Blog Coordinator

« Gardeners—I mean Flickerers—in the NYTimes on Experimental Phil and FW | Main | Agency and Responsibility in Scientific American »



Feed You can follow this conversation by subscribing to the comment feed for this post.

Should be:

There is one notable reply that goes virtually unnoticed: Lynne Rudder Baker’s “The Irrelevance of the Consequence Argument,” Analysis 68.1 (2008): 13-22.

Thanks Joe--I had read both pieces yet had not stuck them together heretofore in my thinking to try to say something about their connection. Let me give you my take.

PvI's article is admirable if for nothing more than it is an archtype of carefully crafted intellectual honesty. By that I mean he not only admits his mistake about Beta, but gives a careful analysis of the error. I think he taught me a more valuable lesson as a philosopher in his up-front attitude than by any particular claims he made.

That said, Baker's piece is also a stunningly clever piece: determinism falls out of the picture due to the revision in Beta to be only about exact logical access. And I think she's right.

But here's my take on why she's right. Revised Beta produces a valid CA because the restriction of exact access--restriction to the actual world (and PvI seems pretty clear on this; read his note 9 on there being no relevant nonactual worlds to consider here)--results in a circular argument. We can't do otherwise in a deterministic world because, well, we can't do otherwise in that world. Baker's argument expands understanding on this in terms of analysis of logical access to nonactual regions, but it is the concept of exact access itself in revised Beta that delivers the logical goods on that. But as I see it--and admittedly as just the excellent second-class philosopher I am I may have missed something more subtle here--revised Beta isn't subject to counterexample because it restricts attention to only the actual world. But it seems to me that that restriction is inadequate to deliver any useful insight about determinism and freedom.

This is admirable, but, speaking of "logical space", where does van Inwagen provide a mechanism that would distinguish "able to", "access to", "bring it about", etc. from random outcomes?

At the risk of thread jacking, a slightly different question, Joe: Let "N(p)" in the third argument stand for "p and no one now has any choice whether p" and let "Psubscript0" stand for a true proposition at any time prior to now. Is the argument now a general argument for incompatibilism?

I agree with your analysis, Al! I think it is spot on.

attlee: your question is VERY interesting. Here's how I would put it: the incompatibilist can't begin to appeal to the consequence argument until he responds to the Mind argument. In other words, if we haven't quite figured out how our actions can be up to us in an undetermined world, it can't be a real problem to note that our actions don't appear to be up to us in determined worlds either.

Neil: Complicated question. I asked John Fischer about the essential features of the consequence argument and he said that it appeal to the past and laws in an effort to argue for incompatibilism. Given that standard, this seems to be a general argument for incompatibilism.

Maybe a better question would be, is your argument a good argument. It doesn't appear to me to be a good argument. I don't see how the kind of necessity that propositions have because of their mere pastness can transfer onto facts about my present and future actions. Just think of drunk-driving cases. An important issue is whether anyone HAD a choice about whether Psubscript0, not just whether one has a choice about it now.

By the way, this is why I scratch my head when it comes to arguments like van Inwagen's second argument and Fischer's basic argument. They don't appear to make assumptions about the remote past yet if they don't, I don't see how they can work.

But Joe, my revised argument is iterative: for any past fact, I had a choice about it only if the argument could not be run about that past fact. But it can. So there are no facts about which I ever had a choice.

Whatever your argument does, it does not show that determinism is a problem for free will. You can't just assume that "there are no facts about which I ever had a choice." That has to be established. But there is simply nothing in this set of assumptions to establish that fact, nothing to ground the no-choice that you hope to thrust onto the rest of the truths (via the thesis of determinism).

Imagine my Adam* world (Analysis 2010), where someone's existence is eternal (or cyclical). You can say Adam never had a choice, so he has no choice now. But I can just as easily say there is no reason to deny choice from Adam previously (nothing to ground the claim of no-choice), thus his choices now are not threatened. I don't see how your assumption of no-choice has more grounding than my assumption of choice. You have to ground the no-choice before you can assume it is passed along. After all, you are the one offering the argument! I don't claim to have a proof of the free will thesis or a proof of compatibilism. I just claim that these theses cannot be shown to be false.

No matter how you cut it, determinism is merely a connecting thesis, merely a thesis connecting the truth of some propositions with the truth of others. I admit that transfer principles like Beta have an intuitive pull. Thus, one might use the connecting thesis of determinism, together with some grounding assumptions about our lack of choice wrt certain propositions, then then thrust that no-choice onto all other truths. But that still leaves you with the task of establishing the grounding thesis. Without that, the argument is incomplete.

