Compatibilism, Incompatibilism, and the Missing Definition of Agency

Two questions are constantly fused. They should not be.

1. Compatibility. If the world were deterministic, could there still be free will, moral responsibility, agency?
2. Fact. Is the world deterministic — or deterministic enough — at the scale of human action?

(1) is a question in modal metaphysics and action theory. (2) is a question for physics and the special sciences. Classical mechanics suggested a yes to (2) for Laplace; quantum theory reopened (2) without automatically answering (1). Indeterminism at the micro-level does not by itself give an agent control; that is the luck and rollback family of arguments (van Inwagen 1983, ch. 4; van Inwagen 2000; Mele 2006). Determinism at the macro-level does not by itself prove that “could have done otherwise” is empty; that is the compatibilist counter.

Determinism forbidding free will is not yet a physics debate. It is a debate about what “forbid” and “free will” mean. Physics can change the available models of causation. It cannot, by itself, pick the sense of “agency” that ethics, law, and political economy need.

A third question is not a substitute for those two, and is the reason this paper cannot stay inside seminar metaphysics:

1. Institution. What do political and economic structures do to the unit that (1) and (2) have not finished defining?

Authoritarian command, representative filters, absentee title, and occupant voice are not proofs of free will. They are observed differences in how far a structure treats persons as sources of their own transitions rather than as ensembles — class, ticker, party, “the people” as a mass. More room is not a solved metaphysics. It is still more room. Center libertarianism (CenLib) is a program to maximize that room in government and in the economy. That is a motivation, not a deduction from van Inwagen.

The Individual Completion program (Shields 2026a; Shields 2026b), written after Einstein’s 1949 Reply, is prospected here as a candidate completion of the individual: before a choice, an unread remainder, so the future of the act is probabilistic at the coarse grain; at the exact time-instance of the act, an agent, and an outcome from the choice. That is not a solution to the Consequence Argument, not a constitution, and not a claim that persons have been shown to work this way. Section 7 names the mapping and dates it against the record. It stays marked as open.

1. Definitions that must not drift

Causal determinism (CD). The past plus the laws fix a unique future. Laplacean in spirit; nomological in the analytic literature. Not the same as “caused,” and not the same as “predictable by us.”

Free agency / FA. The thesis that agents sometimes act freely, or are morally responsible, in a sense that is not a joke. The paper will not pretend this sense is already shared.

Leeway. Alternative possibilities at the moment of action: the agent could have done otherwise.

Sourcehood. The action issues from the agent as origin, even if alternatives are thin.

Libertarian (1789). Belsham’s label for the anti-necessitarian — a claim about agency, not about markets. Target, not compliment. Date it or collide with Déjacque 1857 and with the US party after 1971. See Fractures for the left-hand capture; Foundations of the Right for the American inversion (Russell 1955).

Agency (working, this program). A temporally extended individual system is an agent to the extent that it can prospect possible transitions compatible with its constraints and then be the source of some of its own state-transitions at a time-instance, such that those transitions are attributable to it rather than only to an external commander, a biological imperative, or a statistical ensemble. Two families bound the fiber of those possible transitions without exhausting the individual: institutional constraints, and biological / ecological resource limitations. Those ecological bounds are real; they are not a law that agents only fight over two apples. Cooperation can plant. Mutual aid is why the economic models borrowed a biology that was not closed; this paper keeps the unit from being erased into gene, class, or ticker. Before the act, the future of the choice is an unread remainder — probabilistic at the coarse grain. After the act, there is a time-instance and an outcome from the choice. This is a constraint on later work, not a finished theorem. It is written to be testable against logic (does not already pick a winner in the compatibility debate), history (sits in a dated line of uses), and observation (structures differ in the room they leave; biological command is not exhaustive). The unread-remainder / time-instance clause is prospected from an adjacent physics program (Shields 2026a, after Einstein 1949). It is not closed.

Institutional agency. A related, not identical, quantity: the extent to which a political-economic order leaves that capacity room — voice, exit, title, non-domination (Pettit 1997) — rather than converting persons into ensembles. CenLib is a bid to maximize this. It does not prove FA.

Unread remainder / first-event time (prospected). From Individual Completion (Shields 2026a): a complete individual state S fixes a first-event time T(S); a description that writes only a coarse cell S_coarse leaves a remainder S_⊥, and T is then distributed over that fiber. Mapped onto agency: before the choice, the future is that unread remainder; at the time-instance of the act there is an agent and an outcome. Prospected, not finalized. Not a hidden-variable dressing of the wavefunction, not primitive chance, and not a proof of FA.

What this is not. It is not “more votes = more free will.” Ochlocracy and agenda capture can eat a franchise. It is not “the market is freedom.” Absentee title is a legislature over other people’s time. It is not Berlin’s negative liberty restated as a slogan, and not Marx’s species-being restated as a proof.

2. Map of positions

Let CD and FA be as above.

The labels “compatibilism” / “incompatibilism” are usually credited to Keith Lehrer in the 1960s, for the classical-analytic paradigm (roughly 1965–1985), which treated the problem as: can an ordinary human exercise the ability to do otherwise in a deterministic world? Lehrer’s first free-will paper is 1960; the coinage is “most likely” his, not a signed christening. 1

Split inside incompatibilism. Libertarian incompatibilists keep FA and drop CD (Reid; Kane; van Inwagen as incompatibilist who is not a hard determinist). Hard determinists keep CD and drop FA (d’Holbach; some readings of Spinoza; contemporary “willusionists”). Hard incompatibilists say neither determinism nor typical indeterminism yields the control that basic-desert responsibility needs (Pereboom, Living Without Free Will, 2001). Strawson’s reactive-attitudes line (1962) and later revisionism sit nearby: we may keep practices of holding responsible without the metaphysics the practices seem to need.

