Compatibilism and Indeterminism

One of the benefits of compatibilism about determinism and moral responsibility is that our ordinary conception of ourselves isn't "held hostage" to any arcane scientific discoveries.  Or, at least, this is a sentiment that I've heard expressed by some compatibilists (notably John Fischer).  So even if it turns out that the appropriate interpretation of quantum mechanics (for instance) is deterministic, we don't have to give up our view of ourselves as responsible.

Given this sentiment, I take it that it's as much a part of compatibilism that moral responsibility is compatible with indeterminism, as well.  But one rarely hears this claim defended.  And now I'm wondering about it.

Consider, for instance, Fischer and Ravizza's account of guidance control, which is supposedly both necessary and sufficient for moral responsibility (at least regarding the freedom-relevant condition of MR), no matter whether determinism is true or false.  According to F&R, an agent is responsible for an action that issues from the agent's own, moderately reasons-responsive mechanism.  And they spell out what it means for a mechanism to be moderately reasons-responsive in terms of how the mechanism responds at other relevant (though not necessarily accessible) possible worlds.

My worry is -- is their account of guidance control compatible with indeterminism?  Suppose that the world is fundamentally indeterministic in a way that doesn't get cancelled out, so we have indeterminism at the macro level as well.  If the indeterminism is at the right place, we'll now have two worlds that are exactly the same in terms of the past and the laws of nature, but that differ with respect to the action that issues from the mechanism under consideration.  Is this a problem for moderate reasons-responsiveness?

More specifically, I'm worried that something like the Luck Objection that usually is raised against libertarian accounts of free will might be legitimately raised against this particular compatibilist account as well.  Does the agent really have guidance control if what action actually issues from the mechanism under consideration appears to be a mere matter of luck?  And if in some of the relevant possible worlds, the mechanism issues in A, but in other relevant worlds the mechanism issues in B, is the mechanism still appropriately reasons-responsive, even though the reasons would seem to be the same in both worlds?

I haven't thought this through too much, so I'm not saying that this objection is a good one, but I'd be interested to hear what people think.

Comments

[Mark Smeltzer]

There doesn't seem to be a special problem with respect to indeterminism: even if determinism is true, there can be errant causal chains that over-determine an agent's action, thus negating or superseding the agent's MRR mechanism.

The F&R position has to do with the metaphysics of responsibility, but does not dictate the practical stance we should have on responsibility. It stands to reason that given the possibility of errant causal chains (be they determinant or indeterminate) we should take care to deal out praise and judgment.

This is a large part of why many of us stress the importance of looking at the holistic picture of the agent's character as depicted by their actions over a longer course of time. This has to do with trying to understand an agent’s actions in context and form strong inductive reasons to believe that particular actions did (or did not) issue from the agent's MRR mechanism.

F&R have purposefully remained silent here. Their intent, thus far, has been to develop a generic framework for others to extend.

I am in agreement with you that more work needs to be done in this area. Fischer’s recent paper “The Cards That Are Dealt You” indicates that Fischer is well aware of the problems surrounding the concept of Luck.

My own view is that the problem domain of luck is epistemic in nature. Consequently I believe that the problem of responsibility and the problem of luck are only tangentially related, that they are best addressed separately, and that one’s account of responsibility should guide the approach one takes to solving the problem of luck -- and not the other way around due to the fact that luck does not seem to present a special problem for the metaphysics of responsibility.

[Kip Werking]

To answer Neal's question:

Guidance control can be compatible with indeterminism. Suppose that the spin of an electron in a distant galaxy is indeterministic: whether it has one spin or another doesn't necessarily follow from any prior state of the univese. Then a human in this galaxy might still have guidance control.

I think Neal might be asking this question: is guidance control *always* compatible with indeterminism? No. Suppose that the motion of *every* atom in the universe is indeterministic: where the atom moves (or stays) does not necessarily follow from any prior state of the universe. This would be, except for monstrous coincidences, pure chaos. In that case, one's actions do not follow from one's mrrm, because they do not follow from anything.

So my understanding is that responsibility, on F&R's view, can survive some, but not all, kinds of indeterminism.

Neal notes that philosophers rarely note the compatibilisty of compatibilist-fw with indeterminism. I agree that this point is underemphasized. I think there is another point that deserves more attention: Galen Strawson's claim that determinism is unfalsifyable.

