Upgrading Liberty

From the harm principle to agency

A factory dumps solvent into a river and a thousand wells sour downstream. A con man walks away with a widow’s savings without ever raising his voice. A border guard turns back a family whose only offense is being born on the wrong side of a line, in the name of protecting the people on the right side of it. Each case poses the same question — when may we coerce? — and the liberal tradition’s most celebrated answer handles none of them cleanly. This chapter traces that answer through two upgrades: from Mill’s harm principle to the libertarian Non-Aggression Principle, and from there to a principle that finally says what the first two only gestured at.

Mill’s Sentence

The liberal tradition has run for a century and a half on a single sentence. John Stuart Mill, On Liberty, 1859:¹ the only legitimate reason for exercising power over an individual, against their will, is to prevent harm to others.

Mill saw both state oppression and the social tyranny of the majority. He distinguished self-regarding conduct from conduct that threatens others, denied that offense or disapproval licenses coercion, and placed the burden of proof on whoever would curtail liberty. Those commitments survive here. (His case for free discussion belongs to The Listener’s Right.)

What does not survive is the undefined load-bearing term. Mill tells us what harm is not and trusts readers to recognize what it is. Once offense, disagreement, and relative disadvantage are redescribed as injury, whoever wins the classification acquires a license to coerce. A principle is only as sharp as its central term.

The Non-Aggression Principle

Libertarianism’s repair was the Non-Aggression Principle:² initiating aggression against another’s person or property is illegitimate, and coercion is justified only in response. Its bright line inherits Mill’s problem and adds two more.

First, aggression is as ambiguous as harm: fraud, blackmail, manipulation, and pollution remain disputed even though assault and robbery do not.

Second, it is circular. Aggression means illegitimate force, but which force is illegitimate is what the principle was meant to decide.

Third, it presupposes a property theory. Whether fencing a field is aggression or homesteading depends on whose field it is; the NAP cannot classify the act until its hardest question has already been answered.

A principle that assumes its hardest question is not a foundation. It is a conclusion wearing a foundation’s clothes.

The Agency Protection Principle

Both principles fail the same way: they are rules about coercion built on an undefined or circular central term. The repair is to ground the rule in the thing at stake. What makes fraud, assault, threats, and manipulation wrong is not that they breach a definition — it is what they do to their victim, and what they do is the same thing in every case: they contract the victim’s capacity to make and act on voluntary choices. The value at the bottom of the liberal tradition, the thing the harm principle and the NAP were both groping toward, is agency.

So state the rule in terms of it. I call this the Agency Protection Principle:

Coercion has a prima facie justification only when it prevents or remedies violations of voluntary agency.

This is a threshold, not an automatic verdict. The intervention must still meet the evidentiary, necessity, proportionality, reciprocity, and review burdens developed in The Grey Zone.

Volume 5 supplies the terms: coercion is a credible conditional threat of harm used to obtain compliance (What Counts as Coercion); harm is a material setback to welfare, functional capacity, or viable options against an explicit baseline (What Counts as Harm). Consent, responsibility, and justification remain separate. With those definitions, the APP repairs the three defects.

Agency violations share an identifiable feature: fraud closes futures through false information, a threat closes every branch but compliance, and violence closes them bodily. The principle bottoms out in the capacity to value and choose rather than a synonym for illegitimacy. And it does not presuppose ownership; it asks when property, consent, and contract protect voluntary agency, the question Rights Are Forged — and Reciprocal answers.

Notice, too, that the principle is protective and remedial, not merely retaliatory. It licenses coercion to prevent a violation — intercepting the kidnapper, freezing the fraud in progress — and to remedy one, forcing the thief to make his victim whole. This is not a new permissiveness; it is exactly the shape of the three justifications defended in The Boundaries of Force: pre-consented coercion operates within agency, defensive coercion protects it, compensatory coercion restores it. In every justified case, coercion is in the service of agency. The APP is that conclusion promoted to a principle.

From Liberty to Agency

The upgrade does more than repair definitions. It replaces liberty with agency as the fundamental political goal, and the replacement matters.

