2. Two-way compatibilism and Task 2

Given that most contemporary compatibilists are two-way compatibilists, it may seem initially surprising that, in discussions of compatibilism in the contemporary free will literature, the focus is still almost exclusively on Task 1. Very little attention is paid to how compatibilists should handle Task 2.

It is true that introductions to the free will problem often raise the question why so much attention is still paid to whether compatibilism is defensible, given that so many theorists, including many compatibilists themselves, doubt or deny that Determinism is true. After all, if Determinism is actually false of our world, as many believe, why would it matter whether agents can have free will in deterministic worlds? Wouldn't the question be of merely academic or theoretical interest? Yet the issue of the defensibility of compatibilism is usually presented as if it were highly relevant to the question whether we – actual human agents – actually have free will.

The standard response to this question – about the contemporary relevance of compatibilism – is that even if Determinism is false of our world, something close enough to Determinism might be true to make the worries traditionally raised about the compatibility of free will with Determinism relevant to whether we have free will.Footnote ⁷ However, although this may explain why contemporary discussions continue to treat the question whether compatibilists can fulfil Task 1 as significant, it does not appear to explain the neglect of Task 2.

The most likely explanation for this neglect is an assumption (seldom stated) that the two-way compatibilists’ version of Task 2 is a relatively trivial one. In particular, it may be assumed that, unlike the version of this task that the libertarian faces, it does not make the compatibilist vulnerable to what I called, in the opening section, ‘the Indeterminism and Control Argument’. Why so? Well, it may be said, the two-way compatibilists’ Task 2 is simply the task of showing that agents can act freely in indeterministic worlds. By contrast, the task that the libertarian faces is much more demanding. As McKenna and Pereboom put it:

On the face of it, two-way compatibilists incur no such commitment concerning the locus of the ‘indeterministic breaks’ or undetermined events in indeterministic worlds in which there are free agents. This difference is obscured in simple statements of the Indeterminism and Control Argument such as the one I provided in the opening section. On reflection, one can see that what drives the argument (and similar arguments against libertarianism) is not the idea that indeterminism somewhere in the world would rule out control, but rather that indeterminism concerning the alternative possibilities that are regarded by the libertarian as crucial to free will would rule out control.Footnote ⁹

In what follows, however, I shall argue that matters are not so simple. In particular, I shall maintain that two-way compatibilists cannot so easily evade versions of the Indeterminism and Control Argument that they are so fond of pressing against the libertarian. I shall argue that they need to take seriously what I shall call ‘the challenge’ (to two-way compatibilism):

Although I believe that this challenge can be met, I also think that it is by no means a trivial one. That is the main thesis that I hope to establish in this paper.

3. Two-way compatibilism and ‘irrelevant indeterminism’

I have said that most contemporary compatibilists are two-way compatibilists. The crucial point is that most contemporary compatibilists do not commit themselves to the truth of Determinism, and yet believe that we have free will. Most contemporary compatibilists (i) are either agnostic about whether Determinism is true, or believe that Determinism is false (and hence that Indeterminism is true); and also (ii) are what I shall call ‘optimistic compatibilists’: that is, they believe that we do have free will. Evidently, then, they must, if they are to be consistent, be two-way compatibilists.

In addition, I think that all compatibilists should be two-way compatibilists. They should be so even if, unlike most contemporary compatibilists, they are firmly of the opinion that Determinism is true (and hence are committed ‘soft determinists’, according to the standard terminology). The reason is that the alternative version of compatibilism, ‘one-way compatibilism’, seems grossly implausible.

‘One-way compatibilism’ is the view that free will is compatible with Determinism, but incompatible with Indeterminism. Hence one-way compatibilism entails that agents cannot act freely in indeterministic worlds. But this seems absurd. There are many features that could make a world an indeterministic one, and yet which, it seems, no one could sensibly suppose would undermine the freedom of agents in that world. Consider undetermined events that fall into one of the following categories:

Many, if not all, of such events appear to be irrelevant to free will, as the following examples show.