Split inside compatibilism. Classical / leeway: freedom = absence of external constraint plus conditional ability (“could have if one had chosen otherwise”) — Hobbes, Hume, Moore, Ayer. Source / identification: freedom = action issues from the agent’s own standing desires, higher-order volitions, or reasons-responsive mechanism — Frankfurt 1971; Watson 1975; Fischer and Ravizza 1998. Semicompatibilism (Fischer): moral responsibility is compatible with CD even if the freedom to do otherwise is not. Some classical compatibilists held that freedom requires determinism, because otherwise action would not reliably follow character and reason (Hume, Enquiry VIII).

The 1789 word libertarian, in William Belsham’s Essays, Philosophical, Historical, and Literary, names the incompatibilist who asserts a self-determining power against necessitarians. Belsham uses it as a target, not a compliment. That is the original sense this program recovers before the 19th-century European left use and before the 20th-century US political capture of the word (Dean Russell, FEE, May 1955 — Foundations of the Right).

None of these cells writes a constitution. A hard incompatibilist can still refuse autocracy. A metaphysical libertarian can still build a shareholder oligarchy. The compatibility map and the institutional ladder are two boards. The widget on this page keeps them apart on purpose.

3. History, dated so the gaps show

The compatibility problem is not born in 1965. The analytic paradigm renames a fight that is already about definition: is freedom rational determination, the absence of chains, or the real power to have willed otherwise under the same decrees?

Ancient: voluntary, fate, swerve

Aristotle, Nicomachean Ethics III: the voluntary is what has its origin in the agent, who knows the particulars; the involuntary is from compulsion or ignorance. This is not Laplacean CD. It is already a source condition plus an epistemic condition. Praise and blame track the voluntary. Law still lives here.

Stoic fate plus “what is up to us” (eph’ hēmin; later Epictetus’s prohairesis, the faculty of choice that a tyrant cannot take) is often read as an ancestor of compatibilism: the world is fated; the use of impressions is yours. Epicurus’s atomic swerve is an ancestor of libertarian appeals to indeterminism — and already faces the luck problem: a random swerve is not authorship. Cicero, De Fato, is the Roman transmission of both.

The political observation is already available in this register. A tyrant can command the body and not the prohairesis. That is not yet a theory of assemblies. It is the first recorded insistence that some inner source is not identical with the commander.

Medieval: decrees, premotion, alternatives

Augustine, On Free Choice of the Will: evil is not from God; it is from the will. Later, in the anti-Pelagian writings, grace sits so close to the will that later readers smell determinism. The definition is already the fight.

Aquinas: the will follows rational apprehension of the good; freedom is rational determination, not indifference. Later Thomists (Báñez) add divine physical premotion. Scotus: the will has a real power of opposites in the same instant of nature. Molina: middle knowledge, so that divine foreknowledge does not collapse alternatives. This is already leeway versus source under theological CD. Is freedom rational appetite, or the power to will otherwise under the same decrees?

Early modern: Hobbes versus Bramhall, then Locke, Hume, Kant

Hobbes, Of Liberty and Necessity (1654), against Bishop Bramhall: liberty is the absence of external impediments, not the absence of causes. A river is free when it is not dammed; it is not free of gravity. Bramhall’s Defence of True Liberty (1655) is the royalist-theological insistence on a self-determining will. The Hobbes–Bramhall exchange is the English template of compatibilism versus libertarianism, a century before Belsham names the sides.

Locke, Essay II.xxi: the question is badly posed as “whether the will be free.” The agent is free; the will is a power. That sentence is load-bearing for this program. Later political economy that treats the firm or the party as the agent has already left Locke’s unit.

Hume, Enquiry VIII: the contrary of liberty is constraint, not necessity; we need necessity to make sense of praise and blame, because otherwise action would not follow character. This is the “freedom requires determinism” strand. It is also a political hint: Hume’s liberty is civil, not a swerve.

Kant, Critique of Practical Reason (1788): practical freedom as a postulate of the moral law. Not classical compatibilism (the empirical character is in the order of nature) and not an empirical proof of libertarianism. The noumenal will is not an institution. Date it so it is not used as a party slogan.

Spinoza, Ethics: “men think themselves free because they are conscious of their actions and ignorant of the causes.” d’Holbach, Système de la nature (1770): the hard-determinist popularization. Reid, Essays on the Active Powers of Man (1788): agent-causation against Hume. Belsham 1789 writes as a necessitarian and names the other side “libertarian.” The word that later names a US party begins as a metaphysics insult.

Political liberty is not the same literature, and cannot be ignored

Constant, “The Liberty of the Ancients Compared with that of the Moderns” (1819): ancient liberty is participation in collective power; modern liberty is independence from that power. A structure can enlarge one and shrink the other. Athens is not a template of success; it is a data point about voice versus privacy.

Mill, On Liberty (1859): the harm principle; individuality as a condition of human development. Mill is a compatibilist in the Humean line and a political liberal. The conjunction is not a deduction. It is a historical fact this paper needs: people have always designed institutions as if agency were real, while arguing about whether it is.

Marx, 1844 manuscripts: alienated labor — the product and the activity are not the worker’s; they belong to another. That is an agency claim about economic structure, whatever one thinks of the later method. The worker is treated as an ensemble input. See Fractures for goal versus method; do not smuggle 1917 into 1844.