I'm not getting it. Why doesn't iterating the argument constitute a grounding thesis? If the argument establishes for any p that with regard to p s has no choice whether p, then the compatibilist can't claim that there is some prior proposition with regard to which s might have had a choice (by the way, I remain a compaibilist - it seems to me that though the consequence argument might be sound, there are perfectly good senses of relevant words - able, and so on - that the compatibilist can appeal to).

Maybe I'm not getting it! How does the argument establish "for any p that with regard to p s has no choice whether p"?

Here is a better set of questions. If I stole some candy while I was in High School, can I say now I did it unfreely because I have no choice about the matter now? Can the drunk driver tell the police officer, "Given the amount I drank and that I was driving 100 miles an hour, I had no choice about whether I hit the child in the crosswalk"?

My answer in both cases is "No." So the mere fact that some fact is a fact about the past can't in and of itself show that no one has or ever HAD a choice about it. The mere pastness of the past can't ground no-choice in away that can be transferred through entailment, even if the relevant transfer principles are valid.

In order for your argument to prove incompatibilism, you have to establish that past facts have some feature in addition to their mere pastness that can be transferred onto all other facts. It could be the existence of a remote past. But then this is just a version of the third argument. My impression was you were offering something different. I mentioned the Adam example since if there were some feature besides the mere pastness of the past that you could use to ground the no-choice transfer, you could not it in that case. But what is it?

Suppose I say that given some fact about the past - let's make it MY past to better fit with your response - and the laws, I now have no choice about whether p. You reply that nevertheless I might have a choice about the fact about the past. But I now say that with regard to *that* fact I had no choice because there is a fact about my more distant past which together with the laws entails that at that time I had no choice about that fact. I can repeat it for any fact about my past. Eventually we get to the first fact about my life. For my argument not to be successful, that first fact about my life had better be a choice of mine. I doubt that that's coherent (for one thing, choices take time).

Two things. First, the proposition that your life had a first fact -- that you have a remote individual past, a time prior to your birth -- was not an initial assumption of your argument. This is relevant since at first it seemed as if you were giving a version of van Inwagen's second argument or maybe even Fischer's basic argument, which appears NOT to make such assumptions.

Second, if you do make such an assumption, I have a reply (Analysis 2007). The assumption that you have a remote past is a contingent assumption. But incompatibilism is not a contingent thesis. How can a non-contingent thesis rest on a contingent assumption? It can't. Again, consider Adam who has no remote past. We could just keep playing the same game with him yet we never get to the first fact of his life (there is no first fact of his life; he is eternal).

The key point here that everyone seems to miss is that there doesn't seem to be anything different about you and Adam. You could both exist in determined worlds. Yet your argument shows that you are not free and it can't seem to show that Adam is not free. (1) How is it that determinism is inconsistent with freedom? (2) Why can't you propose an argument to show that Adam is unfree, given determinism is true?

I have a draft of an article on the consequence argument that I'm submitting to the Internet Encyclopedia of Philosophy for consideration. If anyone would like to see it, let me know.

Joe, I can't remember from your interesting article, but do you say that the incompatibilist can fall back on the claim that:
Necessarily, for all x, if determinism is true, then if x is not eternal, x is not free.

Folks have tried to argue for weak compatibilism -- determinism plus some contingent theses entails that no one has free will. As far as I know, no one has argued for the claim above.

Both strike me as ad hoc. Compare similar theses in metaphysics:

Determinism plus some contingent theses entails dualism.

Necessarily, for all x, if determinism is true, then if x is not eternal, presentism is false.

It seems that if determinism is a problem for free will one should be able to show it no matter what the other features of the world (or people in the world) were.

My last comment seems to have got lost. It is that my argument does not depend on a first or remote fact, merely a past fact with regard to which the agent never had control. Even Adam satisfies this condition.

Neil: not eternally looping Adam--such an Adam always controls future events by asymmetrical closed-time assumption. Joe's point is that there are worlds beyond the usual assumption of unidirectional non-looping time--and those are modally relevant. I'm unsure about modal constraints of a topological nature, but Joe's point is forceful if so unconstrained.

Alan, it seems to me false that for every event in Adam's past (which, to be sure, is also his future), Adam controls that event. Such events will include all kinds of nonactions which are not the causal consequence of any action of Adam: his left eyebrow twitching, his perspiring, his heart beating. Given these fact and the laws, the argument seems to me to go through since these facts and the laws entail that he now has no choice whether p.