I think Strawson is right. Any claim that the world does not obey uniform laws is vulnerable to the possibility that we are just ignorant of the real (and more complext) laws of nature (Russell would always say that he suspected physicists would later discover the more subtle laws government quantum physics; his friend and fellow anti-war activist Einstein also felt committed to determinism; neither believed in free will). And any claim that the world does obey such laws is vulnerable to a future exception, or to the possibility that the world follows the same path indeterministically (a monstrous coincidence).

So, if this is right, then libertarianism is not held hostage by physicists (not in any practical sense). That alone is interesting. There is another important question to ask about this virtue, however: is not-being-held-hostage-by-physicists a virtue because it is desirable (and so, a happy coincidence), or is it a virtue because it renders the given proposition (compatibilism) somehow more likely to be true? I think it is important to emphasize that the answer must be the former, and not the latter. It would be a happy coincidence if whichever prevailing concept of moral responsibility is not held hostage, in this sense, by physicists. But this feature doesn't render semicompatibilism any more likely to be correct; or rather, that libertarianism does not have this feature should not be a deficiency of that theory. The nature of moral responsibility might be inconvenient but nonetheless real.

[Neal]

I think there's a middle way between asking if guidance control is compatible with one little bitty bit of indeterminism in some far off galaxy, and asking whether it is compatible with indeterminism reigning supreme.

The interesting question, I think, is whether guidance control is compatible with the picture of human agency that comes out of the most plausible libertarian pictures of agency. That is, if there were indeterminism in just the right places with respect to the mechanisms that produce our actions (the 'right places' being those places where the libertarian wants or needs indeterminism), then can the agent still exercise guidance control?

Libertarianism is supposedly at risk of being shown false because it may turn out that all of our actions are determined. But is guidance control equally at risk of being shown problematic because it may turn out that all our actions are produced by indeterministic processes?

[Jonathan Farrell]

I think Neal is right to think that there is a special problem with regard to indeterminism here, so I disagree with Mark. I also disagree with Mark’s claim that the problems related to luck are epistemic ones (at least I don’t think that they are _only_ epistemic ones, which I think is Mark’s claim). However, I’m not sure that I understand what Neal’s first worry is, so what I say below mightn’t be useful at all. To make sure that my _whole_ post isn’t useless, before I get started on Neal’s worry, I’ve got a point about terminology that I think is useful.

Before I get started, though, there’s the issue of terminology. I’d rather that the term ‘compatibilism’ be used to refer to the doctrine that says that free will is compatible with determinism (and that says nothing about the relationship between free will and indeterminism). I don’t have any particularly good arguments for using the term in this way, apart from that (I think) this is the traditional use of the term. I have, however, come across a different term – ‘hypercompatibilism’ – that refers to the doctrine that free will is compatible with both determinism and indeterminism. (John Bishop uses this term in his review of Timothy O’Connor’s Persons & Causes (Philosophy and Phenomenological Research LXVI No.1 Jan 2003) – I’m not sure if it’s Bishop’s coinage or not.)

Neal, I’m not quite sure what your first worry is (this is probably because I’m not too familiar with the details of the notion of guidance control.). I'll describe what I think your worry might be, and then say why I think you might find it worrying. Finally, I'll say something about why I think that the Luck Objection is going to be as problematic for hypercompatibilists as it is for libertarians.

Here’s what I think your first worry is: if it’s possible (only because of indeterministic factors) that I x and possible that I y, then if I actually x, one of the relevant possible worlds we must consider is that in which I y. So if my x-ing issues from my moderately reasons-responsive mechanism, then it must be that, had I y-ed, my y-ing would also have stemmed from my moderately reasons-responsive mechanism. (If I’ve got the worry wrong, then there’s no need to read most of the rest of this post. There is, however, a paragraph at the end that is concerned with Mark’s claim that there’s no special luck issue with regard to indeterminism that (I think) doesn’t depend on my having understood Neal’s first worry correctly.)

Kane deals with a parallel issue in The Significance of Free Will (and elsewhere) where he requires (roughly) that an agent’s actions are free only if they have reasons for performing all possible actions (i.e. all those that might actually be performed, given the past and laws). It seems that F&R might have to require that a similar condition be met if an indeterministically caused action is to count as one flowing from an agent’s guidance control.