Berlin, “Two Concepts of Liberty” (1958): negative liberty as non-interference; positive liberty as self-mastery, which can be captured by a party that claims to be your real self. Taylor (1979) and Skinner (Liberty before Liberalism, 1998) refuse the thinness of pure non-interference. Pettit (Republicanism, 1997): liberty as non-domination — not just non-interference, but not being subject to another’s arbitrary will even if that will is currently benign. A kindly master still dominates. That sentence is why absentee title and an unaccountable office are agency problems, not only distribution problems.

Arendt, The Human Condition (1958): action as beginning something among others in a public space; labor and work are not the same. The program does not need Arendt’s full ontology. It needs the observation that voice in public is a different species of transition from private consumption.

Sen, Development as Freedom (1999); Nussbaum on capabilities: development as expansion of what persons can do and be. This is not van Inwagen. It is the best-known contemporary statement that economic structure is an agency technology. Hirschman, Exit, Voice, and Loyalty (1970): two mechanisms for being a source rather than a captive. CenLib uses both — voice in assemblies, exit via counter-economy — and has not finished specifying either.

Classical analytic, c. 1965–1985

The field pivots from “do we have free will?” to “is free will compatible with CD?” Van Inwagen, An Essay on Free Will (1983), states the Consequence Argument. Frankfurt, “Alternate Possibilities and Moral Responsibility” (1969), attacks the Principle of Alternate Possibilities (PAP). Austin, “Ifs and Cans”; Moore’s conditional analysis; Ayer’s “Freedom and Necessity” are the mid-century compatibilist kit. Strawson 1962 tells the literature it may have the wrong target: reactive attitudes are not going to be argued out of existence by CD.

After Frankfurt: source, luck, manipulation, brain

If PAP falls, leeway incompatibilism loses its main bridge from “no alternatives” to “no responsibility.” Source incompatibilists rebuild the bridge: what fails under CD is origination. Kane (The Significance of Free Will, 1996) is the load-bearing event-causal libertarian: self-forming actions, efforts of will that are reasons-weighted and undetermined. The luck objection (Mele 2006; others) asks why the undetermined bit is theirs. Agent-causation (Chisholm; O’Connor, Persons and Causes, 2000) tries to make the agent a substance-cause. Hard incompatibilism (Pereboom 2001) says both roads fail for basic desert.

Manipulation arguments (Pereboom’s four-case; Mele) target source compatibilism: if a reasons-responsive mechanism was installed, is that still freedom? There is no consensus cutoff between ordinary socialization and covert control. That cutoff is the political problem in miniature. A party school, a recommender, a childhood, a law — which of these is “the history of the mechanism written by another”?

Neuroscience did not close it. Libet (1983) and later readiness-potential work were popularized as “the brain decides first.” Mele (Effective Intentions, 2009) and others: the data do not show that a decision was completed before awareness, and they do not show that veto is impossible. Wegner (The Illusion of Conscious Will, 2002) is a willusionist reading of phenomenology. Dennett (Elbow Room, 1984; Freedom Evolves, 2003) is a compatibilist who wants Darwinian room, not a swerve. Observation has not delivered a physics of FA. It has delivered constraints on cartoon versions of both sides.

4. The Consequence Argument

Informal statement (van Inwagen 1983, 56): if determinism is true, our acts are consequences of the laws of nature and events in the remote past. What went on before we were born is not up to us. The laws are not up to us. Therefore the consequences of those two things, including our present acts, are not up to us.

Formal core (Third Argument, 1983, 93–95). Let Np = “p is true and no one has, or ever had, any choice about whether p.”

• Rule Alpha: from □p infer Np.
• Rule Beta: from N(p ⊃ q) and Np, infer Nq.

Let P0 be a remote-past state, L the laws, P any truth about a present act.

1. □((P0 ∧ L) ⊃ P) — from CD
2. N((P0 ∧ L) ⊃ P) — Alpha
3. NP0 — fixity of the past
4. NL — fixity of the laws
5. NP — Beta, twice

If NP for every act, no one has a choice about anything they do.

Where the fight is. Almost nobody denies Alpha. The war is Beta (transfer of powerlessness) and the reading of “has a choice.” David Lewis (“Are We Free to Break the Laws?”, 1981) distinguishes a strong ability (alter the actual laws) from a weak ability (do something such that, if one did it, the past or laws would have been different). Compatibilists typically take the weak reading. Incompatibilists say that is not the sense of “could” that responsibility uses.

McKay and Johnson (1996) and others produced countermodels to original Beta (agglomeration failures). Van Inwagen and others revised the rule. The argument is not dead; it is a family.

The political analogue is not a deduction, and it is not nothing. If a remote office plus standing rules fix the act, and neither the office nor the rules is up to the person, then in the institutional sense the person had no choice. That is the Consequence Argument’s shape on a smaller board: fixity of the commander, fixity of the rule, transfer of powerlessness. Autocracy is the clean case. Absentee title is a quieter case. Representative filters are a partial case. The analogue does not prove CD. It shows why a program that cares about “could have done otherwise” cannot ignore offices.

5. Frankfurt cases, sourcehood, luck

PAP: an agent is morally responsible for doing A only if she could have done otherwise.

Frankfurt 1969: a counterfactual intervener stands ready to force Jones to do A if Jones is about to refrain. Jones does A on his own. The intervener never acts. Jones could not have done otherwise (the intervener would have stopped him), yet — Frankfurt claims — Jones is still responsible.

If PAP falls, leeway incompatibilism loses its main bridge from “no alternatives” to “no responsibility.” Source incompatibilists rebuild the bridge: what fails under CD is not alternatives but origination.

Ginet and others reply that there is still a flicker of an alternative (a prior sign the intervener uses), so PAP is not really refuted. The literature after 1969 is largely about whether the flicker can be driven to zero without begging the question against indeterminism.

