The Impertinence of Frankfurt-Style Argument

The Impertinence of Frankfurt-style Argument

(For INPC)

4,203 words

In the debate between compatibilists and incompatibilists about the conditions for moral responsibility, the Frankfurt-style counterexample argument against the principle of alternative possibilities has been defended with remarkable energy and ingenuity.[1] This counterexample strategy is, nevertheless, a red herring. What will be on offer here, however, is not a defense of the principle Frankfurt cases purport to falsify. Rather, I will argue that even if the cases are taken to be legitimate counterexamples to the traditional requirement of alternative possibilities, they can make no contribution to the compatibilist cause. Contrary to popular philosophical perception, the examples are quite simply irrelevant to the free will debate.

To refresh our memories, recall that Frankfurt cases typically involve some fail-safe mechanism that is designed to go into effect just in case a given agent is about to deviate from a single desired path. Thus, imagine that Paul has insulted John. Paul’s insult comes after his own normal and rational deliberation. Nothing suggests the presence of any responsibility-undermining factors at work in the action sequence. However, Frank, a third agent, has all along wanted Paul to insult John. Indeed, Frank has implanted a chip in Paul’s brain that allows him to “read” his intentional states and make changes to them if necessary. Since Paul’s intentional states were at each point consistent with his insulting John, Frank has simply continued to monitor Paul’s mental content. If, however, Paul had begun to show any inclination not to insult John, then Frank would have activated the chip in such a way as to bring about an inclination in Paul to insult John. This seems to be a case in which Paul is morally responsible for his action even though he lacked alternatives with respect to it; even though his insulting behavior was unavoidable. This is to say that the following principle of alternative possibilities (or any one of its near kin) seems to be false:

(PAP) a person is morally responsible for something she has done only if she could have done otherwise.

What I want to emphasize is that in the crush of effort over the past 35 years to defend and attack the counterexamples themselves, the argumentative limits of these examples have, for the most part, been ignored. I propose to address this deficiency and explore the boundaries of Frankfurtian reasoning. My conclusions will be largely skeptical. The question I will be focused on is: what can reasonably be thought to follow from the falsity of PAP as demonstrated by the counterexamples?

To motivate my minimalist answer to this question, consider the following case, to which I will frequently return. Suppose Dave and Patty are watching the night sky. They notice a bright object moving in various unpredictable ways through space, first left, then up for a few seconds, then down and to the right, etc. Patty suggests that it must be a plane or helicopter or some other “man-made” object. It can’t be a comet or meteor, she claims. Dave is the argumentative sort. He challenges Patty’s conclusion. In response, Patty outlines the following argument:

(1) That object is moving in unpredictable ways.

(2) If an object moves in unpredictable ways it is not a natural object.

(3) Therefore, that object is not a natural object. It must be an artifact.

Suppose, now, that Dave raises an objection to premise (2) of Patty’s argument by noting that, according to the received view in physics, sub-atomic particles move in unpredictable ways, but they aren’t artifacts. Given Patty’s broad confidence in science, Dave is eventually able to convince Patty to concede that sub-atomic particles and only sub-atomic particles move in unpredictable ways. Dave then points out that Patty now has a counterexample to (2). There are objects that move in unpredictable ways but that are also natural objects. When Patty gathers her thoughts, however, it seems to me that she should not be very impressed by the objection.

Why not? Well, in one sense, it is rather obvious that Dave’s objection is simply impertinent. We can see, along with Patty, that sub-atomic counterexamples to (2) are irrelevant to the context of argument. We don’t see sub-atomic particles in the night sky, after all, so we don’t expect Patty’s argument under these conditions to take them into account. Though, strictly speaking, Dave will have forced Patty to give up her second premise, we in no way expect Patty to abandon her argument’s conclusion. This impertinence provides the intuitive explanation for why Patty should not be impressed with Dave’s objection. Given a valid argument with intuitively plausible premises, an objection is impertinent if it depends on a proposition (or propositions) that can serve equally well in establishing the conclusion, holding fixed the context of the initial argument.

With respect to the debate between compatibilists and incompatibists over the conditions for moral responsibility, I now accuse proponents of Frankfurt-style argumentation of a similar impertinence.[2] The famous counterexamples to PAP, I contend, depend on considerations that ultimately give no solace to compatibilism and provide no rampart against incompatibilism.

