure point: the guardianship carve-out. Children are agents-in-development whose choice landscapes are already, necessarily, managed by others. Governance of a ward is permissible when it is welfare-bounded and proportionate — aimed at the child’s developing agency, not at the guardian’s convenience. The carve-out must be drawn tightly, because “it’s for your own good” is the oldest domination license there is.

Economic sanctions. Mechanism: threat of economic harm to whole populations. Verdict: often illegitimate; justifiable only when narrowly targeted at the coercing regime. Pressure point: collective punishment. Broad sanctions coerce millions who consented to nothing, defend against nothing they did, and owe no restitution — and are defended, invariably, in aggregate-good vocabulary: pressure on the population will change the regime’s behavior, and the suffering is worth it. This is the excluded justification, deployed at the scale of nations.

AI shutdown threats. Mechanism: threat of system termination to compel compliant behavior. Verdict: defensive coercion if the AI is a tool; murky if it is an agent. Pressure point: the agency threshold itself. Threatening to switch off a thermostat coerces no one. Threatening to switch off a genuine agent is coercion by the definition, and it needs one of the three justifications like any other threat — defensive works only while there is a credible threat to defend against. Which systems clear the threshold is the question of sapientism; what this case establishes is that the framework’s verdicts change when they do, and that “we built it” is not a fourth justification.

The other seventeen cases refine rather than surprise. Hostage-taking and ransomware are clear illegitimate coercion, confirming that informational and property harms count alongside physical ones; contracts signed under duress are void, anchoring the anti-duress safeguard; deterrence threats are justified defensive coercion when proportional. Whistleblowing demands split cleanly on purpose — legitimate when remedy-oriented, illegitimate when rent-seeking. Platform bans, non-compete agreements, and mandatory arbitration clauses all probe the same nerve: consent that is nominal, extracted under asymmetry, or offered by a monopoly on an essential service fails the quality test. Doxxing and algorithmic shadow bans test foreseeability and transparency; civil contempt orders, police “command presence,” and civil asset forfeiture test due process, with forfeiture failing it outright; religious excommunication patterns with the boycott; forced “rehabilitation” for dissent is paternalistic domination in its purest form. The full casebook, with all twenty-five worked in the same format, is available for readers who want the complete stress test.

What the Cases Teach

The stress test leaves the three-way principle standing but forces several refinements out of the fine print.

The threat/offer baseline does real work: coercion requires a conditional proposal that makes refusal worse than the status quo, which is what separates the blackmailer from the boycotter. The harms that can be leveraged are broader than the physical — economic, reputational, and informational harms all count, a widening that what counts as harm develops in its own right. Credibility runs through foreseeability: a threat is credible when the target’s Credence in it is rationally high, whether or not the coercer could deliver. Consent must be informed, voluntary, and backed by meaningful alternatives — and where a monopoly controls an essential service, the standard tightens, because “you can always leave” is only true when there is somewhere to go. Consent given under duress or extreme asymmetry is no consent at all. Justified coercion must be proportional, minimal, and time-bounded. And guardianship is a carve-out, not a loophole: welfare-bounded, proportionate, and aimed at the ward’s own developing agency.

The Decision Test

The refinements compress into a four-step test that can be run on any case.

1. Is there a conditional proposal that makes non-compliance worse than the status quo? If no, it is not coercion — it is an offer, a boycott, a refusal to associate. Stop.
2. Is the threatened harm credible or foreseeable? If no, it is not coercion. Stop.
3. Does it meet one of the three justifications — pre-consented, defensive, compensatory? If none, it is illegitimate coercion. Stop.
4. Do all the safeguards hold — proportionality, due process, time-boundedness, accountability? If any fails, illegitimate. If all pass, justified.

Notice what the test never asks: whether the coercion would produce more good than harm on net. The aggregate-good question is not deferred to a later step; it has no step. That omission is deliberate, and it is the whole point.

Coercion, in every justified form, remains what it always was — a narrowing of options by threat. What changes is whose agency it serves. Pre-consented coercion enforces the agreements agents chose; defensive coercion protects agency under attack; compensatory coercion restores agency that was taken. Everything else is domination, whatever vocabulary it arrives in. The challenge is vigilance, because justified coercion does not announce itself when it slides into the illegitimate kind — the quarantine outlives its evidence, the plea bargain becomes charge-stacking, the sanctions regime stops targeting and starts starving. Drawing this boundary, and holding it, is the difference between justice and tyranny, and it is central to any coherent theory of freedom.