The institutional analogue is the kindly master and the standing army. Pettit’s domination: you are unfree if another could interfere arbitrarily, even if they do not. Frankfurt’s intervener is a philosopher’s Black. An autocrat who happens not to call tonight is still the source of the room you move in. A listed residual claimant who happens not to fire you this quarter still holds the legislature over the plant. CenLib’s quarrel with “as long as they don’t interfere, you are free” is this analogue. It is Berlin’s worry about positive liberty captured, and Pettit’s non-domination, not a new Frankfurt paper.

Luck: if the difference-maker is undetermined, why is it yours? Kane’s self-forming actions are the best-known attempt. They remain contested. Hard incompatibilists take the contest as a loss for FA. This paper does not need to pick the winner. It needs the gap named: authorship is the thing every camp is trying to keep or deny, and none has an uncontested account of it.

6. The definitional gap

Both camps smuggle a theory of the agent into the word they claim to be testing.

The compatibility debate therefore cannot close until “agency” is specified independently of the conclusion one wants. That is not a verbal quibble. Law needs a unit that can be culpable. Economics needs a unit that can hold title and consent. Political theory needs a unit that can authorize or refuse. Biology of organisms (Noble and Noble 2018, 2023; active inference) needs a unit that can select among constrained trajectories. Foundations of physics, on the Individual Completion line (Shields 2026a, after Einstein 1949), needs a description of the individual system, not only the ensemble. That physics is prospected into the working definition in §7. It is not a deduction of FA.

List and Pettit, Group Agency (2011): groups can be agents in a functional sense. The warning for this program is the converse. Once a group is treated as the primary agent, the person becomes an organ. Party, class, ticker, “the people.” CenLib’s unit-constraint is a refusal of that promotion, not a denial that groups act.

Organism-level agency in physiology (Noble 2012; Noble and Noble 2018) is adjacent, not a solution. Downward causation and the harnessing of stochasticity may constrain models of the individual system. They do not, by themselves, refute Beta or restore PAP. Do not smuggle a physiology paper into van Inwagen. Do not smuggle it into a property law either. See Competition & Mutual Aid. Mises’s acting man (Human Action, 1949) is the economist’s cousin of this unit — purposeful behavior as the primitive of catallactics. The file is Foundations of the Right. This program takes the cousinhood. It refuses praxeology as apodictic proof of a title regime, just as this paper refuses physiology as a deduction of FA.

7. A definition that tries to satisfy logic, history, and observation

Working definition. A temporally extended individual system is an agent to the extent that it can prospect possible transitions compatible with its constraints and then be the source of some of its own state-transitions at a time-instance, such that those transitions are attributable to it rather than only to an external commander, a biological imperative, or a statistical ensemble.

Two families bound the fiber of those possible transitions without exhausting the individual:

1. Institutional constraints — command, title, law, agenda. The office and the rule that are not up to you.
2. Biological and ecological resource limitations — body, niche, caloric and material budget. Hunger, thirst, elimination, sleep, injury, the finite stuff of a niche.

Before the act, the future of the choice is an unread remainder: only a coarse cell is given, many remainders still fit, and the future looks probabilistic at that grain. After the act, there is a time-instance, an agent acting, and an outcome from the choice. A thinking agent is not exhausted by biological needs plus a causal chain to that moment. The only outcome, given what was written down, would not have been the only outcome.

That is not yet a finished definition. Open problem 1 stays open. The unread-remainder / time-instance clause is prospected, not closed. The whole is a constraint with three tests, plus a fourth that only the prospected clause has to meet: it must not already pick a winner, it must sit in a dated line, it must match observation — and it must say what would count as its own failure.

Logic. It does not require CD to be true or false. It does not require PAP. It does not identify freedom with the absence of causes (Hobbes can sit here) and does not identify it with a swerve (Epicurus cannot sit here without authorship). It distinguishes the person from an ensemble, so class, ticker, party, and wavefunction-without-completion are all marked as changes of subject. It allows constraint: a body, a world, laws, other persons. Agency is not omnipotence. A definition that needs omnipotence is not a definition of persons. The prospected clause does not smuggle FA in through physics: if the complete individual state S were given, the first-event time T(S) would already be given — that looks like individual completeness, not a proof of leeway. If only a coarse cell is given, T is distributed — that looks like chance at the grain of our evidence, not primitive chance in the law. Compatibilists can sit at the complete grain; libertarians can sit at the unread remainder. The unique addition is the cut, not a winner.

History. It sits behind Aristotle’s “origin in the agent,” Locke’s “the agent is free,” Belsham’s self-determining power (without taking Belsham’s necessitarian side), Berlin’s refusal to let a party be your real self, Pettit’s non-domination, Sen’s capabilities as what persons can do. The prospected clause sits behind Einstein’s 1949 successor-theory demand — complete the individual; statistics as the theory of what was not read — not behind Epicurus’s swerve, not behind a quantum-pop-science headline, and not behind the US Libertarian Party’s identification of agency with a particular title regime. Date the word. Date the physics.

Observation. We do not observe a completed physics of FA. We do observe:

• persons who can refuse and persons who cannot (the slave’s yes is not consent);
• offices that can be ignored and offices that must be priced (autocracy versus a bound executive);
• title that is a legislature over other people’s time (absentee residual claim) and title that is occupancy;
• assemblies in which the person is a voice and systems in which the person is a sampled mass;
• biological imperatives that press and biological imperatives that are overridden (fast, hunger strike, continence, delayed elimination, pain held for a reason);
• that these differences survive the metaphysical stalemate. People keep designing as if the unit were real.

The dilemma of ill-defined agency is therefore not a reason to stop designing. It is a reason to design under a constraint that does not pretend the metaphysics is finished. Maximize the room. Do not claim the swerve. Do not claim Laplace has cancelled the person. Do not claim Individual Completion has closed van Inwagen.