To sum up: all three of these examples of indeterminism seem to be ones that a compatibilist should regard as irrelevant to freedom – irrelevant because they are obviously not ‘freedom-prohibiting’. That is, if one thinks that it is possible for agents to act freely at all (as compatibilists do), one should not think that the existence of these three types of indeterminism undermines their freedom of action.Footnote ¹⁰ (It is worth noting that ‘irrelevant indeterminism’ cannot be equated with ‘indeterminism that is isolated from the causal history of the agent's action’. As the second and third examples show, such isolation is not necessary for irrelevance, even if it is sufficient.Footnote ¹¹)

4. Is there an easy answer to ‘the challenge’?

The remarks I've just made may seem to reinforce the suggestion that two-way compatibilists have a very easy way of responding to what I have called ‘the challenge’, without making any concessions to the libertarian.

Two-way compatibilists must say that some kinds of indeterminism are ‘freedom-consistent’. (That is, after all, what defines them as two-way rather than one-way compatibilists.) But they can still say that some other kinds of indeterminism are ‘freedom-prohibiting’. Suppose, then, that there is a kind of indeterminism that libertarians regard as ‘freedom-enabling’ (i.e. as necessary for free will and free agency) but which the two-way compatibilist regards as ‘freedom-prohibiting’ (i.e. incompatible with free will and free agency). If correct, this hypothesis provides the two-way compatibilist with an easy answer to ‘the challenge’. And the plausibility of the hypothesis may seem to be supported by the fact that the types of indeterminism mentioned in the previous section seem irrelevant to freedom not merely from the point of view of a sensible compatibilist, but also from the point of view of a libertarian.

In order to evaluate this hypothesis, however, we must first identify exactly what type of indeterminism libertarians do regard as necessary for free will – as ‘freedom-enabling’. This is the task of the next section.

5. L-indeterminism

Let us consider what type of indeterminism the typical libertarian regards as a necessary condition for free will. I call this ‘L-indeterminism’. The typical libertarian (henceforth simply ‘the libertarian’) holds the principle (L):

A few comments on this are in order. First, the qualification ‘directly’ reflects the fact that some libertarians hold that an action can be free (indirectly free) if it is the result (at least in an appropriate way) of a previous directly free action. Secondly, for simplicity, I shall treat decisions and choices as belonging in the class of actions. Thirdly, the ‘does something other than A’ in condition (U) is to be interpreted in such a way that an involuntary ‘action’ such as dropping dead is not to count as ‘doing something’. That is, I take it, one of the lessons of my example of Alan and his book purchase.Footnote ¹³

Obviously, principle (L) involves an incompatibilist conception of the ‘could have done otherwise’ that many regard as a requirement for free agency. According to (L), an action done at time t is (directly) free only if the agent could have done otherwise, in a sense that entails that the agent's acting differently at t is consistent with the past up to t and the laws of nature.

Return to Alan and his decision, at 10 a.m. last Tuesday, to buy a book. The libertarian who holds principle (L) will say that if Alan's decision was a (directly) free one, then it was genuinely undetermined, by the past and laws of nature, up until 10 a.m. on Tuesday, whether he would make that decision. And not because it was genuinely undetermined whether he would stay alive to do anything at all, but genuinely undetermined whether he would make that decision rather than (voluntarily) doing something else instead.

In more detail: the principle (L) may be seen as incorporating the following:

and

1. (b) there is some other action B, such that it was undetermined by the past up to just before t plus the laws of nature whether the agent would do A at t or B at t. Footnote ¹⁴

Some libertarians add:

Once the libertarian's commitment to (L) is made explicit, it might be supposed (in line with the hypothesis suggested in the previous section) that the disagreement between the two-way compatibilist and the libertarian concerning the relation of free will to Indeterminism consists in the following. The libertarian thinks that the satisfaction of principle (L) is a requirement for (directly) free action. The indeterminism that (L) requires is not simply that the world be indeterministic, but also that the agent's action be itself an undetermined event that satisfies (a) and (b) of (L1). By contrast (according to this hypothesis), the two-way compatibilist will hold that, in order to be a free action, an action must not be one of which both (a) and (b) are true. (It might be an action that satisfies condition (a) of (L1),Footnote ¹⁵ but it must not satisfy condition (b) as well.)

In §7 I shall question whether this is, in fact, the appropriate stance for the two-way compatibilist to take. But to have a clearer sense of what is at issue, let us next consider what problems the libertarian's commitment to (L) appears to confront.

