Free Will and Determinism

The last article asked what a cause is. This one asks what follows if everything that happens has causes of a certain thorough kind, and the thing that might not survive is us. You decide to leave the office early, and it feels as though you could have stayed. If the whole history of the universe, together with its laws, fixed that you would leave, then it is not obvious in what sense staying was ever open to you. The worry is old, but it took its sharpest modern form in the 1970s, and it is a worry about possibility: about what was really possible for you, as opposed to what is merely consistent with the description of the case.

Determinism and its rivals

The classic statement is Laplace's, in 1814: an intelligence that knew all the forces of nature and the position of every body at one moment could, with enough calculating power, foresee the entire future. Notice what the thesis is not. It is not the claim that everything is predictable by us, which is a claim about our limitations. It is not fatalism, the claim that things will happen whatever we do. A determinist can say that your choices are causes, that they make a difference, and that the future depends on them. Determinism says only that those choices are themselves the products of earlier states of the world and the laws.

Whether the actual world is deterministic is an empirical matter that physics has not settled. Classical mechanics is deterministic in its usual formulation, though even there certain idealised setups produce odd results. Quantum mechanics divides by interpretation: collapse theories such as the Ghirardi, Rimini and Weber model are indeterministic, whilst Bohmian mechanics reproduces the quantum predictions with a fully deterministic dynamics. So philosophers of free will do not usually stake their position on the physics. They ask a conditional question: if determinism were true, could anyone act freely?

The question matters beyond the metaphysics because of what we ordinarily do with the answer. We blame, praise, punish and resent, and these practices seem to assume that people could have done otherwise.

Free will is usually understood as whatever capacity is required for moral responsibility in this sense, though that is itself a substantive claim, as the last sections will show. Everything below turns on whether that capacity is compatible with determinism.

Two kinds of answer

There are two questions here, and the standard map of positions comes from combining the answers.

Its opposite, incompatibilism, holds that free will requires the falsity of determinism. Incompatibilists then split. Some accept that we are free and deny determinism. Others conclude that, since either determinism or indeterminism is true and neither leaves room for the right kind of control, we are not free at all.

This is the metaphysical position, not the political one, and the two should not be confused. The libertarian about free will need hold no views about taxation. What the libertarian does hold is that in a free choice the agent really could have chosen differently, with the entire past and the laws held exactly fixed.

Compatibilists have traditionally replied that this demand is confused. Hobbes and Hume thought that freedom is the absence of external obstruction: you act freely when you do what you want and nothing prevents you. A prisoner is unfree and a person strolling in a park is free, and the difference has nothing to do with whether the strolling was caused. G. E. Moore gave this a modern form in 1912 by analysing "could have done otherwise" as "would have done otherwise, if I had chosen to". On this conditional analysis the sentence is perfectly consistent with determinism, since a determined agent can truly say that had they chosen otherwise, they would have acted otherwise.

The consequence argument

The most influential case against this comfortable picture is Peter van Inwagen's, first published in 1975 and given its full form in An Essay on Free Will in 1983. It is less an argument about free will than an argument about what "can" means.

Abbreviate "p, and nobody has, or ever had, any choice about whether p" as "Np". Let $P_0$ be a complete description of the world at some time long before anyone existed, let $L$ be the laws of nature, and let $P$ be any true proposition about what you do now.

The fourth premise is the engine, and van Inwagen defended it using a rule: from $Np$ and $N(p \to q)$, infer $Nq$. If you cannot affect the past and cannot affect the fact that the past necessitates your act given the laws, you cannot affect your act. The thought is the same as the one that makes you doubt anyone can change the outcome of a game already played, given that the outcome follows from how it was played.

What the argument has going for it is that it does not require libertarian metaphysics to get started. It does not say that determinism feels oppressive. It says that the premises are about the same relation we use in everyday life when we ask whether something is up to us, and it applies that relation consistently.

Compatibilist replies

There are three places to attack the argument, and each has a defender.

The first attacks the rule. Thomas McKay and David Johnson argued in 1996 that rule beta is invalid, because it entails that $N$ distributes over conjunction. Suppose a fair coin is in fact not tossed, though you could have tossed it. Let $p$ be "the coin does not land heads" and $q$ be "the coin does not land tails". Nobody has any choice about $p$, since no one can make a tossed coin land heads, and the same goes for $q$. But the conjunction says the coin lands neither way, and you do have a choice about that, since you could toss it. So $Np$ and $Nq$ hold whilst $N(p \wedge q)$ fails. Van Inwagen responded by reformulating the rule, but the exchange shows the argument depends on an inference less secure than it first looks.

The second reply concerns the second and third premises. David Lewis's 1981 paper "Are We Free to Break the Laws?" distinguishes a weak thesis from a strong one. The strong thesis is that I am able to do something that would be a breaking of the laws. The weak thesis is that I am able to do something such that, if I did it, a law would have been broken, because had I done it, the past would have been slightly different, with a small miracle somewhere before my act. The compatibilist, says Lewis, need only accept the weak thesis. In a deterministic world, suppose I do not raise my hand but could have. Had I raised it, the past would have differed slightly beforehand, through a small divergence miracle relative to the actual laws. My raising my hand would not have caused that miracle, and so I would not have broken any law, but a law of the actual world would have failed in that counterfactual world. So I am able to do something such that, were I to do it, the actual laws would not have held. That is why the premise "nobody has a choice about the laws" is false on the reading the argument needs.

