Libertarianism

Libertarianism (in the metaphysics of free will — not the political doctrine) is the incompatibilist position that affirms free will and therefore denies determinism. It holds that (i) free will is real, (ii) free will is incompatible with determinism, and so (iii) the world must be, at least at the moments of free choice, indeterministic. Where the compatibilist redefines freedom to fit a determined world, and the hard determinist accepts determinism and surrenders freedom, the libertarian keeps the strong, "could-have-done-otherwise-with-the-very-same-past" notion of freedom and pays for it with a commitment about how the world is.

The libertarian's distinctive burden is therefore twofold: a metaphysical task — to say what kind of indeterministic causation amounts to free agency rather than mere randomness — and an empirical bet — that the actual world supplies indeterminism in the right place. This page sets out the motivation, the main models (agent-causal, event-causal, non-causal), the history, the arguments for the view, and the objections it must answer — above all the luck objection.

The core commitments

Libertarianism is the conjunction of two theses, one negative and one positive:

• Incompatibilism (negative): if determinism were true, no one would have free will. The libertarian typically motivates this with the Consequence Argument and with the conviction that genuine freedom requires categorical alternative possibilities — that holding the entire actual past and the laws fixed, the agent could still have done otherwise.
• Libertarian freedom exists (positive): at least some human choices are free in exactly this sense; hence determinism is false where it matters.

Two further notions recur and explain why libertarians want indeterminism rather than treating it as an awkward cost:

• Leeway (alternative possibilities): more than one future must be genuinely open given the actual past — a real branching, $⟨\text{past}⟩\to A$ or $⟨\text{same past}⟩\to B$. Determinism permits exactly one continuation, so it forecloses leeway.
• Sourcehood / ultimacy: the agent must be the ultimate originator of the choice — the buck must stop with the agent, not with prior causes stretching back before her birth. Robert Kane calls this Ultimate Responsibility (UR): to be ultimately responsible for an action, the agent must be responsible for anything that is a sufficient reason or cause of it.

Libertarians disagree about which of leeway and sourcehood is fundamental, but most modern versions treat sourcehood/UR as primary and indeterminism as what sourcehood requires: if every sufficient cause of my choice traces to factors I did not originate, the regress of responsibility never terminates in me. Indeterminism is the only way to make the regress stop.

Why indeterminism is necessary (the regress of sourcehood)

The driving intuition: if my choice is fully determined by my prior character and motives, then to be truly responsible for the choice I would have to be responsible for that character — but my character was formed by earlier choices, upbringing, heredity, and circumstance, regressing back to factors entirely outside my control. A determined world hands the ultimate explanation of every act to the remote past plus the laws, never to the agent. To break this regress, at least some choices must not be settled by the prior state of the world: there must be points where the agent's will is genuinely undetermined and yet the resulting choice is the agent's own doing. That second clause — undetermined and genuinely the agent's — is the whole problem, and the three models below are competing solutions to it.

The three models of libertarian agency

Event-causal libertarianism

On the event-causal view (Robert Kane, Laura Ekstrom, Mark Balaguer), the only causation in nature is ordinary causation between events/states — there is no special agent-substance causation. Free choices are caused, in the familiar way, by the agent's reasons, beliefs, and desires — but this causation is indeterministic: the very same prior mental events leave more than one choice causally possible, each with some objective probability.

Kane's account is the most developed. He locates indeterminism in rare but pivotal self-forming actions (SFAs) — the will-setting choices (typically cases of torn decision, where duty conflicts with self-interest, or prudence with passion) through which we form the characters that, thereafter, more or less determine our ordinary acts. In an SFA, the agent is genuinely torn: there are two concurrent, goal-directed deliberative efforts of will, one toward each option, and the indeterminism is "located in" the agent's effort itself (Kane analogizes to amplified quantum indeterminism or to the chaotic, noisy processing of neural networks). Whichever way the choice resolves, the agent wanted it, made an effort for it, and endorsed it as her reason — so the outcome is not a random intrusion but the agent's own resolution of the conflict. Kane's signature move against the luck objection is plural rational control: because the agent rationally wills, and will own and ratify, the choice on either branch, the indeterminism does not undermine responsibility the way a coin-flip would. Most ordinary free acts then inherit their freedom derivatively, from the SFAs that built the character they flow from.

Strengths: naturalistically modest (no exotic causation); ties freedom to reasons and to a plausible developmental story. Pressure point: because the prior state does not settle the outcome, critics (Mele, Haji, Pereboom) argue the difference between the two branches is, at the crucial instant, not settled by anything about the agent — which is precisely the luck problem (below).