At least two interesting consequences seem to follow from having such a condition (although the second is really a continuation of the first). The first is that it would seem to make free actions less common than we might otherwise think. This is because, presumably, at least sometimes, we ‘tick all the boxes’ with regard to the action that we perform (i.e. meet the conditions that must be met concerning what actually occurs), but not with regard to the actions that we don’t, but (actually) might have performed. This may or may not be a problem for libertarians (I don’t think that Kane sees it as a worry, for instance), but doesn’t seem to cause difficulties for hypercompatibilists (they can just note that free agents in deterministic universes will, other things being equal, tend to perform more free actions in a given period of time than those in indeterministic universes).

The second consequence is that the situation seems to be worse if there are more than two possible futures, given the past and the laws. For example, it might be that given the actual situation, I will x with a probability of 0.5, y with a probability of 0.499, and z with a probability of 0.001. And it might be that if I either x or y then my action will issue from my moderately reasons-responsive mechanism, but that if I z, then my z-ing will not stem from such a mechanism. If the condition I described above must be met, then if I x or if I y, neither of these actions will be free.

Presumably, situations in which all three possible future actions ‘tick the boxes’ will be rarer than those in which only two possible actions must pass muster. And presumably the situations will be rarer still if there are four, five, or more possible actions.

There seem to be at least three ways of proceeding if we want to claim that free will is possible in indeterministic universes. We could try to provide good reasons for thinking that there are only two possible actions (this is the way Kane goes). Or we could aim to explain why it is that unlikely-but-possible actions can be ignored, i.e. that there is some threshold of probability of z-ing below which what happens if I z is irrelevant to whether or not my x-ing or y-ing is free (perhaps by arguing that the non-actual world in which I z is not particularly nearby (I’m not sure how this could be argued for, though)). Or we could admit that we act freely less often even than the first consequence had led us to believe. There may be other options, but it seems to me that these three each require more than, or are more unattractive than, the situation faced in deterministic universes. So it seems that there is a problem particular to indeterminism here.

I also agree that the Luck Objection will be relevant, if I understand this objection correctly. Even if my only possible actions are either to x or to y, and whichever I in fact perform issues from my moderately reasons-responsive mechanism (so my suggested condition is met), it still seems to be a matter of brute chance whether I actually x or y. And if it is just a matter of good/bad luck that I x rather than y, then this seems relevant to whether or not I x-ed freely.

[Neil]

I think there is a problem here too, but I suspect that it's not quite as big as one might think. If you take F&R to defend the view that MMR leaves no hostages to empirical fortune, Neal might be able to show that they're wrong. But Fischer (at least) doesn't think that. In "Frankfurt-Style Compatiblism" (Watson, ed, Free Will , 2nd ed), he writes:

I do not believe that our personhood and moral responsibility should be insulated from every empirical discovery about the world. Rather, I believe that these central notions should be resilient with respect to this particular issue - whether the equations that describe the macroscopic universe are universal generalizations or probabilistic generalizations with extremely high probabilities attached to them' (p. 211).

So Fischer seems to agree that enough indeterminism is incompatible with his view.

[Mark Smeltzer]

"So, if this is right, then libertarianism is not held hostage by physicists (not in any practical sense). That alone is interesting."

Wouldn't the actions of L-free agents be susceptible to the same causal conundrums that you believe C-free agents are susceptible to?

I'm an agent causal compatiblist. I don't believe that agents are reducible to physical properties and/or physical relations (viz., emergent properties) -- most agent causal libertarians would side with me on this point. However, all actions that issue from the agent's choice mechanism (whether it has guidance control or regulative control) would then be subject to the laws of physics.

The only kinds of libertarians don't seem to have any special problems with indeterminism are event causal (Kane) and simple indeterminists (Ginet) accounts... but those are totally wussy, and I think most of us would prefer something a bit more bad ass.

[Mark Smeltzer]

Jonathan,

"I also disagree with Mark’s claim that the problems related to luck are epistemic ones (at least I don’t think that they are _only_ epistemic ones, which I think is Mark’s claim)."

Certainly hard incompatibilists will be inclined to disagree with me here, but for different reasons that I intended to address. Outside that context, what I meant was something to the effect that even if we can provide an adequate account of the metaphysics of responsibility, the problem of luck still remains -- and the problem of luck is metaphysics neutral.