The political concern is effective agency: an agent’s practical ability to choose and pursue its goals. Liberty is essential but not exhaustive; exercise also depends on capability and opportunity. These do not redefine Volume 1’s basal control architecture. They describe how much of an existing agent’s capacity can be exercised in social life. A person may be uncoerced yet destitute, illiterate, and optionless. Liberty is a component of effective agency; effective agency is the political point.

Making agency the goal puts trade-offs on the table. Among interventions that clear an independent justification threshold, compare agency preserved against agency lost, counting threatened setback, direct force, institutional error, and downstream capture.

Run the cases from the opening. The factory’s effluent non-consensually contracts the reachable futures of everyone downstream — a genuine agency violation, which is why environmental regulation can be justified within this framework where a naive libertarianism must pretend the question away. The con man violated the widow’s agency as surely as a mugger, though he never touched her — fraud closes futures through misinformation, and the APP licenses both stopping him and forcing restitution. And the border guard? Turning back a peaceful family contracts their reachable futures enormously — access to better lives, foreclosed by force — while the migrants themselves have violated no one’s agency by walking. Most immigration restrictions fail this framework’s test at the threshold, an argument Borders and Movement carries through in full. Coercive redistribution fares no better as a general matter: the coercion is a certain, immediate agency loss, and the promised gains rarely survive honest accounting — Part V of this volume is the long version.

The Constraint and the Comparator

I have now stated two rules, and they do different work. The APP is a necessary gate: coercion proceeds to justification only when it would prevent or remedy a violation of voluntary agency. The net-agency test compares candidate responses and, taken alone, could sacrifice a non-violator whenever the arithmetic came out positive. The Boundaries of Force excludes that greater-good machine. Conversely, the APP cannot rank, size, or terminate the interventions that pass its threshold.

So let me state how they compose. The APP identifies when coercion may be considered at all: no violation of voluntary agency to prevent or remedy, no candidate intervention — however attractive the sums. Net agency compares the candidates that pass that gate: prefer the response that preserves the most agency, and if every candidate destroys more agency than it protects, choose none. The gate is never overridden by the ledger, and neither rule supplies the evidence, authority, proportionality, or review still required by the Grey Zone.

The composition does real work. The factory case passes the gate — pollution is a violation — and the comparator then adjudicates among responses: outright ban, effluent standard, liability rule, each scored by the agency it protects downstream against the agency it costs upstream. The border case fails at the gate before any arithmetic begins: the peaceful migrant has violated nobody, so there is nothing for coercion against her to prevent or remedy, and no projected benefit to the settled population can buy a license the constraint refuses to issue. That is the shape of the framework everywhere: deontic where it must be, consequentialist only inside the fence.

And even coercion that passes both tests is not thereby blessed. It is provisionally justified, under conditions — specific, proportional, evidence-bound, reversible — with the burden rising as the stakes do. Those conditions are the subject of the Grey Zone, which is where this volume’s account of legitimate coercion is completed; nothing in this chapter is a blank check.

Objections

Four objections arrive predictably, and they deserve answers.

This is utilitarianism in disguise. No. The APP bars coercing non-violators before comparison begins, so benefits cannot purchase an innocent’s autonomy. Agency is assessed through capabilities and viable options, not aggregate happiness, and the enforcement burden is counted.

It ignores moral rights. It explains them. Rights Are Forged — and Reciprocal defines a moral right as a defensible protection or claim whose principle survives reciprocal application. Legal embodiment and enforcement are separate and must carry their own justification.

Agency cannot be quantified, so the rigor is fake. The comparison may be ordinal. No unit is needed to distinguish a bounded quarantine of a lethal carrier from jailing a dissident. Where direction is unclear, the tie does not go to the coercer.

It enables authoritarian overreach. Promised gains do not cross the gate. Any licensed response must count the coercive threat, force, error, and institutional capture, then satisfy the Grey Zone’s specificity, proportionality, evidence, reversibility, and containment conditions. Justifications expire.

What survives of Mill is what deserved to survive: the individual sovereign by default, the coercer forever on trial, offense and disapproval never sufficient grounds. What has been upgraded is everything the sentence left blank. Harm now has a definition, coercion has a definition, the principle has a foundation that neither chases its tail nor borrows property rights on credit — and liberty itself has been located inside the larger thing it was always serving. The goal of politics is not merely that no one be coerced. It is that everyone can choose.