The prospected addition: unread remainder and the time-instance

Individual Completion (Shields 2026a) converts Einstein’s 1949 successor-theory demand into a predictive primitive for the case Einstein used as his working example — the decay of a metastable individual and the mark it writes. Einstein’s Reply refuses the ψ-function as a complete description of an individual. If a complete description of individuals is achieved, statistical quantum theory would stand to that description as statistical mechanics stands to classical mechanics. The laws of the complete description, per se, have nothing to do with statistics. 2

The primitive is a first-event time T(S) fixed by a complete individual state S. Laboratory preparations specify only a coarse cell S_coarse; the unread remainder S_⊥ decides the instant.

S = S_coarse ⊕ S_⊥. One individual, written so one can say what was fixed and what was left unread. ⊕ means “together with,” not a calculation. If S is given, T(S) is given. There is no probability in that sentence. If only S_coarse is given, T is distributed over the fiber of compatible remainders. That distribution is the statistical theory of the unread remainder.

Axioms of that program, compressed (Shields 2026a, §3):

• A1. Individual completeness. There is a fact about when the mark is written. Statistical quantum theory does not contain that fact.
• A2. Laws without primitive chance. Probability is not a term in the law. It appears when the complete description is replaced by a cell in a coarse partition.
• A3. Events are field facts. The counter pulse and the ink mark are concentrated regions of the same field.
• A4. Time belongs to the description. The decay instant T* is an event time on the worldtube of the system.
• A5. Statistics survive as coarse-graining. When many systems share a coarse cell and differ in unread structure, first-exit times may look exponential at that grain. This is the kinetic-theory relation Einstein asked for.
• A6. Predictive content is conditional on S. A theory that only outputs a rate is still the incomplete theory. Completion means: narrow S, narrow T. If better individual data never shrink the uncertainty in T, the program has failed.

The Rank-1 analog (metastable well plus a finite Fourier background with definite phases) is a characterized schema, not a claim that nuclei — or persons — are built this way. Specified phases yield a unique T; unstructured phases yield a waiting-time histogram; partial readout of phases narrows Var(T). Nuclei are the last test, not the first. Characterized analog systems are. The demand is not a hidden-variable dressing of ψ, and it is not a proof that nature is deterministic. It is a consistency requirement across scales: do not declare an incomplete description final while treating the laboratory record as definite (Shields 2026a; Shields 2026b).

The mapping onto agency is analogical and prospected. It is not a theorem of Shields 2026a, which is a paper about metastable decay. It is the unique addition this research program is now putting on the table:

Before a choice is made, we have the unread remainder. The future is probabilistic at the grain of what was written down: many remainders still fit. After a choice, at the exact time-instance, we have agency: an individual acting, and the outcome is from the choice. A thinking agent is not determined only by its biological needs and a causal chain to that moment. Given what the coarse cell listed, the only outcome would not have been the only outcome.

That last sentence is the leeway claim at the coarse grain. It is not a claim that the complete individual is chancy. If S were given, T would be given. We do not have S. We have S*_coarse — body, niche, office — and a remainder we have not read. Probability here is the statistical shadow of that remainder, not primitive chance in the law (A2, A5). That is why the mapping does not already pick a winner in the compatibility debate, and why it is not Epicurus.

Dated against the record

The unique addition has to be shown as unique. That means dating it against the claims it could be mistaken for.

*Epicurus’s swerve; Cicero, De Fato. Atomic deflection as a physical interruption of fate. The luck problem is already in the ancient transmission: a random swerve is not authorship. Aristotle’s voluntary is origin in the agent who knows the particulars — a source condition, not a hiccup. Individual Completion refuses primitive chance*. The distribution is the unread remainder of this individual, not a physical hiccup that then needs to be made “theirs.” That is the first difference. It does not yet pay the authorship debt. It relocates chance from the law to the coarse grain.

Kane’s self-forming actions (1996). The load-bearing event-causal libertarian: reasons-weighted undetermined efforts of will, at which the agent sets the will that later determined actions express. The luck objection (Mele 2006; van Inwagen 2000) asks why the undetermined bit is theirs. If S is complete, T is given — that looks like individual determinism, not Kane’s undetermined effort. If only S_coarse is given, T is distributed — that looks like Kane’s indeterminism at the grain of our evidence, with the completeness claim that the remainder was already of this individual. The unique addition is: the “undetermined” bit is unread structure of this system, not an injected chance event. Whether that remainder is authored is still the luck gap. Prospected, not paid. Kane remains the best statement of the debt. This mapping does not retire him.

Noble and Noble, harnessing stochasticity (2018). Organisms use variation; agency as a physiological use of noise, not a ghost in the machine. Adjacent, as §6 already said. Downward causation and biological relativity (Noble 2012) constrain models of the individual system. They do not refute Beta or restore PAP. Individual Completion is closer to Einstein’s kinetic-theory relation than to a biology of free will: a characterized analog with an empirical constraint (A6), not a metaphor of “the cell decides.” The Nobles give a biological reason to refuse erasure of the organism into a gene-ensemble. They do not give a first-event time, and they do not write a constitution. Keep them adjacent. See Mutual Aid.

Einstein 1949, the successor-theory demand. Complete the individual, not the correlators (Shields 2026b). Statistical quantum theory demoted to the role statistical mechanics plays inside classical mechanics. The unique addition to the agency literature is to treat the agent as an individual system whose act-time is a fact about one worldtube, and to treat “the future was open” as the statistical shadow of an unread remainder. This is Einstein’s demand applied analogically to the person. It is not a hidden-variable dressing of ψ. It is not a claim that persons are radioactive atoms. It is not unification-as-quantization. Contact with already catalogued statistics is a graduation constraint, not an entrance examination (Shields 2026b). Nuclei are not the first test of the physics; persons are not the first test of the physics either. The agency mapping is a further analog, and it is marked as such.