6. Problems for the compatibility of free will and L-indeterminism

If L-indeterminism (characterized as ‘the indeterminism needed for the satisfaction of principle (L)’) is what libertarians think free will requires, their view faces well-known problems: problems that suggest that this kind of indeterminism would actually undermine free will, by introducing an element of randomness or chance or luck that is incompatible with free and responsible agency.Footnote ¹⁶

There are various versions that this problem can take. I'll focus on two, which I call ‘the problem of control’ and ‘the problem of luck’.

The problem of control says that if an action satisfies condition (U), then the agent is not in control of the fact that the action occurs. Combined with the plausible assumption that unless an agent is in control of whether an action occurs, the action is not free, the problem of control leads to the conclusion that, far from providing a way in which actions can be free, the indeterminism required for condition (U) would actually prevent an action from being free.

The problem of luck says that if an action satisfies condition (U), then it is a matter of luck that the agent performs that action rather than some alternative action. Combined with the plausible assumption that if this a matter of luck, the action is not free, the problem of luck again leads to the conclusion that the indeterminism required for condition (U) would prevent an action from being free.Footnote ¹⁷

6.1. L-indeterminism and the problem of control

I'll illustrate what I call ‘the problem of control’ using an example taken from Peter van Inwagen's introductory book Metaphysics.Footnote ¹⁸ Van Inwagen says:

Let us look carefully at the consequences of supposing that human behavior is undetermined. Suppose that Jane is in an agony of indecision; if her deliberations go one way, she will in a moment speak the words, ‘John, I lied to you about Alice’, and if her deliberations go the other way, she will bite her lip and remain silent. We [are supposing] that there is a physically possible future in which each of these things happens … Let us suppose that there is a certain current-pulse that is proceeding along one of the neural pathways in Jane's brain and that it is about to come to a fork. And let us suppose that if it goes to the left, she will make her confession, and that if it goes to the right, she will remain silent. And let us suppose that it is undetermined which way the pulse will go when it comes to the fork …

Now let us ask: Does Jane have any choice about whether the pulse goes to the left or the right? … it is very hard to see how she could have any choice about that …

There is no way for Jane to influence the pulse. There is no way for her to make it go one way rather than the other … and leave the ‘choice’ it makes an undetermined event. If Jane did something to make the pulse go to the left, then, obviously, its going to the left would not be an undetermined event.Footnote ¹⁹

I have depicted the situation in the diagram below (Figure 1). Suppose that Jane makes the decision to speak (DS), and does so freely, according to the libertarian. According to the libertarian view we are considering, this implies that, up until the moment of decision, it is undetermined, by the past and laws of nature, whether she will decide to speak, or do something else. We also suppose (for the sake of the example) that the ‘something else’ is deciding to remain silent (DQ).

Figure 1.

According to the libertarian view we are considering:

1. (i) Both DS and DQ are physically possible continuations of the state of the world before t.

2. (ii) When Jane decides (at t) to speak, Jane's decision (DS) is a free action.

Since Jane's decision DS is a free action:

1. (iii) When Jane decides (at t) to speak, Jane controls which of DS and DQ comes about.

But this combination of views confronts an objection. The objection is that (i) and (iii) are inconsistent. How could Jane control which of DS and DQ comes about, given the assumption that both DS and DQ are physically possible continuations of the state of the world before t? Here, of course, it is important to note that the state of the world before t – the state that, ex hypothesi, does not determine which of DS and DQ comes about – includes all the facts about Jane, including her mental states – which in turn includes her preferences, desires, intentions, etc., up to t. Hence, if Jane does somehow control the outcome in this example, she must do so in a way that leaves those prior-to-t mental states intact.

The objection may be put in terms of the following simple argument, adapted from van Inwagen's discussion:

1. (1) The outcome of an indeterministic process is not determined by anything.

2. (2) To control the outcome of an indeterministic process would be to determine the outcome.

Therefore:

1. (3) It is not possible to control the outcome of an indeterministic process.Footnote ²⁰

So, in particular, Jane cannot control the outcome of the indeterministic process that (in our example) actually leads to DS – although DS is, according to the libertarian, supposed to be a free decision of Jane's.