We are creatures in a world where some things existed before we did and time isn't circular. What if our world is deterministic? Suppose we have an unimpeachable argument to the effect that, in that case, we don't have free will. What then to say about creatures (sort of) like us who are eternal or in whose world time is circular?

Who knows? But their similarity to us is grounds for saying that they, too, lack free will. Theirs is a weird case that we might judge on the basis of the clear ones.


Perhaps I'm misunderstanding your argument, but I don't think it works. In the original P_0 is a complete description of the world, and so (if determinism is true) we can use it and the laws of the universe to deduce any partial description of the (later) world. But your P_0 looks to me like only a partial description of the world (one small fact that Adam/x has no choice about), and so it seems there might be many propositions that do not follow from your P_0 and the laws the universe, even given determinism. Thus we'd be missing one of the LHS elements of your equivalent of Beta.

(Of course, the fact that no one's mentioned this before makes me think I might be missing something very obvious....)

Mark, some searching on the Internet shows that PVI does indeed define P_0 as a description of the world. But neither Joe (in his book) nor Vihvelin her SEP article follow him; rather they define P_0 as a true proposition about the remote past (remote is Joe's stipulation). It seems that that might not be enough, because such a proposition could entail nothing about S's action at t. But a complete description of a world at a time is not needed either: rather, any past fact that is not controlled by S and which is part of the causal chain leading to S's phiing at t will do.

Thanks, Al!

Neil: My last comment to you was dilemmic, so confusing. I wasn’t sure whether you were giving something like van Inwagen’s (1983) first or third argument OR something like his second argument. The former do but latter does not explicitly endorse contingent assumptions about the remote past.

It seems clear now that you’re giving something like the second argument, since you say: “my argument does not depend on a first or remote fact, merely a past fact with regard to which the agent never had control.” Call arguments like this “nontransfer versions” (of the consequence argument), since they do not explicitly make use of either transfer principles or principles about the remote past.

Suppose we assume that you think that I have a response to the first or third argument: they make use of contingent assumptions and thus are not arguments for the thesis of incompatibilism. (More on this in the next post in response to Randy’s point.)

You write, in response to Alan: “… it seems to me false that for every event in Adam’s past (which, to be sure, is also his future), Adam controls that event …” etc.

But this strikes me as either clearly false or at the very least unsupported. Why assume that “every event in Adam’s past” is such that he lacks control over it? I’ll grant that if we switch back to the first or third argument, then you can ground that assumption, for in both of those arguments there are explicit assumptions about the practical necessity of the remote past.

Later in response to Mark’s nice post, you write: “… a complete description of a world at a time is not needed either: rather, any past fact that is not controlled by S and which is part of the causal chain leading to S's phiing at t will do.” Where “doing” is grounding the practical necessity (in Joe-speak).

I don’t see how this can work. Determinism can only provide what computers provide: garbage in, garbage out. Of course, it might follow that substance in, substance out (by transposition). Nonetheless, the kind of necessity that determinism thrusts on the world is purely relative. If you don’t have a grounding principle, you’ve got nothing but your own incompatibilist intuitions.

If you try to say otherwise -- any old time-slice will do -- you are led to classical free will skepticism (no one can do otherwise) without the aid of determinism. If any old time-slice will do, then each time-slice will do. And you are going to have to deny a lot more than determinism to contain this. Why doesn't this apply to this very moment, for instance, or the very next moment (without the assumption of determinism)?


What do we know? We know that the third argument suggests that determined creatures with a remote past are unfree; we know that the same argument does not lead to the same conclusion about eternal Adam. There doesn't appear to be any relevant difference.

This seems to be a draw. We can go from the apparent fact that noneternal creatures are unfree & the claim that there is no difference to the conclusion that Adam is unfree; or we can go from the apparent fact that Adam is free & the no difference claim to the conclusion that the third argument is deceptive but not conclusive.

I don't see a reason for going one way rather than the other. But that just means that the third argument is not conclusive.

One other set of points. I don't think that the weirdness of the Adam case matters. For one thing, the weirdness seems irrelevant to the issue of freedom. I wonder whether I'm free. Does the question of whether I've always existed or not matter? I can't see why or how.