Agent-causal libertarianism

The agent-causal view (Thomas Reid historically; Roderick Chisholm, Richard Taylor, Timothy O'Connor, E. J. Lowe, Randolph Clarke in modern dress) holds that the event-causal picture is missing an ingredient. A free action is caused not merely by prior events but by the agent herself, conceived as an enduring substance that exercises a fundamental, irreducible causal power. The agent is an uncaused causer — a prime mover unmoved, in Chisholm's deliberately strong phrase: she initiates a new causal chain without herself being causally determined to do so.

This is designed to solve the luck problem directly: the choice is not left to chance among undetermined events, because the agent settles which way it goes — substance causation supplies the missing factor that determines the outcome without prior events determining it. The agent, not an antecedent state, is the terminus of the explanation, which is exactly what sourcehood/UR demanded.

Objections. (i) Intelligibility: what is "substance causation," and how does it differ from event causation involving the agent's states? Critics charge it is a label for the mystery, not a solution. (ii) The timing/exercise problem: if the agent-cause is not itself triggered by any prior event, what explains why she exercises the power now, toward $A$ rather than $B$ — and if nothing explains it, has the luck problem really been escaped, or merely relabeled "agent-causal"? (iii) Naturalistic cost: an irreducible causal power of substances sits uneasily with a physics of event-to-event law; O'Connor and Clarke have worked to make it compatible with (or a supplement to) the scientific image, with contested success.

Non-causal (volitionist) libertarianism

The non-causal view (Carl Ginet, Hugh McCann, Stewart Goetz) is the most radical: free actions — or their core, the volitions/willings — are not caused at all, neither by prior events nor by an agent-substance. A basic action has an intrinsic "actish" phenomenal quality; its connection to the agent's reasons is not causal but a matter of the agent's acting in order to fulfil a reason (a non-causal, teleological/intentional relation). Freedom is secured simply because the act is uncaused and hence not necessitated by anything.

Objections. (i) The control problem in its starkest form: if the volition is uncaused, it is hard to see how it is up to the agent at all rather than a brute, unowned happening — control seems to require some causal role for the agent. (ii) The deviant-causation / reasons-explanation worry: without causation, what makes it the case that the agent acted for this reason rather than that one, when both were present? Non-causalists owe an account of reason-explanation that does no causal work, which many find elusive.

A map of the positions

A short history

• Aristotle — the locus classicus of origination: virtuous and vicious actions are "up to us" (eph' hēmin), and we are "moving principles" of our actions; responsibility for character traces to earlier voluntary acts (an ancestor of Kane's self-forming actions).
• Epicurus — often read as positing the atomic swerve (clinamen): an uncaused, indeterministic deviation in atomic fall, introduced (on the standard interpretation, via Lucretius) precisely to make room for free volition against Democritean necessity — arguably the first appeal to physical indeterminism for free will.
• Medieval voluntarism — Scotus and others emphasize the will's self-determination and its power for opposites at the moment of choice.
• Thomas Reid (18th c.) — the founder of modern agent causation: the agent has an "active power" to cause volitions and is, in the strict sense, the cause of her free acts.
• Kant — a complex, non-naturalist libertarianism: transcendental freedom, in which the noumenal self is a spontaneous, atemporal cause outside the deterministic phenomenal order. (Often treated as a separate, two-standpoints theory rather than a model of in-time indeterminism.)
• C. A. Campbell (mid-20th c.) — a classic statement: free will is exercised paradigmatically in the effort of will to resist the strongest desire, against the line of least resistance, in a way introspectively given and not determined by character.
• Roderick Chisholm ("Human Freedom and the Self," 1964) — revives agent causation in analytic terms: the agent as prime mover unmoved.
• Robert Kane (The Significance of Free Will, 1996) — the leading event-causal theory: SFAs, Ultimate Responsibility, plural voluntary control, indeterminism located in the effort of will.
• Timothy O'Connor, Randolph Clarke, E. J. Lowe — sophisticated contemporary agent-causal accounts seeking compatibility with a broadly scientific worldview.
• Carl Ginet, Hugh McCann — contemporary non-causal views.
• Mark Balaguer (Free Will as an Open Scientific Problem, 2010) — argues the existence of libertarian free will turns on a genuinely open empirical question about whether certain neural events are indeterministic, refusing to settle it from the armchair.