I. Counterfactual Stability

Let’s return to Frankfurt’s original paper. In it, the classic examples certainly appear to demonstrate the falsity of PAP, but they do not explain it. The explanation comes only after the examples have motivated the relevant intuition. Here is the analysis Frankfurt offers:

The fact that a person could not have avoided doing something is a sufficient condition of his having done it. But, as some of my examples show, this fact may play no role whatever in the explanation of why he did it. It may not figure at all among the circumstances that actually brought it about that he did what he did, so that his action is to be accounted for on another basis entirely. Even though the person was unable to do otherwise, that is to say, it may not be the case that he acted as he did because he could not have done otherwise. Now if someone had no alternative to performing a certain action but did not perform it because he was unable to do otherwise, then he would have performed exactly the same action even if he could have done otherwise…

This, then, is why the principle of alternate possibilities is mistaken. It asserts that a person bears no moral responsibility—that is, he is to be excused—for having performed an action, if there were circumstances that made it impossible for him to avoid performing it. But there may be circumstances that make it impossible for a person to avoid performing some action without those circumstances in any way bringing it about that he performs that action.[3]

Frankfurt’s confidence, then, in the irrelevance of alternative possibilities for moral responsibility comes from recognizing that it is possible for an agent to find himself in circumstances that make a certain course of action inevitable, but that ultimately play no role in the his choice or action. The agents in the counterexamples can, then, be held responsible because they would have done the same thing even if, contrary to the stipulated facts, alternatives had been available to them. The inevitability-creating circumstances turn out to be otiose.

We can put this by saying that while agents under Frankfurtian circumstances may not possess alternative possibilities, they do appear to possess “counterfactual stability.” Counterfactual stability can be defined in this way:

A person S is counterfactually stable with respect to some action x = def (1) S has done x, and (2) S could not have done other than x, and (3) S would have done x even if she could have done other than x.

Notice, now, that counterfactual stability plays a crucial role in provoking us to accept the conclusion that Paul is responsible for insulting John in spite of his inability to do otherwise.

It appears, then, that we are prepared to countenance the strict irrelevance of alternative possibilities for moral responsibility in cases where agents possess counterfactual stability with respect to their actions. We can put this point by saying that Frankfurt-style cases seem to appeal to a latent commitment to something like the following principle of responsibility without alternative possibilities:

RWA: If a person S is morally responsible for doing x, then if S does not have alternative possibilities with respect to doing x, then S is counterfactually stable with respect to doing x.

The issue is now whether, in accepting the falsity of PAP on the basis of Frankfurt-style reasoning, we are (implicitly, at least) committed to RWA.[4] Put slightly differently, the question is whether or not the counterexample strategy depends on RWA (or something similar). I believe that it does. I have already pointed out Frankfurt’s repeated appeals, in his effort to generate agreement, to considerations from which the principle has been constructed. But here is the most important point. In accepting that PAP is false on Frankfurtian grounds, one needs to be accepting nothing more than that a person can be morally responsible in the absence of alternative possibilities only if the person is counterfactually stable. The examples themselves force us to concede nothing further. RWA has been constructed so as to capture this minimal commitment.

After all, RWA forcefully explains our intuitions in Frankfurt-style cases. Even if we grant (what might be denied in another context) that these examples genuinely eliminate alternatives, we are still powerfully inclined to concede that agents in these circumstances are nevertheless responsible. Since the agents act for their own reasons, we have every reason to believe that RWA is satisfied. They would have done the same thing even if they could have done otherwise—even if the intervener had not been present. What drives the counterexample argument against PAP, then, is the RWA intuition. If we accept that PAP is false, it is, in part, because we accept that RWA is true.[5] We have found, then, that our rejection of PAP relies on a premise that, as we shall soon see, can do the same work as the one it has functioned to supplant.

II. Making the Impertinence Explicit

To make the dialectical impertinence of the Frankfurt-style counterexample strategy explicit now, I need to make good on the promise of an argument. Specifically, the argument needs to lead to the incompatibilist conclusion that causal determinism would, if it obtained, undermine moral responsibility. Furthermore, this argument needs to depend only on the elements of the initial context of argument and on that which is entailed by our acceptance of the argument for the falsity of PAP; namely, RWA.