So far, the libertarian's position may seem hopeless. However, as van Inwagen is well aware, some libertarians respond to this argument by appeal to a special kind of causation – agent-causation.Footnote ²¹ Such agent-causal libertarians may argue as follows:

‘There are two senses of “indeterministic process”. In one sense (Sense 1) an indeterministic process is one whose outcome is not determined by anything. In another sense (Sense 2), an indeterministic process is one whose outcome is not determined by preceding events and states of the world plus laws of nature. We may grant premise (2) of the argument above – that to control the outcome would be to determine the outcome. We thus agree that an agent cannot control the outcome of an indeterministic process in Sense 1. Nevertheless, an agent can control the outcome of an indeterministic process in Sense 2. Why so? Because an agent is not an event or state. And an agent can determine an outcome (by causing the outcome to occur, in an act of agent-causation) even if the outcome is not determined by previous events or states (including previous events or states involving the agent).’

This appeal to agent causation – as a way in which agents may determine outcomes that are otherwise undetermined – is notoriously controversial.Footnote ²² Even many libertarians regard it as unsatisfactory.Footnote ²³ Although I cannot pursue the issue here, it is fair to say that the ‘problem of control’ illustrated here is generally recognized to be a serious challenge to libertarians. And the problem arises, in part, because of the libertarian's insistence on condition (U) in principle (L).

6.2. L-indeterminism and the problem of luck

As well as facing this problem of control, L-indeterminism also confronts a problem of luck. The problem can be illustrated by considering a variant of the previous example, in which, instead of two possible decisions (DS and DQ) for Jane in the actual world, we consider the contrast between the actual world, in which Jane makes the decision DS, and a possible world that is a duplicate of the actual world up until just before time t, and has the same laws of nature as the actual world, and yet in which Jane (or a ‘Twin-Jane’ who is a duplicate or counterpart of Jane; it makes no difference) makes the opposite decision DQ. The libertarian must say that, if the availability of the alternative decision (DQ) in the actual world was required for Jane's actual decision (DS) to be (directly) free, then a world in which Twin-Jane makes that alternative decision is a possible one. We thus have the set-up represented in the diagram below (Figure 2):

Figure 2.

The existence of the possible world (which the libertarian seems bound to admit) appears to be an embarrassment to the libertarian. Given the possibility of the world in which Twin-Jane makes the decision DQ, how can it be other than a matter of luck that DS comes about in the actual world rather than DQ? Adapting the words of Mele's discussion of a similar case:

If there is nothing about [the agents’] powers, capacities, states of mind, moral character, and the like in either world that accounts for this difference, then the difference seems to be just a matter of luck. And given that neither world diverges from the other in any respect before t, there is no difference at all [between Jane and Twin-Jane] in these two worlds to account for the difference in [their] decisions.Footnote ²⁴

This problem of luck (if indeed it is a problem) seems to arise even for the libertarian who adopts the agent-causal response to the previous ‘problem of control’. Suppose we grant, for the sake of argument, that in the actual world Jane does determine that DS occurs, by agent-causing it. (She is the ‘undetermined determiner’ of DS, according to the agent-causal theorist.) For the sake of parity, we should say that Twin-Jane, in the alternative possible world, determines that DQ occurs, again by agent-causing it. But then why isn't it a matter of luck that Jane agent-causes (and determines) DS rather than DQ?Footnote ²⁵

7. Two-way compatibilism and L-indeterminism

We now have a characterization of a type of indeterminism – L-indeterminism – that the libertarian regards as necessary for free agency (because it is necessary for the satisfaction of the libertarian principle (L)), and yet which (in the light of the problems of luck and control) may appear also to be incompatible with free agency.

Now, of course, the compatibilist cannot accept the principle (L), since (L) entails that free action is incompatible with Determinism. In other words, compatibilists cannot accept that L-indeterminism is what I have called ‘freedom-enabling’ – that it must be involved in anything that counts as a (directly) free action.

But it does not follow, from this, that the two-way compatibilist must say that L-indeterminism is never involved in free action. There are, at least in theory, two possible options for the two-way compatibilist, Option A and Option B.

We might describe compatibilists who take Option A as ‘recalcitrant two-way compatibilists’, and those who take Option B as ‘liberal two-way compatibilists’.