Arguments for libertarianism

1. The argument from incompatibilism + the reality of freedom. The skeleton of the view: (i) the Consequence Argument shows free will is incompatible with determinism; (ii) we do have free will — a datum at least as secure as the metaphysical premises used against it; therefore (iii) determinism is false and our freedom is libertarian. The compatibilist saves freedom by weakening it; the libertarian takes the strength of our self-conception at face value.
2. The argument from sourcehood / Ultimate Responsibility. Genuine responsibility — desert-grounding responsibility, the kind that makes praise and blame fair — requires that the buck stop with the agent. As argued above, a determined chain never lets it stop with the agent (the regress runs to the remote past). Only an undetermined, agent-originated choice secures UR. This is the libertarian's positive engine, independent of any particular physics.
3. The phenomenological argument. In deliberation we experience the future as open and the choice as up to us now — a "garden of forking paths." We cannot deliberate at all except under the idea that more than one option is really available. Libertarianism takes this pervasive datum to be veridical rather than illusory. (Compatibilists reply that the experience is neutral between open alternatives and mere epistemic ignorance of what one will determinately do.)
4. The moral-desert argument. Basic desert, and the appropriateness of the strongest reactive attitudes, seem to presuppose that the agent could have done otherwise in the categorical sense — a presupposition only libertarian freedom satisfies.

Objections (and libertarian replies)

The luck / randomness objection — the central problem

This is the objection every libertarian must answer. If a choice is not determined by the agent's prior state — her character, reasons, deliberation — then, holding all of that fixed, the agent might have chosen either way. But then what makes it the case that she chose $A$ rather than $B$? Nothing about the agent settles it — it is, at the crucial instant, a matter of luck. Pressed as a thought experiment (the "rollback" argument, van Inwagen, Mele): imagine God rewinds the universe to the exact instant before the choice and replays it many times. With everything about the agent identical, sometimes she chooses $A$, sometimes $B$, in some proportion. The differing outcomes cannot be credited to or explained by anything about her — so the particular outcome on this run seems no more her doing, and no more a basis for responsibility, than a quantum coin-flip. Indeterminism appears to subtract control rather than add it. (This is the luck objection foreshadowed in the introduction, now aimed squarely at the libertarian, and it is the mirror image of the compatibilist's Humean argument.)

Libertarian replies:

• Event-causal (Kane): plural voluntary/rational control. On both branches of the rollback the agent acts for her own reasons, by her own effort of will, and endorses the result as what she was trying (in part) to do — like two simultaneous efforts where either success is a success the agent willed. So neither outcome is a mere chance intrusion; both are rational, voluntary, and owned. Indeterminism makes the choice undetermined without making it inadvertent. Critics (Mele's "still doesn't explain the difference") respond that this explains why each individual outcome is rational, but not why this one occurred rather than the other — the contrastive explanation is still missing, and that residue is luck.
• Agent-causal: deny the inference's key premise. It is false that "nothing settles it" — the agent settles it. The rollback shows only that no prior event settles it; agent-causation supplies a settling factor that is not a prior event, so the outcome is the agent's doing by definition. Critics retort that this just names a settler without explaining the contrast (why exercise the power toward $A$ this time?), so luck reappears at the level of the agent-causal exercise.
• Non-causal: reframe — demanding a factor that settles the choice presupposes a causal model of control the view rejects; the act's being up to the agent is basic, not constituted by a settling cause, so "what settles it?" is the wrong question. Critics find this closes the gap by fiat.
• General "no worse off" reply (Kane, Clarke): the same contrastive-explanation demand, pressed on the compatibilist, also goes unmet — a determined choice is "explained" only by factors the agent didn't ultimately originate, which is its version of luck (constitutive luck). If neither side can give ultimate contrastive explanation, indeterminism is not a special defect, and the libertarian at least secures sourcehood.

The intelligibility objection

Stated since antiquity (and by Hume): a free act must be neither determined (else not free) nor undetermined (else random) — but these seem exhaustive, so libertarian freedom is incoherent, a will-o'-the-wisp between necessity and chance. Reply: the dichotomy is not exhaustive if there is a third category — agent-causation (a kind of causation that is neither event-determination nor chance) or non-causal active origination. The entire libertarian project is the attempt to articulate that tertium quid; whether it succeeds is exactly what the three models contest.

The empirical / scientific objection

Libertarianism makes a falsifiable-ish bet: that the brain exploits genuine indeterminism at the moment of choice. But (i) quantum indeterminism may wash out at the macroscopic, warm, wet scale of neurons (decoherence), leaving brain dynamics effectively classical and deterministic; and (ii) even if there is neural noise, it looks like thermal randomness, which is the luck problem, not freedom. Replies: Kane appeals to chaotic amplification of quantum fluctuations by sensitive neural networks; Balaguer insists the relevant question — are the torn-decision neural events causally undetermined in the right way? — is simply open, to be settled by future neuroscience, not refuted a priori; agent-causalists may locate the causal power at the level of the person rather than requiring micro-indeterminism to "do the work" unaided. The honest situation: the empirical premise is unproven in either direction (see quantum mechanics and the interpretation-dependence of physical indeterminism).

