I agree with 90% of what Michael Moore writes in this very interesting, clever and thoughtful paper. In particular, I’m with him in thinking that those interested in arguing that Libet’s experiments show people to lack responsibility, of the sort that the criminal law cares about, on the grounds that they show people to lack free will is a complete non-starter. (Moore says that the argument for this claim would make a sophomore philosophy major blush. I might add the better freshmen to this group, too.) I also think that he’s right in claiming that the conclusion cannot be reached by interpreting the experiments as showing that conscious awareness of one’s choices plays no causal role in generating action. (In contrast to Moore, I would reach this conclusion by arguing that when consciousness is necessary for responsibility it is not so by virtue of its causal role in generating action except in a sense of “consciousness” that Libet’s experiments do not show to be absent from the causal chain leading to action; but let that pass.) I’m also with him in holding that the remaining option, namely that the experiments show that our choices are causally irrelevant to what we do—that brain states cause two independent things, bodily movements and, separately, choices—is not supported by the experimental results; there’s just no reason to think that this “epiphenomalist” picture accurately describes what’s going on. But, and here is the 10% where we disagree, Moore also holds that even if it were shown that our choices are causally irrelevant to what we do—even if our choices were in the “dangler” position in the diagram of the causal structure of action—ordinary responsibility judgments would not be threatened. They would be, it seems to me.

Moore’s argument for the contrary has the following broad structure: We can learn from a solution to Newcombe’s Paradox that there are (seemingly) odd circumstances in which people can be in control of future events even when their choices to act are not among the potential causes of those future events. If we interpret the Libet experiments as showing that our choices are causally irrelevant to what we do, then it turns out that the odd circumstances in which this is possible are actually ubiquitous; the way in which the man in Newcomb’s Paradox exercises control is actually the way all of us do it all the time. But, since control is what is needed for responsibility, it follows that Libet’s experiments do not threaten responsibility. This is a very clever approach. The trouble is that the solution to Newcombe’s Paradox fails in its own right and fails to show what Moore takes it to show about control; in fact, things that Moore says in this very paper about the nature of action, and in many of his published works, commit him to rejecting the lesson about control that he wants to take from his “solution” to Newcombe’s Paradox.

The solution to Newcombe’s Paradox that Moore offers is motivated by a perfectly accurate observation about certain ordinary cases. Sentences like the following are often true:

(a) If you follow through, you’ll hit the ball squarely.

Since (a) is true, it makes sense to give the following advice to the man standing at the tee with a desire to hit the ball squarely: “Follow through!” If he follows your advice and hits the ball squarely, it will also be true that he hit the ball squarely by following through. That would be bad advice to give him, if (a) were not true of him.

Now Moore takes the case of Newcomb’s Paradox to be parallel. Looking at Moore’s Figure 8, consider the following sentence:

(b) If you decide to take one box (at t4), you’ll get more money than you would get by taking two.

Is (b) true? As I know Moore is aware, it’s actually not entirely clear. After all, whatever is in the opaque box at t4, if you decide to take both boxes you get that plus $1000, which is more than you would get by taking just the opaque box. But, if we interpret (b) as what is sometimes called a “backtracking conditional”, then it’s true: if you decide to take one box, then that’s what the Predictor predicted earlier; in which case, there’s $1,000,000 in that box; but it’s empty if you decide to take two. The conditional is backtracking because we imagine its antecedent satisfied and then see what must have been true earlier for that to be the case; we then check to see what follows from that earlier condition, and if the consequent follows, then the conditional is true. Of course, the same is true of (a): we imagine you following through and then see that, if you did, some prior condition must have been met and that prior condition would have resulted in your hitting the ball squarely. So let’s grant to Moore the truth of (b). Since (b) is true, it follows (thinks Moore) by analogy to the golf example that it makes sense to give the following advice to the man standing before the two boxes trying to decide whether to take one or both: “Take just one!”. As before, if this man follows your advice and ends up with more money than he would have gotten by taking both boxes, he will have done so by choosing to take just one box.