It might seem obvious that the two-way compatibilist should take Option A, thus entirely avoiding the problems of control and luck presented in the previous section. This corresponds to the suggestion mentioned earlier in this paper (§4) concerning the two-way compatibilist's response to what I called ‘the challenge’ of reconciling free will with Indeterminism.

In spite of this, though, I don't think it is obvious that the compatibilist should go for Option A rather than Option B. There are two reasons for this.

8. Two-way compatibilism and the problems of control and luck

Suppose, then, that two-way compatibilists do go for Option B. Then, of course, they must provide answers to the problems of control and luck. How are they to do that, without undermining their case against the libertarian?

It seems clear that, in response to the problem of control, the compatibilist (from now on this means ‘the Option B two-way compatibilist’, unless otherwise indicated) should jettison the notion that, in cases of free action where L-indeterminism is present, the agent manages somehow to determine the otherwise undetermined outcome. The question is why this won't play into the hands of the libertarian.

The compatibilist, we are now envisaging, says that an action can be free even if it satisfies condition (U), and it was thus undetermined, up to the time of action, whether the agent would do that action or something else. The suggestion is that an action can satisfy this condition, and yet be free, even if it is not, strictly speaking, true that the agent controls or determines whether that action occurs or, instead, some other action. In other words, the two-way compatibilist is envisaging that, where it is the case that there are alternative possibilities of the kind involved in L-indeterminism, and the action is nevertheless free, its being a free action does not require control over which of the alternative possibilities comes about. According to this view, the typical libertarian is doubly mistaken: mistaken in holding that such control is a requirement of free action, and mistaken in thinking that L-indeterminism provides scope for such control.

The libertarian's appeal to undetermined alternative possibilities is typically motivated by the idea that unless there are such alternative possibilities, the agent cannot really determine what happens in a way that is required for free agency. By contrast, the two-way compatibilists’ view can be that the alternative possibilities provided by condition (U) are not freedom-enabling, and need not even be freedom-enhancing. Two-way compatibilists can regard the indeterminism in question as, so far as it goes, a hindrance to freedom. In other words, according to the two-way compatibilist who takes Option B, we sometimes act freely not because of, but in spite of, L-indeterminism, just as we sometimes act freely in spite of the fact that we might have dropped dead before the time of action.

This last consideration bears on the contrast between the two-way compatibilist's attitude to L-indeterminism and that of Robert Kane's event-causal version of libertarianism.Footnote ³¹ Kane rejects the agent-causal response to the problem of control in favour of a view according to which, whichever way perplexed Jane decides, she is responsible for the decision, although she may have ‘diminished control over each option considered separately’.Footnote ³² He also concedes that ‘indeterminism … does seem to diminish rather than enhance agents’ voluntary control’.Footnote ³³ Nevertheless, Kane's view differs significantly from that of the Option B two-way compatibilist. For Kane, unlike the compatibilist, thinks that the presence of L-indeterminism is required for an action to be directly free. He therefore faces, as the compatibilist does not, the problem that his view makes free agency conditional on there being actions (and a sufficiency of such actions) that involve L-indeterminism.Footnote ³⁴

The two-way compatibilist who adopts Option B can, I have argued, handle the problem of control without capitulating to the libertarian. But what of the problem of luck? On the face of it, the Option B two-way compatibilist appears to confront a stark dilemma. Either L-indeterminism brings with it a degree of luck or chance that prevents the action from being free, or it does not. If it does, then Option B is incoherent. If it does not, then there is no challenge for the libertarian to face.

In reply, however, we can observe that there is an important difference in the resources available to the Option B compatibilist and the libertarian in responding to the problem of luck. The compatibilist may say that what makes it the case that the element of luck (if it is present) does not undermine freedom is that the action satisfies sufficient conditions for free action of a type that would be unacceptable to a libertarian – unacceptable because they can also be satisfied by actions in which no element of L-indeterminism is involved. Evidently this response is unavailable to the libertarian. Hence (whether or not the response is ultimately an adequate one) Option B compatibilists may consistently make this response without undermining their case against the libertarian's claim that L-indeterminism is not only consistent with, but actually required for, free agency.