The "indeterminism in the wrong place" / disappearing-agent objection

Even granting indeterminism, libertarians must put it where it helps. Indeterminism upstream of the choice (which considerations occur to me) makes deliberation chancy; indeterminism at the moment of choice makes the choice chancy; indeterminism after the will is set does nothing. Pereboom's disappearing agent objection presses that in the purely event-causal picture, once we subtract determination there is no thing left to tip the balance — the agent "disappears" as a settler, and only agent-causation could restore her, at the intelligibility cost above. Reply: Kane denies the agent disappears — she is present as the plural, effortful, reason-giving source on both branches; O'Connor agrees the event-causal version is inadequate and embraces agent-causation precisely to keep the agent in the picture.

Galen Strawson's causa sui regress (the basic argument)

A challenge that cuts against the libertarian's own goal: to be ultimately responsible (UR) for a choice, one must be responsible for the self — character, reasons, priorities — from which it issues; to be responsible for that self, one must have chosen it, on the basis of a prior self, ad infinitum. Ultimate self-origination would require being causa sui, which is impossible. So the very UR the libertarian invokes is unattainable — indeterminism does not help, because the regress is about origination of the self, not about determinism. Reply: Kane argues UR does not require being causa sui in Strawson's strong sense: it requires only that some of one's character-forming SFAs were themselves undetermined and willed — responsibility is built up incrementally through a finite history of self-forming choices, each undetermined, none requiring a prior fully-formed self that chose it. Whether this truly halts the regress, or merely pushes it into each SFA (which faces the luck problem afresh), is the crux of the dispute.

Common misconceptions

• "Libertarian free will means actions are uncaused / random." Only the non-causal model says uncaused, and even it denies the act is random (it is reason-directed). Event- and agent-causal libertarians insist free acts are caused — by reasons (indeterministically) or by the agent (substance-causally) — just not determined. Undetermined ≠ uncaused ≠ random is the libertarian's most important distinction.
• "It requires that everything be undetermined." No — most libertarians need indeterminism only at select nodes (Kane's rare self-forming actions); ordinary acts can be virtually determined by the character those nodes formed, and inherit freedom derivatively.
• "This is the same as political libertarianism." Entirely unrelated; the shared word is a historical accident.
• "Quantum mechanics proves we have free will." It does not. Physical indeterminism is interpretation-dependent (Bohm and Everett are deterministic — see the introduction), may not survive to neural scales, and even if real is thermal luck unless an account like Kane's or O'Connor's converts it into control. QM at most makes indeterminism available; it does nothing to make it freedom.
• "Libertarianism just denies science." The careful versions (Balaguer, O'Connor, Kane) are explicitly naturalistic and treat the key claim as an open empirical question, not a defiance of physics.
• "Agent causation is obviously magic." It is contested, not obviously incoherent; its defenders argue ordinary event-causation is no less primitive a posit, and that substance-causation is needed to keep the agent (not just her states) in the causal story.
• "If my choice could have gone either way, I'm not in control." This assumes the control problem is unanswerable — which is exactly what plural voluntary control (Kane) and agent-causation are designed to dispute; conflating "undetermined" with "out of control" begs the question against the libertarian.

Where libertarianism sits

Against the master dilemma, libertarianism keeps (1) free will exists and (3) free will is incompatible with determinism, and therefore rejects (2): the world is not (wholly) deterministic. Its great appeal is that it preserves the full-strength, ultimacy-grounding freedom our moral self-conception and our deliberative experience seem to demand — the freedom the compatibilist is accused of watering down. Its great burdens are the luck objection (does indeterminism add control or only chance?), the intelligibility objection (is there a coherent tertium quid between necessity and randomness?), and the empirical bet (does the actual brain supply the right indeterminism?). The three models trade these burdens off differently: event-causal libertarianism is naturalistically cheap but most exposed to luck; agent-causal libertarianism answers luck but pays in intelligibility; non-causal libertarianism is metaphysically spare but strains hardest on control.

The next pages take up the hard incompatibilist / free-will-skeptical response — which agrees with the libertarian that compatibilism fails and with the determinist's critic that libertarianism cannot beat luck, concluding we lack free will either way — and then turn to the bearing of physics and quantum mechanics on whether the indeterminism libertarians need is real and usable.