In diagnosing his own case, Frankfurt insists that Jones’ responsibility can be recognized by paying attention to the fact that the factor that made it impossible for him to do otherwise “play[s] no role whatever in the explanation of why he did it.”[6] The factor can simply be subtracted from the case and things will unfold the same way. Now, counterfactual stability under the looming eye of an intervener is one thing. Under deterministic causal laws, it is something entirely different. Returning to our initial case, recall Paul’s misfortune at having found his way into a Frankfurt scenario. Frank has implanted the chip in his brain and is waiting to see if he will insult John. Paul insults John on his own, so Frank can sit back and enjoy the fun. As we recognized before, we are all likely to agree that Paul is responsible for the insult. If Frankfurt is right, this is because if we subtract Frank’s chip from the case, it makes sense to say that Paul still remains—with his autonomously formed inclinations to insult John. But suppose now that Paul cannot do otherwise than insult John not because of Frank’s presence and power, but because the world is causally determined. Then to assess Paul’s responsibility for insulting John, we need to know if Paul would have insulted him had determinism (rather than a computer chip) been subtracted from the story. Does Paul remain after this subtraction? There is no good reason to think so. At least, there is no good reason to think that a morally relevant version of Paul remains. After all, though the chip is embedded in Paul’s brain, it is not embedded in his causal history in the way determinism is. Thus, the chip can come out without systemic affect. Determinism, on the other hand, is entangled in his psyche. Since the actual world is deterministic, it must be true that Paul’s present inclinations, dispositions, and intentions are, at least in part, a product thereof. That is, they are, at least in part, a product of the past states of the world together with the laws of nature. To subtract the factor that made it impossible for him to do otherwise in the actual world would be to subtract these inclinations, dispositions, and intentions. This would seem to undermine any intuition that Paul would still do the same thing. If he wouldn’t, he can’t be responsible for what he has done. Even granting the falsity of PAP, determinism still undermines responsibility--via RWA.

This has been the casual and motivational version of the argument. No doubt there are obscurities and points of controversy. Still, I think the above line of reasoning can be largely vindicated with somewhat more precise argument. Again, remember that my strategy is to show that Frankfurtian considerations are simply irrelevant. I do this by showing that the incompatibilist can get her argument out of the very same resources that the counterexamples presuppose. Here, then, is the more formalized argument for incompatibilism employing RWA now rather than PAP. Let us begin by assuming that an agent S has performed some action x in the actual world Wa. Then,

(1’) If determinism is true in the actual world (Wa), then S does not possess

alternative possibilities with respect to doing x in Wa.

(2’) If determinism is true in the actual world (Wa), then S does not possess

counterfactual stability with respect to doing x in Wa.

(3’) If S does not possess alternative possibilities with respect to doing x in (Wa) and

S does not possess counterfactual stability with respect to doing x in Wa , then S is

not morally responsible for doing x in Wa.

(4’) Therefore, if determinism is true in the actual world (Wa), then S is not morally

responsible for doing x in Wa.

This argument, I grant, is not as crisp or intuitive as its traditional counterpart. But it need not be. It needs only to be forceful; which it can be shown to be. Since this argument is perfectly general, it will apply to any S and x. This is to say that (4’) amounts to the standard incompatibilist conclusion [i.e. premise (3) in the traditonal argument].

Notice that (1’) is simply (1) from the traditional incompatibilist argument instantiated for the actual world and a particular person. That is surely part of the original argumentative context and so cannot be a legitimate point of contention. (3’) is the contrapositive of RWA, also instantiated for the actual world and a particular person. If the proponent of Frankfurtian reasoning hopes to resist this argument, then, the concern will have to be with premise (2’). I will now argue that there are very good reasons to believe that (2’) is true.

I will argue for (2’), then, in two stages. First, I will show that there are insurmountable obstacles to demonstrating that it is false. Second, I will offer a probability argument to the conclusion that it is true.

To demonstrate that (2’) is false for any given agent S, the defender of Frankfurt’s argument would need to persuade us that, even though determinism is true in S’s world, S is indeed counterfactually stable with respect to some action of hers. That is, the defender of Frankfurt’s argument would need to persuade us that S would have performed the very same action even if, contrary to fact, determinism had been false. This, I claim, cannot be done. My reasons for insisting on the impossibility of this demonstration are all tightly connected to various problems that emerge, on the assumption of determinism, for assessing the counterfactual claim that an agent would have done what she in fact did even if she could have done otherwise.[7]

Assume, now, that Justin has lied to Lori in the actual deterministic world. The principle reason that Justin cannot be shown to possess counterfactual stability under these conditions is that there appear to be logically or modally incoherent concepts at work in the relevant counterfactuals. To see this, notice that confidence in Justin’s counterfactual stability under determinism requires that there be a coherent and relevant positive answer to the question “Would Justin have lied to Lori in another possible world (W) just as he did in the actual world (Wa) [where (Wa) is governed by strictly deterministic laws and (W) is not]?”

Is there, then, another possible world at which Justin lies to Lori even though he could have done otherwise? There are two ways of making such a world coherent. First, this possible world could be exactly like the actual world in being causally determined right up to the moment that Justin is about to decide whether or not to lie to Lori, at which point the laws of this world now permit indeterministic outcomes. Call this W1. A second potential way of making the required world coherent would be to think of it as a world whose laws are indeterministic throughout its history and in which there is an indeterministic realization of Justin under the relevant conditions. Call this W2.