*Locke, Essay II.xxi. The agent is free; the will is a power. Badly posed as “whether the will be free.” Individual Completion: the individual is the bearer of T; the coarse description is not. Later political economy that treats the firm or the party as the agent has already left both Locke’s unit and S.

Frankfurt 1969, and Pettit’s kindly master. Jones does A on his own; the intervener never acts. The act at T is Jones’s. The standing intervener is still the institutional S*_coarse — the kindly master, the absentee claimant, the autocrat who does not call tonight. Individual Completion does not dissolve domination. Non-domination (Pettit 1997) remains the institutional test. The two boards stay two boards.

The Consequence Argument, scale and ensemble. Informal statement (van Inwagen 1983, 56): past plus laws fix the act; neither is up to us; the act is not up to us. If S is the complete individual, T is fixed — the argument looks true at that grain. If S*_coarse is what we have, the act is not fixed by what we wrote down. The unique addition is a scale remark, not a refutation of Beta: the argument’s “laws” may be laws of the complete description (no probability, A2) while our evidence is always a coarse cell (A5). That is open problem 2 restated with a candidate primitive. It does not close the argument. It names the grain at which “has a choice” might mean something other than a joke, without having shown that it does.

Libet (1983) and the readiness potential. Popularized as “the brain decides first.” Mele 2009: the data do not show that a decision was completed before awareness, and they do not show that veto is impossible. Individual Completion’s T is an event time of the individual, not a scalp latency. Do not smuggle a readiness potential into T. Do not smuggle T* into a Libet headline.

Two constraint families

The fiber of possible transitions is not the set of all logically possible worlds. Two families bound it. They shape the ability of choices. They do not exhaust S.

Institutional. Command, title, law, agenda. The Consequence-Argument analogue of §4: fixity of the commander, fixity of the rule, transfer of powerlessness. Autocracy is the clean case. Absentee title is a quieter case. Representative filters are a partial case. A pole that can draft, sanction, unplug is Ostrom’s principle 7 refused (Empires; Equilibria). These are S_coarse of the polity: what the preparation called “the situation” can write down. They leave a remainder or they do not. When they do not — when another will is the source of the body’s transitions — the person is an instrument. That is observation. It does not wait for FA.

A story that arrives before the act is the courier spend of this family: Creel’s Four-Minute Men, an incubator testimony, a firehose that names dissent as occupancy. It does not delete the unit. It fills the unread remainder until some transitions are unsayable, and a Sedition Act (or a platform ban) makes saying them costly. That is less institutional agency, not a solved metaphysics. Propaganda is that instrument at section depth.

Biological and ecological. Body, niche, caloric and material budget. Hunger, thirst, elimination, sleep, injury, climate, soil, the finite stuff a life is spent on. These are S_coarse of the organism and its place. They press. A definition that ignores them is a definition of a ghost. A definition that lets them exhaust the individual is a definition of a thermostat.

The two families interact. Famine is ecological and becomes institutional when an office allocates the grain. Injury is biological and becomes institutional when the plant will not be voted. CenLib’s bid to maximize institutional room does not abolish ecology. It refuses to add a commander on top of the niche.

Prospecting, then acting

A thinking agent is not determined only by biological needs and a causal chain to that moment.

Prospecting is the use of the fiber: possible futures compatible with the coarse cell, held as possibilities, weighed. Acting is the first-event: T*, a time-instance, an outcome from the choice. Prospecting is not the act. A person can prospect and freeze. A person can act without having prospected well. The definition needs both because Locke’s unit is the agent, not the will as a detached organ, and because Individual Completion’s primitive is an event time of one individual, not a histogram of an ensemble.

The causal chain to that moment is the Consequence Argument’s P0 plus L, or the history written as coarse. Biological needs are the body’s list. If those together were already S, there would be no remainder to read, and “the only outcome would have been the only outcome” would be true at every grain. The mapping’s claim is that they are S_coarse, not S. The remainder is of that same individual. The act decides the instant.

This is why the definition can sit with Hobbes on constraint (agency is not the absence of causes) and refuse Epicurus on the swerve (a random interruption is not authorship), without having paid Kane’s debt. Prospecting is the nearest this program currently has to authorship: the agent as the one who holds the fiber and then acts. That is not yet agent-causation (Reid, Chisholm, O’Connor). It is not yet a self-forming action (Kane). It is a constraint on later work: do not write a theory of the act that never mentions the prospecting, and do not write a theory of prospecting that never mentions a time-instance.

Override of biological command is observation, not theology

Yes, we pee when we need to, and eat when we need to. Biological imperatives are real. They are S_coarse of the body.

They are not an exhaustive commander. Persons have been observed to override them. Fast. Hunger strike. Continence. Delayed elimination. Sleep postponed. Pain held for a reason. Traditions that train the override — ascetic, contemplative, political-sacrificial — are evidence that the imperative is a constraint, not a commander. The secular cousin is delay of gratification: a future possibility prospected against a present press. Neither is a proof of FA. A hunger strike can be read as a different cause (vow, party, shame, audience). The observation is only this: *the body’s list is not S.* If it were, override would not occur. It occurs. So there is a remainder, and there is a choice under constraint.

This paper does not need a theology to say that. It does not need a proof that the remainder is a soul. Spiritual language in the source of the observation names a psychology of override, not a second ontology. A3 in the physics program is the analogous refusal: measuring instruments are not a second ontology. Do not add a ghost to pay for a fast.