In other words, I disagree with philosophers like Galen Strawson and Thomas Nagel who claim that the problem of luck should inform our views about the nature and/or value of responsibility.

Neil,

I think it stands to reason that Fischer would accept that any kind of causal force (whether deterministic or not) which significantly impedes the proper of function of an MRR mechanism would militate against guidance control.

Kip,

You might also want to check out Randolphe Clarke's recent article "Agent Causation and
The Problem of Luck" (Pacific Philosophical Quarterly, 2005). He seems to agree that luck is a problem that libertarians need to deal with.

[Kip Werking]

Neal,

To answer your middle-ground question: such a question (is this indeterministic account of human action compatible with free will) needs to specify a particular indeterministic account of human action. Jonathan mentions Kane's view, here, and I think Kane's efforts to show that an agent might have duel reasoning or control over different choices suffice to show that such an agent might still be reason responsive. To give a crude example, if I have a reason to kill a man, and a reason to not kill him, and how whichever reason "wins" is indeterministic, then I might still have still have F&R style moral responsibility, because I'm still reason responsive. There would be a continuun of responsibility, which tracks how much indeterministic "noise" there is in human nature, and both Kane and Fischer endorse concepts of "partial" responsibility.

Indeed, a responsible Kanean agent (one who satisfy both Kane's AP and UR requirements) would seem to be sufficient, but not necessary, condition for being a responsible F&R agent. Kane achieves this result by adequately addressing how human nature might be indeterministic but still largely reason-responsive. But he fails to adequately address the larger concerns of, for example, Strawson and Nagel. In that case, one wonders (as very many have), what soft libertarianism can achieve that compatibilism cannot---is the indeterminism even necessary?

Kane tries to satisfy such concerns by appealing to the concept of Ultimate Responsibility. But careful analysis shows that this concept does not provide the libertarian with what he really wants: being causa-sui.

Mark:

I've had Clarke's paper on my desktop since Bloglines alerted me to its presense. I've scanned it but have not read it in depth (I also read Clarke's recent article on the Basic Argument). Given my own Strawson/Nagel-style view, I find such articles to be fascinating. I'm also looking forward to Mele's new book on luck.

I describe the larger concerns which Strawson and Nagel have about luck in my new article "Who's Afraid of Creeping Exculpation?":

"Once one recognizes the importance of being causa-sui, one also recognizes a time shift within the free will debate. Traditionally free will has been conceived of as just an immediate or “time-slice” power. But what these considerations show is that the problem which motivated the free will dispute is historical. PAP attempts to compensate for the fixity of the past by denying the fixity of the present. But once the libertarian grants that a given self is problematic, there is no self remaining to exploit these current alternative possibilities. Agents cannot create themselves ex-nihilo. From this perspective, one recognizes the charge that free will does not exist, if such a charge is to capture the genuine dilemma which motivated the dispute, is not a claim about immediate or “time-slice” powers. Rather, the charge is that one’s entire life is ultimately arbitrary. Everything, on this view, is an accident of birth. Because one’s life is ultimately arbitrary, one cannot claim responsibility for it."

[Paul Torek]

Jonathan,

It is not that "there is some threshold of probability of z-ing below which what happens if I z is irrelevant to whether or not my x-ing or y-ing is free." Rather, as long as the probability is low, you are still moderately reasons-responsive, because your reasons are highly relevant. It is only when you have a high probability of doing something for no reason, that your freedom and responsibility plummets. At least, that strikes me as a natural reading of moderate reasons responsiveness.

[Jonathan Farrell]

Neil and Paul: thanks for your remarks. Your two comments seem to imply that Neal’s first worry (or at least what I thought Neal’s first worry was) isn’t in fact particularly worrisome. (It’s still probably going to be the case that agents in indeterministic universes perform less free actions than those in deterministic universes, but perhaps this isn’t too terrible a consequence.) It seems that the luck objection remains, however – even if the possibility of (non-reasons-responsively) z-ing doesn’t stop my actual act of x-ing from stemming from a moderately reasons-responsive mechanism, it still seems to be the case that I have been lucky to x, while my counterpart in another possible world has been unlucky to z.