A reader may feel that this concedes the point whilst changing the words. It is hard to reply without begging the question, because the premise and Lewis's denial of it are both about the same notion of what is within one's power. A harder-line compatibilist, such as Helen Beebee and Alfred Mele in 2002, argues that on a Humean view of laws as summaries of the pattern of events, the laws really are partly up to us, and so the premise is simply false.

Alternate possibilities and responsibility

So far, the debate is about what an agent can do. A rival tradition says that this was the wrong question, because what really matters is responsibility, and responsibility does not depend on alternatives.

The principle is what links the consequence argument to responsibility. If responsibility requires being able to do otherwise, and determinism removes that ability, then determinism removes responsibility. Many incompatibilists simply took the principle for granted. In 1969, Harry Frankfurt published a challenge to it that has shaped the field since.

Frankfurt's example, in its usual retelling, runs as follows. Black wants Jones to shoot Smith. He has secretly installed a device in Jones's brain which, should Jones show any sign of deciding not to shoot, will force him to decide to shoot. In fact Jones, for his own reasons, decides to shoot without any interference, and the device never activates. Jones could not have done otherwise: had he wavered, Black would have intervened. Yet it seems that Jones is fully responsible for the shooting, since the device played no role in what happened.

If this is right, the principle of alternate possibilities is false, and responsibility does not require the ability to do otherwise. The consequence argument might still show that determinism takes away that ability, but this would not matter, because the ability was never required. Compatibilists such as John Martin Fischer and Mark Ravizza have built accounts of responsibility on this foundation. What matters is that the act issues from the agent's own reasons-responsive mechanism, and not that other mechanisms were available.

There is a deeper alternative, due to P. F. Strawson's 1962 essay "Freedom and Resentment". Strawson suggested that the question "is it rational to hold people responsible if determinism is true?" is the wrong way round. Our reactive attitudes, such as resentment and gratitude, are part of what it is to be involved in relationships with other people, and a theoretical thesis could not rationally require us to give them up. Responsibility is then not something we discover by looking at metaphysics, but a practice that sets the terms in which metaphysical claims are assessed.

The libertarian's cost

If the consequence argument stands and the compatibilist replies fail, then free will requires indeterminism. But indeterminism alone does not help. An event that merely happens by chance is not an exercise of control. This is the luck objection, and it is the main source of difficulty for libertarianism.

Imagine a decision that, given the past and the laws, has a fifty per cent chance of going one way and fifty per cent the other. Suppose it goes the first way. Nothing about the agent explains why it went that way rather than the other, since both were equally possible given everything about them. Van Inwagen's version of this uses the idea of rewinding the world to the moment before the decision. If it were replayed many times, the choice would go each way in proportion to the chances. That looks like chance, not control.

Libertarians have three main replies. Robert Kane's event-causal view holds that in genuinely difficult choices, where the agent is torn between moral and prudential reasons, there is an indeterministic effort of will, and whichever way it goes the agent has tried to bring about that outcome, so the outcome is something they were trying to do. Critics respond that trying does not remove the luck about whether the effort succeeds. A second reply, associated with Thomas Reid and in the twentieth century with Roderick Chisholm, is agent causation: the agent, as a substance, is a cause of the action without being caused to cause it by some prior event. This sidesteps the luck worry by making the agent the explanation. The price is a form of causation not reducible to events, which many find mysterious, and the question of what, if not luck, determines which way the agent causes.

The third reply is to deny the problem. Some libertarians hold that the reasons for an action can explain it without determining it, and that there is no further demand for a contrastive explanation of why one option was chosen rather than the other.

Given these costs, a fourth position gains ground. Hard incompatibilists such as Derk Pereboom accept the consequence argument, accept the luck objection, and conclude that we lack the free will required for the sort of moral responsibility that justifies blame as desert. We may keep many of our practices, but on a forward-looking basis such as deterrence and protection.

What is at stake

Everything turns on whether the compatibilist has answered the incompatibilist, or merely changed the topic. The incompatibilist charge is that the compatibilist's freedom, however sophisticated, is not the freedom people actually care about when they ask if they are the ultimate source of their actions. The compatibilist reply is that the freedom the incompatibilist wants is either unintelligible, as the luck objection suggests, or unnecessary, as the Frankfurt cases suggest.

The question is also not only abstract. A jury, a prison system, and anyone deciding whether to forgive a friend are all implicitly committed to a view about what a person is answerable for. Which view is correct determines whether blame is something deserved or only something useful.

Whichever way these issues fall, the argument has shown how heavily the debate leans on one notion: what could have happened, given what was fixed. The dispute is about which things are held fixed and what "could" quantifies over, and that is the question of how to give a systematic account of possibility and necessity themselves, which the next article takes up through the idea of possible worlds.