What would count as failure

A6, analogically: if better individual data — not better ensemble rates, not a better Libet headline, not a better party story about “the people” — never shrink the uncertainty in when and what this person does, the mapping has failed for persons.

What would not count as success: a quantum pop-science swerve; a hidden-variable dressing of ψ sold as free will; a proof of CD at the scale of persons; a proof of FA; a constitution deduced from Einstein 1949; CenLib as a theorem of Shields 2026a. Individual Completion does not write CenLib. CenLib maximizes institutional room for a unit this paper has not finished defining. The physics is a candidate completion of that unit. The two remain two.

The Rank-1 analog is not how persons are built. Nuclei are not the first test of the physics; persons are not a substitute test. Characterized analog systems are the first test of Shields 2026a. The agency mapping is a further analog, prospected in this paper, and marked as such. Open problem 3 is now this A6 analog, not a parked slogan.

8. Political structures constrain agency

More freedom, in the institutional sense, means more room for the unit. That is an observation with a literature, not a slogan.

Autocracy / command. One will is the source of public transitions. Others are instruments. Censorship, arbitrary arrest, a police that is not under standing law (Locke Second Treatise §22: liberty is not “do as one lists”; it is not being subject to another’s inconstant, uncertain, unknown, arbitrary will). The Consequence-Argument analogue is clean: the commander plus the rule are not up to you; the act is not up to you. Hard determinism about physics is unnecessary for this verdict.

Authoritarian party-state. The party claims to be the real subject (Berlin’s warning about positive liberty captured). Applied Marxism-Leninism did not retire the office; it staffed it. See Fractures. The person remains an administrative ensemble. “Could have done otherwise” is a party question.

Representative democracy. Periodic suffrage, careers, a boule-like filter. More room than autocracy: exit of the ministry is possible; voice exists on a schedule. Still a mass to be aggregated. Michels (1911): oligarchy inside the organization that claims to speak for you. Agenda oligarchy — staff, recommenders, petition industry — can return you to a commanded agenda with a franchise on paper. Hansen (1991) on Athens: sortition was real; notable families still dominated the volunteer pool. The Pnyx did not hold the body. “We have never implemented a true democratic government” is the program’s claim; Athens does not refute it.

Direct assemblies, allotted agenda, bound emergency executive. The CenLib polity axiom (Center): the person as unit of voice, not a sampled mass. This is Constant’s ancient liberty rebuilt with modern ledgers, plus Pettit’s non-domination as the test of the office. It is not a proof that the will is free. It is a structure in which political transitions are more nearly attributable to persons. Ochlocracy (Polybius) is the failure mode if the “few” (allotted agenda) and the “one” (narrow emergency) are unspecified. Open problem 15.

Scott, Seeing Like a State (1998): high-modernist command sees populations as ensembles to be made legible. That is an agency paper, whatever Scott thought he was writing. A state that must simplify persons in order to rule them has already changed the subject of Paper 01.

Game of Empires is why this section cannot be only domestic. A pole can draft, sanction, unplug. Ostrom’s principle 7 — external recognition of the right to organize — is what a pole can refuse. Institutional agency at home is a play on a polarity board.

9. Economic structures constrain agency

Different economic systems leave different room. That, again, is observation.

Slavery and serfdom. The person is title, or is bound to land. Labor-time is not theirs. No serious theory of consent starts here.

Command economy. Gosplan as the source of productive transitions. The worker is an input quota. Local knowledge (Hayek, The Constitution of Liberty, 1960, and the calculation debate) is an agency argument: situated judgment cannot be fully substituted by a center. Hayek’s conclusion for a title regime is not this program’s. The observation is.

Wage labor under absentee residual claim. The person can often exit a job (Hirschman’s exit, at a cost) and cannot vote the plant. Shareholder primacy is a private legislature over hours. Marx 1844 called this alienation of activity. CenLib calls it a species error: the scaled firm is treated as a person; the persons inside it are treated as an ensemble. Listed stocks and bonds are that legislature made liquid. The negative rule (no listed stocks/bonds) is an agency rule before it is a finance rule.

Sole proprietorship, partnership of operators. The person, or a small set of persons who work the plant, is the source of productive transitions. This is the baseline CenLib keeps. When demand requires hired labor, species change: employees become partial owners — occupant use-shares, not a secondary market in control. That is an attempt to keep the economic unit lined up with the agency unit. It is not finished (threshold rule, investment instrument: open problems 8 and 9). It is the motivation named.

Automation without hiring. If models and robots make hiring optional under shareholder primacy, the CPR may have no human users left. Species-change at hired-labor scale does not, by itself, answer that. Open problem 20. It is an agency problem: a structure that can prefer non-persons as inputs has already stated its unit.

Sen’s capabilities and Nussbaum’s version: an economy is to be scored by what persons can do and be, not only by output. That scoring is an agency metric. This program has not written the metric (open problem 24). It refuses the opposite metric: fitness-as-market-share, which Mutual Aid has already marked as an unrepaired metaphor.

10. Observation without a physics proof

The dilemma: agency is ill-defined in the sense Paper 01 began with, and we still have to build law, firms, and assemblies that take a unit as given.

What would count as empirical progress without a proof of FA (open problem 3):

• constraints on models of the individual system — the Individual Completion mapping of §7, prospected, with A6 as the failure condition: if better individual data never narrow when and what this person does, the mapping has failed for persons;
• better maps of which institutional differences actually enlarge attributable transitions (voice that is not captured; exit that is not formal; title that is not dominion);
• refusal of neuroscience cartoons that skip Mele.