But let’s slow down. Is the truth of (a) reason enough to advise the golfer to follow through, or does more need to be true to make that good advice? More needs to be true. In particular, the following must be true:

(c) Advising the man to follow through might cause him to hit the ball squarely.

If (c) is not true, then the fact that (a) is true fails to provide sufficient reason to advise the man to follow through. And, in fact, (c) is true, for what the advice might do is to cause the man to form an intention to follow through and that intention will cause him to engage in a bodily movement that results in the ball being hit squarely.

But then it seems that “Take just one box!” is good advice to the man facing Newcombe’s Paradox only if the following is true:

(d) Advising the man to take just one box might cause him to get more money than would have been gotten by taking two.

The trouble is that (d) is false. At time t4, no matter what happened earlier, there is more money in the two boxes together than there is in just one of them. So the man will make more by taking both.

Notice that in granting that (b) is true, I am granting that the one-boxer makes more money than the two-boxer—$999,000 more, to be exact. Thus, as Nozick was well aware when he first discussed Newcombe’s Paradox, we find ourselves in the odd position of having to say that sometimes what it is best to advise a person to do is not what you know will allow him to get the best of his possible options. Rather than trying to deny this result, as Moore does, we ought to embrace what it means: the norms governing rational deliberation and decision are not grounded merely in the fact that by following them we do better than we would otherwise.

As Moore has surely noticed by this point in reading this commentary, my objection to his solution to the Newcombe Paradox depends on interpreting it as a paradox of rationality, which is how it is usually interpreted. As usually interpreted, the puzzle is that it seems that Reason tells you to do something that will result in your failing to get the best of your options; that’s supposed to be paradoxical given that the only reason to be rational is that by being that way you get more good stuff. But what it makes sense to advise someone to choose, and what it is rational to choose are one and the same thing. Given that, what Moore has not shown is that it is rational to choose one-box. However, what Moore is really concerned to say is not something about rationality, but something about control of the sort that is necessary for responsibility. We might agree that it is irrational to choose one box while still granting that it is in the chooser’s control whether he receives $1,000,000 or only $1,000. Should we grant that? If we do, then Moore has probably succeeded in showing that the epiphenomenalist interpretation of the Libet experiments fails to undermine ordinary responsibility judgments. But I don’t think we should grant that. In fact, I think that Moore himself is committed to denying it by other things that he says.

In holding that the man in Newcomb’s Paradox can control, at t4, whether the opaque box contains $1,000,000 or nothing, Moore is accepting the following general principle about control:

(e) If (1) S chooses to A at t2 if and only if F occurs at t1, & (2) F occurs at t1 if and only if E occurs at t3, then at t2 S is in control of E’s occurrence at t3.

But this principle admits of counterexamples. Say that I suffer from a compulsive handwashing disorder. It’s quite severe: I inevitably choose to wash my hands every day precisely at 11 AM even though I never have good reason to do so. I make that choice as a result of my disorder and would not make it in the disorder’s absence. Yesterday, in fact, I had to leave an important business meeting in the middle in order to wash my hands. This made my boss so angry that he put a pink slip in the mail to me yesterday; it will arrive at noon today. Right now it is 10:59 and I’m starting to feel my urge to wash my hands. Now, if I choose not to wash my hands at 11:00 then I don’t have the disorder; any world in which I manage to choose to refrain is one in which I lack the disorder. Consider such a world: in it, I also did not leave the meeting in the middle yesterday—after all, I only left because of my disorder and in the world we are imagining, I don’t have the disorder. In that world, my boss didn’t have any reason to get angry and so didn’t send me a pink slip, and so I won’t receive a pink slip at noon today in that world. Does it follow that it is up to me, under my control, whether I actually receive a pink slip today at noon? Of course not. The pink slip is already in the mail! Maybe I can control whether or not I choose to wash my hands at 11:00 today. But even admitting that, and even admitting that I can control everything that would be different as a causal consequence of that choice on my part, does not commit one to the absurd result that I can now control whether or not I receive the pink slip. My choice at 11:00 today can’t have any effect on the progress of yesterday’s mail. Here, S is me, A is washing my hands at 11:00 today, F is my disorder, and E is my receiving the pink slip at noon today. So much for the principle on which Moore relies, namely (e).

