Tuesday, October 12, 2010

Alternate possibilities and forcing someone to choose something

Standard counterfactual-intervener Frankfurt stories require two ingredients: (i) a sign telling the intervener what the agent would choose were she not interfered with and (ii) a way for the intervener to make the agent choose what the intervener wants. There has been a lot of discussion in the literature on the first ingredient and not enough on the second.

For one might argue as follows:

  1. (Premise) If x freely chooses A over B, then x could have chosen B over A instead.
  2. (Premise) If it is possible to force someone to choose A over B, then (1) is false.
  3. Therefore, it is not possible to force someone to choose A over B.
Here, (2) follows from the Frankfurt cases, assuming there is some solution to the sign problem (I think the best bet might be to insist that the sign does not need to be infallible to rule out an ability-entailing "could-have"—cf. this piece by Harrison). And (1) is a reasonable version of the Principle of Alternate Possibilities (PAP). It is worth noting that PAP is really very plausible. Testimony to its plausibility is that for several centuries even compatibilists tried very hard to preserve it.

Now, most people working on free will accept that it's possible to force someone to make a particular choice. But I think this may be mistaken.

As a warmup, observe that ordinary threats don't do the job if one understands the contrastive nature of choices—that a choice is a choice of one option over one or more others. Suppose I am deliberating between Pad Thai and Green Curry Chicken. I love both, and it's a hard (but insignificant) choice. Now, threatening my life can make me choose Pad Thai. But it doesn't make me choose Pad Thai over Green Curry. Rather, it makes me choose Pad Thai and life over Green Curry and death. So it doesn't force me to make one of the two choices open in the original choice situation—it replaces those two options with a new pair of choices that is much easier to choose between.

Now, one might think that brainwashing could force me choose Pad Thai over Green Curry Chicken. But I am not so sure. Brainwashing could make me be utterly disgusted at Green Curry Chicken and unable to consider eating it. If it did that, however, I wouldn't be choosing Pad Thai over Green Curry, because Green Curry wouldn't be an option for me. (When I choose to wear a shirt in the morning in the summer, I don't choose a shirt over a fur coat—the latter option does not enter into the deliberation.) A less extreme form of brainwashing could make me very likely to choose Pad Thai over Green Curry Chicken, by accentuating my liking for the former or creating an intense dislike for the latter. But I see no reason to think this would make it impossible for me to choose Green Curry Chicken, unless it made me not see any value in choosing Green Curry Chicken over Pad Thai. And if I saw no value at all in Green Curry Chicken over Pad Thai, I wouldn't be choosing between the two. It would not be choice, but a shoo-in. If I went for Pad Thai, it would no more be a case of my choosing Pad Thai over Green Curry Chicken than in this world I choose to eat Pad Thai over eating a chopstick. Similarly, God and the saints in heaven do not choose good over evil.

Maybe you're not convinced by what I've just said, but it's at least not absurd. And I think the folk do have some intuitions in favor of (3). That no one can make you choose something is prima facie a pretty plausible sentiment. We are willing to abandon it when we think about threats and brainwashing, but if those cases can be handled—and I think they probably can—then (3) remains ultima facie plausible. And (1)-(3) provides additional reason to believe (3).

I think (3) is fairly plausible if choices are necessarily exercises of a non-natural causality. But, interestingly, (3) could be true even if naturalism is true. Suppose, along the lines of Kane's view, that a choice of A over B is constituted by the collapse of a quantum state a|A>+b|B>, mixed between a state |A> favoring A and a state |B> favoring B, with both a and b non-zero. The state can collapse into the pure state |A> or into the pure state |B>, and which way it collapses determines which intention you form. Now, there is a difference between a mixed quantum state collapsing into a pure state, and a mixed quantum state being changed into or replaced with a pure state. A physical process can perhaps force a change of a mixed state into a particular pure state, but likely that change wouldn't count as a collapse, just as a non-gravitational process can make a particle be attracted to another, but it wouldn't count as gravity. As far as I know, it is quite reasonable to say that no physical process could force a particular mixed state to collapse into a particular pure state. Could a non-physical process force a mixed state to collapse in a particular way? I doubt it. It seems essential to the nature of the collapse of a|A>+b|B> that its probabilities be governed in the right way by the weights a and b, while if some non-physical process forced a transition from a mixed state to a particular pure state, the probabilities of the transition wouldn't seem to come from the weights a and b of the mixed state in the way they do in collapse, but from the nature of the forcing process.

The question whether God could force a quantum state to collapse in a particular way is a difficult one. But it's not obvious that both: (a) God's causing a transition from a mixed to a pure state would count as determining or forcing or the like (here a very insightful paper by W. Matthews Grant is relevant); and (b) the transition would still count as a collapse.

Even if free choices are not naturalistic or are naturalistic but are not collapses, nonetheless the collapse case gives us a picture, I think, of how something could be such that it couldn't be forced without destroying its nature. Someone can force me to transition from a state of being undecided between A and B to being decided for A, but not all such transitions are choices, and on the view I am defending, none of the forced ones can be choices.

If the above is right, then (1) can be strengthened by omitting "freely" in its antecedent. All choices have alternate possibilities. I once heard Nuel Belnap remark, after a talk on free will by a job candidate, something along the lines of: "I always thought the matter was simple. To make a choice, you have to have choices."

If (1) holds without "freely" in its antecedent, this leaves the interesting question whether all choices are free, or is it merely that all choices have alternate possibilities. I think the claim that all choices are free may be correct, but needs to be correctly understood. Some choices have severely constricted alternatives, and choices are contrastive in such a way that the alternatives need to be mentioned to fully describe what was chosen. The ordinary murderer can choose to shoot x over leaving x alone. The person brainwashed into murder may only be able to choose to shoot x over stabbing x. Both are in some sense freely choosing to shoot x, but what they are responsible for depends on the fuller story about what they are choosing over. (If shooting x is less cruel than stabbing x, then the brainwashed person may actually be praiseworthy for shooting x—for he is shooting x instead of stabbing x.) And I am inclined to say that what they are doing is different. I should also add that a fuller description of the choice would be that one chooses to do A for R over doing B for S, where R and S are bunches of reasons that one is impressed by.

Note: I am not claiming here that choices are the only thing one is responsible for. Though they may be the only cases in which one assumes responsibility.


Heath White said...

FWIW, Thomas and Descartes (and probably others, Rogers Albritton more recently) hold that it is a matter of logic that, if one has free will, the exercise of one's will cannot be forced. If it's forced, it's not one's (necessarily free) will, and contrapositively.

Alexander R Pruss said...

Thanks for the Albritton reference. Here is a recent Analysis paper along the same lines.