What would not count: a quantum pop-science swerve; a Libet headline; a party that says it is your freedom; a ticker that says the price is your will; a hidden-variable dressing of ψ sold as free will; a proof of FA deduced from Einstein 1949.

We observe that persons under autocracy have less room than persons under a bound law. We observe that persons whose labor-time is legislated by absentee title have less room than occupant-operators. We observe that a franchise without agenda control is a thinner voice than an assembly that can set the question. Those observations survive if CD is true, and they survive if it is not. That is why CenLib does not have to wait for van Inwagen to retire.

11. Why CenLib: maximize institutional agency

Center libertarianism rejoins the 1789 agency sense with a property split and a voice structure. The motivation, stated without a manifesto:

Keep the person as the unit of title and of voice. Do not let the scaled firm or the representative office become a private legislature. Change species at hired-labor scale rather than abolish personal property. Exit by building the alternative rather than seizing the office that then will not retire.

Maximize agency here means: maximize institutional room for the unit Paper 01 has not finished defining. More democratic leaning, in this program’s dated sense — nested direct assemblies, allotted agenda, bound emergency — is more room for political sourcehood. Occupant-operator title is more room for economic sourcehood. Neither is omnipotence. Neither is a swerve. Both are refusals to treat persons as ensembles.

The right-libertarian error is to keep Locke’s unit through a firm that is no longer a person. The applied-Marxist error is to drop the unit into a class and then into an office. Berlin’s captured positive liberty is the second error’s English. Absentee title is the first error’s English. CenLib is the attempt not to make either — named center because it is the join of the two libertarianisms at property. The right-hand file is Foundations of the Right.

It will fail if Game of Empires is ignored: a pole can unplug a local commons. It will fail if Ostrom is ignored: a commons without monitors is Hardin’s parable. It will fail if the definition of the unit is quietly replaced by a party, a ticker, or a model-weight monopoly. Those failure modes are why this paper exists as Axis A, not as a preface.

12. Gaps the literature — and this program — still leave open

1. Definition that does not pick a winner. Section 7 is a constraint. The unread-remainder / time-instance clause is prospected. It is not closed. Open problem 1.
2. Scale gap. Consequence arguments are about whole-world CD. Human action sits at a mesoscale. Few papers specify the level at which “laws” must fix action for Beta to apply. Section 7 restates the gap with a candidate primitive (S versus S_coarse). Open problem 2.
3. Ensemble gap. Statistical theories describe collectives. Responsibility is predicated of individuals. Individual Completion is a candidate completion of the individual system. It is not yet that completion for persons. Moving from ensemble to individual without it remains a change of subject.
4. Luck gap. Libertarians who buy indeterminism to save alternatives owe authorship. Kane remains contested. The mapping relocates chance from the law to the unread remainder. It does not pay the debt.
5. Manipulation gap. No consensus cutoff between socialization and covert control. Institutions are that cutoff in practice.
6. Political gap. The analytic literature mostly still does not ask what offices do to agency. Sections 8–11 are this program’s entry. They are not a finished metric.
7. Semantic capture. “Libertarian” in 1789 ≠ 1857 ≠ 1971 ≠ Kane. Undated use is a method error. Open problem 4.
8. Metric of institutional agency. Voice, exit, title, non-domination — specify a bundle that can be compared across regimes without becoming GDP-of-freedom. Open problem 24.
9. Whether maximizing institutional agency requires FA to be settled. This paper says no. That “no” is a research stance, not a proof. Open problem 25.
10. Whether the agency mapping of Individual Completion survives A6. If better individual data never shrink Var(T) analogically — never narrow when and what this* person does — the mapping has failed for persons. Open problem 3.

13. What this paper does not claim

It does not claim that quantum indeterminacy yields free will. It does not claim that CD has been proven at the scale of persons. It does not treat Individual Completion (Shields 2026a; Shields 2026b) as a finished solution to van Inwagen, as a hidden-variable dressing of ψ, as a proof that nature is deterministic, or as a constitution. The unread-remainder / time-instance mapping is prospected, not closed. It does not claim that more votes are more agency regardless of agenda control. It does not claim that markets are freedom. It does not claim that CenLib has been shown to maximize anything by a theorem, or that CenLib follows from Einstein 1949. It does claim that further work on agency has to say what the unit is, before it says whether that unit is compatible with a given law-story — and that work on government and economy is already work on that unit, whether the seminar has finished or not.

14. Attachment to the other axes

Game of Empires is the environment in which institutional agency is or is not recognized (Ostrom 7). Isolated experiments die on that board. Dalio’s inner conflict, taken there, is this paper’s conversion of persons into ensembles (the draft, the class, the mass, the wartime “we”). Outer conflict is states treating persons as inputs to a security dilemma. A pole can draft, sanction, unplug — including at the energy unit (Follow the Oil: a caloric cut, a strait closed, a magnet plant halted). A pole can also fill the remainder before the act (Propaganda: Creel, blowback, a third scare). Scott, Seeing Like a State, is the inner file.

Equilibria is the formal apparatus for cooperation without pretending Nash is nature. Solution concepts are not persons. Do not smuggle PCE into van Inwagen.

Fractures dates the political words on the left. The 1789 sense of libertarian is this paper’s; the 1857 and 1971 senses are not substitutes. Foundations of the Right dates the American inversion and the Lockean file: self-ownership as an agency claim is real; title through the scaled firm is the spend.

Competition & Mutual Aid is why biology does not underwrite a property law, and why Noble organism-agency is adjacent rather than a deduction. Individual Completion is the adjacent physics: a candidate completion of the individual, prospected into the working definition, not a deduction of FA.

Center libertarianism is the proposed stack whose motivation this paper has now named: maximize the room for the unit. Open specs stay visible.