Mark: I think you’re right to say that “even if we can provide an adequate account of the metaphysics of responsibility, the problem of luck still remains” (although I might prefer ‘… problems related to luck …’). But it doesn’t seem to follow from this that luck isn’t also relevant to the metaphysics of responsibility. (Which I think is what you’re saying. But perhaps you’re not. Or perhaps you’re not saying that this _follows_, but only that it’s true. Or perhaps we’ve got different ideas of what ‘the problem(s) of luck’ is/are.)

[Mark Smeltzer]

Jon,

As I said earlier, I think the problem of luck is an epistemic one. We can lay out a principle account of responsibility that has to do with the metaphysics of responsibility (e.g. Fischer's account of guidance control), but when it comes to applying that account in the real world, the problem of luck bears practical significance.

The problem of luck is epistemic because if we knew with Cartesian certainty which actions an agent bears responsibility for, it would not exist.

[Randy Clarke]

I've argued in several places that as long as actions are caused, and caused by the things and via the routes that causal theories of action require, indeterminism doesn't threaten responsibility. E.g., "Indeterminism and Control", APQ 1995, and chs. 3 & 5 of _Libertarian Accounts of Free Will_. I don't think the possibility of an empirical discovery that these causal transactions are governed by nondeterministic laws should worry us.

On the other hand, a discovery that the things we think are actions are not caused at all would, I think, undermine our view of them.

[Neal]

Randy,

Good point...I've read your arguments and was impressed by them. Given that there are good arguments out there defending libertarianism from the problem of luck (or similar problems), perhaps I should weaken my claim to the following:

If the problem of luck were a problem for the libertarian, then it would be a problem for the compatibilist (of the MRR variety).

What's interesting about this, I think, is that it's not open to a compatibilist to argue against libertarianism by invoking the problem of luck. The compatibilist should be just as interested as the libertarian is in addressing the problem.

Does that seem right?

[Randy Clarke]

It seems to me there are two problems that can be raised by invoking luck, and I'm often unsure which of them the critics of libertarian views are raising. One could charge (1) that indeterminism at the required juncture wouldn't make a favorable difference--it wouldn't make us responsible if determinism at that juncture precluded responsibility. A compatibilist could raise that objection without herself worrying about indeterminism. Or one could charge (2) that indeterminism at the required juncture would make an unfavorable difference. Compatibilists who make that charge might well worry about how the empirical work will turn out.

[Kip Werking]

Neal:

What is this problem of luck? I only remember two problems of luck from the literature: (i) the luck objection to libertarianism and (ii) the Strawson/Nagel view that "luck swallows everything. According to (i), whichever way an agent indeterministically decides is not up to the agent but up to luck. Given that compatibilism does not invoke any indeterminism, I don't see how (i) could also be an objection to compatibilist views.

Perhaps you are referring to (ii), which is an objection to all "free willist" views?

[John Fischer]

Just to say a few words. I (and Mark Ravizza) presented a set of conditions for guidance control which we claimed were consistent with both causal determinism and causal indeterminism. We then said that the account of guidance control could be part of an overall defense of compatibilism; also we suggsted that be employed by incompatibilist. What we did not claim is that the account in itself solves all the problems for either the compatibilist or the incompatibilist. In previous work I have been more concerned with seeking to address the problems for the compatibilist. Soon I hope to write something defending the litertarian against the "luck" objection. I suppose the best way to see the account of guidance control, in an indeterministic context, is as part (although not all) of a defense of moral responsibility.

[V. Alan White]

Just as an exercise in perspective, replace this discussion of guidance control with that of agent causation. It seems to me that much of what Tim O'Connor (e.g.) wishes to wrench out of his concept of agents as causes is the same as what John Fischer wishes to wrench out of reasons-responsiveness: the role of reasons as control over what we do. Of course O'Connor adds in the feature of indeterminism which I think Fischer does not. It seems significant to me that even van Inwagen recognized in his "Freedom Remains a Mystery" that the indeterminism of an O'Connor view simply fails to provide any useful sense of an agent's being confident of his/her promise to be able to do something since he/she must acknowledge the chance (under what van Inwagen believes is a plausible interpretation of indeterminism) that he/she could do other than what is promised. (Thus the stunning implication by PvI that a revised Mind argument may well show that free will is incompatible with indeterminism.) Is this not in essence an argument that moral luck is an inherent component of any view that merely adds indeterminism into its otherwise reasons-responsive account of deliberation?