Another thing. It is weird to think that I might hover two feet above the earth for no reason at all. But this is precisely why nomological necessity is different from absolute necessity. The weirdness is irrelevant to the issue. Sometimes, weird possibilities that have absolutely nothing to do with the actual world establish important metaphysical claims.

Warning: long response ahead (it's too late for me; save yourselves).

Joe, actually I think we agree more than you think we do. You say to Randy that the third argument is not conclusive. But that’s just what I think too. I said previously that I think there is one way of understanding key terms on which it is sound and another on which it is unsound; understanding the terms in one way begs the question against compatibilism and in the other begs the question against incompatibilism. I thought that Oscillating and Eternal Adam were supposed to break the stalemate and show that we have good reason to prefer the compatibilist reading to the incompatibilist; that’s what I deny.

At the risk of boring you further, a quick response to each horn of your dilemma. First, on the any-old time slice will do claim. I have here a very good argument: from authority. On p. 49 of Campbell’s authoritative book Free Will, he defines P_0 as “any true proposition about the remote past”. Since I claim, and you don’t dispute (in this context) that pastness plus lack of control is sufficient, rather than remoteness, it seems I can help myself to Campbell’s stipulation with the remoteness condition dropped. Actually I went a little further than Campbell, and said that the past fact had better be causally linked to Adam’s phi-ing. If the past fact plays some kind of role in shaping Adam’s phi-ing – by causing or entailing some sufficient or necessary condition of his phi-ing – then it can play the role of P_0. I agree that much more needs to be said here about which past facts can play that role. Very detailed philosophical work is needed here (which, frankly, I am not remotely equipped to do). Prima facie, though (it seems to me) that it is likely for either Oscillating or Eternal Adam that there will be myriad such past facts available to play the role.

Now the second horn. I take it the claim is this: if you drop the remoteness condition, then you have no right to suppose that for every event in either Adam’s past, Adam lacked control over that event. Well, I can rely on the authority of Campbell once more to claim that it is at least rational to go either way on this question (or so I interpret his remarks on Galen Strawson’s claim that being an ens causa sui is incoherent: Campbell puts forward Eternal Adam as a counterexample, and then confesses that he completely unsure whether the counterexample succeeds). To make further progress here, I think a great deal of very detailed work is needed. If we make progress on the first horn – analysing determinism and the causal relations it involves – we might also succeed in settling whether facts outside an eternal or oscillating agent’s control must exist. The other thing we will need to do is to provide an analysis of ‘control’. In Levy’s forthcoming book, he argues that control, in the sense at issue in the free will debate, is a rational power, and that this fact entails that control depends upon options with weighs (values, significances) that are unchosen. This claim it utilized to argue, for instance, that Mele’s Diana, who can shed all her values, is incoherent: rather, an agent who exercises control must have a deliberative standpoint. But it is not clear that it also entails that an Adam who can choose his own values is also incoherent. An argument for this claim might be constructed along the lines of a Williams’ style argument against doxastic voluntarism. Just as a belief cannot be voluntarily acquired because voluntary acquisition is incompatible with the aim of belief, so we might think that holding a value is inconsistent with knowing that one has the value only because one has chosen it. This would require defense of a controversial account of values. It might be responded that eternal Adam could forget that he has chosen his values; whether that line of argument would work is also unclear to me (it might require dividing Adam into effectively two agents, such that one controls the other).

Joe, what do you think of an argument that employs an agent- and time-indexed operator? Let 'Nstp' say that p and it isn't up to agent s at time t whether p. Suppose that P concerns only some temporally intrinsic state prior to t. Then, plausibly, for any human agent s, NstP.

Now, for an agent s like us but who exists at every time, at every time t there's some truth about the past that isn't up to s at t. Let P be the whole temporally intrinsic truth about some such past time. Then, it seems, if determinism is true, we can run the consequence argument for every one of s's actions.

I suppose this won't do if time is circular.

My point about weird cases is that sometimes they are SO weird that we rightly treat them as spoils to the victor.

Even if we don't do that, if we have a proof that for agents such as ourselves in worlds such as ourselves, if determinism is true, then there's no free will, that would establish an important point.

I'm inclined to agree with Randy: I think we are primarily interested in agents such as ourselves in worlds such as our worlds. It would be an interesting and important result if agents such as us, in worlds such as ours, are not free if causal determinism is true.

Aren't we essentially interested in whether WE would be free and morally responsible, if causal determinism were true in the actual world?

The comments to this entry are closed.

Books about Agency

3QD Prize 2014: Marcus Arvan