Compatibilism (Free Will)

The compatibilist move

Start with the apparent conflict. Determinism says: given the laws of nature and the state of the world a million years ago, only one future was physically possible — including every choice you've ever made. Free will seems to require: when you chose dessert, you genuinely could have done otherwise. The two seem to clash.

Compatibilists pull the conflict apart by examining what “free” means in ordinary usage. A prisoner isn't free; a person at gunpoint isn't free; a kleptomaniac isn't free. But in none of these cases is the unfreedom about causation per se — it's about being constrained by something external to or alien from the agent's own desires. Compatibilist freedom is the absence of such constraints.

The classical compatibilist analysis (Hobbes, Hume, Mill, Ayer):

An agent acts freely if she does what she wants, and could have done otherwise if she had wanted to.

Note the conditional. The kleptomaniac wants to steal, but had she wanted not to, she'd have stolen anyway — her hand is forced by compulsion. By contrast, you take dessert because you want to; had you wanted to skip it, you'd have skipped it. You meet the conditional; the kleptomaniac doesn't.

Argument structure

The compatibilist argument can be set out in three steps.

1. Free will doesn't require uncaused choice. The relevant ordinary contrast is “freely” vs “under compulsion”, not “freely” vs “caused.”
2. What freedom requires is the right kind of cause. Acting on your unconstrained desires, where those desires would have responded to reasons, counts as free.
3. Determinism is consistent with the right kind of cause. A determined choice can still issue from the agent's own values, be responsive to reasons, and be unconstrained.

Therefore: free will is compatible with determinism.

Compatibilism vs its rivals

	Compatibilism	Libertarianism	Hard determinism	Hard incompatibilism
Determinism true?	Maybe (irrelevant)	No	Yes	Yes or no
Free will exists?	Yes	Yes	No	No
Need alternative possibilities?	No (for many versions)	Yes — agent-causation or indeterministic event	—	—
Moral responsibility?	Yes	Yes	No	Forward-looking only
Champion	Hume, Frankfurt, Fischer	Kane, Chisholm	d'Holbach, hard-line Spinoza	Pereboom, Caruso
Lay intuition fit	Mixed	Strong (we feel uncaused)	Threatening	Threatening but less so
PhilPapers 2020 share	59%	19%	11%	(within 11%)

Worked example: the Frankfurt case

Harry Frankfurt's 1969 paper introduced an influential thought experiment. Imagine Black, a neurosurgeon, has secretly wired Jones's brain so that if Jones is about to deliberate toward not assassinating the senator, Black will intervene and force the deliberation the other way. Black is a fail-safe, not an active manipulator.

Now suppose Jones, on his own and without any intervention from Black, deliberates and decides to assassinate the senator. He pulls the trigger entirely on his own steam.

• Did Jones do it freely? Most people say yes — he chose, his reasons, his action.
• Could Jones have done otherwise? No — Black would have prevented it.

If both answers are right, then free, responsible action doesn't require the ability to do otherwise. The Frankfurt case is a wedge between freedom and the “could have done otherwise” condition. Compatibilists use it to deflate libertarian demands. (Libertarians have replied with “flicker of freedom” defences — that some alternative remains in the deliberation phase, even if the ultimate action is fixed. The literature is enormous.)

Variants of compatibilism

• Classical / conditional analysis (Hume, Moore, Ayer). Free = could have done otherwise if you had wanted to. Critics: this collapses under iteration — could you have wanted otherwise if you had wanted to want otherwise?
• Hierarchical (Frankfurt, 1971). Free = first-order desires aligned with second-order volitions. The willing addict is free; the unwilling addict isn't. Critics (Watson): why should second-order desires be authoritative? Why not third-order?
• Real-self (Watson, 1975). Free = action flows from the agent's evaluative system, not raw appetite. Avoids regress at the cost of needing a theory of the “real self.”
• Reasons-responsiveness (Fischer & Ravizza, 1998). Free = the action issues from a mechanism that would respond to a sufficient range of reasons. Even if Jones in fact pulls the trigger, the deliberation he used would, in slightly different circumstances, have stopped him. Becomes semicompatibilism: maybe libertarian free will is incompatible with determinism, but moral responsibility is compatible.
• Strawsonian / reactive (P. F. Strawson, 1962). Free will isn't a metaphysical property at all — it's the practice of holding ourselves and each other accountable through reactive attitudes (resentment, gratitude, indignation). Determinism doesn't undermine the practice; it's part of the human form of life.
• Mesh / structural (Bratman, Velleman). Free action arises from a stable structure of plans, policies and identifications. The agent is constituted by the mesh; free action is action through it.

Objections and replies

• Consequence argument (van Inwagen, 1983). If determinism is true, then any choice is a consequence of the past and the laws. We can't alter the past or the laws. So we can't alter our choices — we're not free.
  Reply. The argument equivocates on “can.” In one sense we can't alter the laws (no agent has nomological power over physics); in another we can do otherwise (i.e., the world has counterfactual structure such that, had things been minimally different, our choice would have differed). Compatibilists insist only the second “can” matters for free will.
• Manipulation argument (Pereboom, 2014). If a neuroscientist could engineer a person to act exactly as they do, the person isn't responsible — but a determined agent is in the same boat. So determined agents aren't responsible either.
  Reply. Most compatibilists accept the disanalogy: covert manipulators bypass the agent's reasons-responsive deliberation; ordinary causal history doesn't.
• Luck objection. If your character was determined by genes and upbringing, your good or bad actions are matters of luck. Holding you responsible is unfair.
  Reply. Compatibilists distinguish the question of original responsibility (which they sometimes concede) from forward-looking responsibility (which justifies blame, training, and reform).
• Wretched subterfuge (van Inwagen). Compatibilism redefines free will until the redefined version is trivially compatible with determinism, then claims victory.
  Reply. The redefinition tracks ordinary usage. If “free” never meant “uncaused” in everyday speech, no subterfuge has occurred.

Why compatibilism matters

• Criminal justice. If free will is metaphysically deep and rare, the case for retribution weakens. Compatibilists save the conditions of moral and legal responsibility under any plausible physics.
• Neuroscience. Libet-style experiments showing brain activity precedes conscious decision worry libertarians; compatibilists shrug — determined neural activity is exactly what the view predicts.
• AI ethics. If compatibilism is right, sufficiently sophisticated reasons-responsive systems could in principle have free will. Whether they do is then an empirical question about their architecture.
• Theology. Compatibilist free will reconciles divine foreknowledge or sovereignty with human responsibility — a long-running concern in Calvinist and Thomist traditions.

Common confusions

• Compatibilism doesn't claim determinism is true. Only that if it were, free will would survive.
• Determinism ≠ fatalism. Fatalism says outcomes are fixed regardless of intervening causes (“you'll die in Samarra no matter what”). Determinism says outcomes are fixed by intervening causes — including your deliberation and choice.
• “Could have done otherwise” is ambiguous. The categorical reading (in the very same world, with the same laws) is what libertarians demand; the conditional reading (in a slightly different world, with different antecedents) is what compatibilists offer.
• Free will ≠ conscious will. Libet experiments show readiness potentials before awareness, which is a fact about timing, not about whether the action was the agent's. Compatibilist free will isn't undercut.
• Compatibilism doesn't trivialise responsibility. An agent acting under coercion, compulsion, or manipulation still fails the compatibilist test, just as in lay usage.