To see this more clearly, consider Moore’s case of Paul Revere. In the example, if there’s a brain event of the sort that is registered as the readiness potential, then that causes two things: a choice to flex and, through a separate causal stream, a light to go on. As in the usual story, one if by land, two if by sea. Now imagine that Revere chooses to flex just once, say, and he does it knowing, as was actually the case, that two lights are required to warn the rebels properly. Only one light goes on in such a case. (This is diagramed in Moore’s Figure 9a.) Moore makes two claims about such a case: He thinks that Revere performed the act of misleading the rebels. And he thinks that Revere “should be fairly hung for treason”. But not so fast. Consider first the claim that Revere misled the rebels, that he performed that act. Notice that if we accept that claim, then we are denying a causal theory of action of just the sort that Moore claims, in the early part of his paper, and in a great number of his published works, to endorse. Nowhere in the causal history of the rebels coming to have the mistaken belief about the British attack is a choice on Revere’s part in favor of that result, or even a choice in favor of the result that causes it, namely the appearance of only one light. Revere made that choice, but it didn’t cause the light to go on. Tracing backwards, the causal chain goes from the rebel’s false belief to the single light’s appearance, to the brain event registered as the readiness potential, and back to some prior set of brain events (which happen to also cause the choice). If the relation that Moore takes to hold between the rebel’s false belief and the choice to flex here—they are both part of the same causal network, for instance, even though they are not in the same causal stream—is enough to make it the case that Revere misled the rebels, then so much for the causal theory of action. On the causal theory of action, Revere misled the rebels only if his choice to do so caused the rebels to have the false belief, and this is not true here. It will take a lot more than this to convince me to abandon the causal theory of action. If there are intuitions to the effect that Revere misled the rebels, then I side with theory over intuition here.

When it comes to the second claim, namely that Revere should be hung for treason, Moore is tapping an intuition, to be sure, but it is not the one that he thinks it is. To be “fairly hung for treason”, Revere needs to perform a treasonous act, and be culpable for it. Attempting treason, which he has done, arguably, by flexing only once, is not enough. For treason itself, Revere would have to at least have performed the act of turning on only a single light. Since Moore can only reach the conclusion that Revere has performed a treasonous act by rejecting the causal theory of action, and since he himself does not want to take that step, I think he should rethink his claim that Revere ought to be hung. Still, there’s a case to be made that Revere is culpable for treason. To be culpable for treason, the causal history of the single light’s appearance must include that in virtue of which a person is culpable. This might be true in Revere’s case, even though his choice is not in that causal history. Someone who holds, for instance, that choices are important to culpability because of what they tell us about the agent’s settled principles for recognizing and responding to reasons might very well reach this conclusion. Both the brain event registering as the readiness potential and the choice to flex are caused by Revere’s commitment to treat facts about what will hurt the American cause as reasons to act in those ways. If that’s what matters for culpability for treason, then, sure, Revere in this example is culpable for treason. But he has not performed a treasonous act, for the causal history of that which hurt the American cause—namely the appearance of only one light—does not include his choice to send a misleading signal. Choices contribute to the voluntariness only of the events they cause. But even granting that, they still might matter to culpability not because of that contribution, but because of what they tell us about their causes. In this case, the state of the agent that the choice discloses (the commitments about reasons) also causes the injury, but the choice does not. It seems to me, therefore, that the disapproving attitude one feels towards Revere, in the example, might be based simply on the recognition that he is culpable for the injury to the American cause. However, on this line, he performed no voluntary act that caused that injury and so is not “fairly hung”.

It seems to me, therefore, that someone who wants to push the anti-responsibility line that Libet’s experiments are sometimes used to support ought to try to develop evidence for the epiphenomenalist conception of what is going on when we act, the conception under which our choices are causally irrelevant to what we do. But, as Moore notes, the Libet experiments don’t support that view and, at the moment anyway, there’s no other reason to think it is correct.

can i quote you?