Free Will and Necessitation
June 8, 2004 — 7:35

Author: Jon Kvanvig  Category: Divine Foreknowledge  Comments: 25

Here’s a hard problem that I’ve been thinking about and making very little progress. I was reminded of it reading Hugh McCann’s reply to Rowe in the January 2001 issue of Faith&Philosophy. Rowe took a Reid line against McCann’s version of Thomism about the way in which God concurs with human beings in acting, claiming that “God leads us around by the nose.” The criticism seemed appropriate, since what God wills is logically sufficient for us doing the actions we do.
McCann’s reply? That logical sufficiency doesn’t undermine even libertarian free will, because the logical relationship between our wills and God’s will is symmetrical: God’s general concurrence is sufficient for us acting as we do, but our willing as we do is also sufficient for God’s concurring as He does. (This is a bit mysterious, but I’ll let it pass for now.) So if logical sufficiency undermines free will (here I mean by that libertarian free will), then God can’t have free will either. McCann’s resolution of the issue is that the loss of free will has to be tied to causal or nomological implications, not logical ones.
This same issue arises in the literature on freedom/foreknowledge. Bill Craig and David Hunt both claim that the mere fact that some event or state of affairs logically implies that I will mow my lawn tomorrow is irrelevant to whether I will freely mow my lawn tomorrow, even if that event or state of affairs is one strictly about the past. For both, the question of freedom here is tied to causal or nomological implications, not logical ones.

On the face of it, this is an interesting move. Libertarians differ from compatibilists and hard determinists on the question of causal determinism, so the relevance of causal and nomological implications is obvious and the relevance of logical implications less so.
On the other hand, logical fatalism can be understood as the position that there is only one logically possible world, leaving everything that happens logically necessary. Logical fatalists take it as obvious that their position threatens free will of the libertarian sort, since there is not even a weak sense in which one could have done otherwise.
Then there is Frankfurt freedom, which can be libertarian without allowing the possibility of having been able to do otherwise, so the logical fatalist position is not as compelling as it might have been apart from Frankfurt-style counterexamples.
So the question I’m interested in is this: is a logically sufficient condition for some action of mine an underminer of free will, libertarianly conceived? The only argument I can think of for saying “yes” involves appeal to the principle of alternative possibilities cited above and which Frankfurt constructs counterexamples to. Those counterexamples can be avoided by going internal: instead of saying you have to have been able to do otherwise, you can say you have to have been able to try to do otherwise (or choose to do otherwise). Yet, if these internal options are themselves actions, variations on Frankfurt’s counterexamples can still be used–once we let in possible evil demons who watch to see what we’ll do, we can let in such demons who monitor our internal states and will block any alternative choices or tryings (after all, even choices and tryings have beginnings and take time, and a smart demon can see the beginning of a choice or trying and prevent it, since the beginning of a choice or trying is not itself a choice or a trying).
So I remain puzzled. I think Frankfurt-style counterexamples are much more powerful than rebutters have realized (see, e.g., Tom Flint’s rebuttal in his book on Molinism, which takes the line I argued against in the last paragraph), just as Gettier’s original counterexamples contained the seeds of something much more powerful and overwhelming than originally thought. Yet, the best argument I know of for insisting that logically sufficient antecedents of an action rob it of libertarian freedom have to appeal to the principle of alternative possibilities. So maybe McCann, Craig, and Hunt are right: only causal or nomological antecedents are relevant to the question of libertarian freedom.

  • I think you’ve just convinced me that the response to Frankfurt I heard from John Hawthorne won’t work. I’ll have to think about it more to be sure.
    This sounds like the same problem with ordinary fatalism, which I take to be the view that the present truth of propositions about the future logically entails that the future will be a certain way. When put that way, it sounds obviously true (even if we’re free), which makes me wonder why people then think it’s problematic for human freedom. They look at the logical connection between the truth of a proposition and the reality that the proposition is about and then think there’s some causal connection between the proposition’s being true now and the reality in the future, and they think we must not be free. They deny that there could be truth about future contingent propositions, but what they should deny is that logical entailment has anything to do with freedom.
    So in short I think what you’re saying is just an extension of something I’ve always thought obvious. I’ve been talking about logical entailment between one thing and another. You’re talking about logical necessity (not just necessary given something contingent). Maybe what seems bothersome about your issue is just that difference, though. Should it be bothersome? I’m not sure.

    June 8, 2004 — 14:37
  • jon kvanvig

    The difference between this problem and the logical fatalism problem is this: any interesting account of what is strictly about the past will make “it has always been the case that I will mow my lawn tomorrow” not a claim strictly about the past. If, however, you think “God believed yesterday that I will mow my lawn tomorrow” is strictly about the past, then there is a difference between the logical and theological case. If libertarian freedom is unaffected by logical constraints, then there is no difference (and we don’t need to worry about whether God’s beliefs are strictly about the past).
    So, I think your response is exactly the right one toward logical fatalism. But theological fatalism is different, because it uses a claim which at least appears to be strictly about the past.

    June 8, 2004 — 15:01
  • A few comments:
    1) I’m attracted to what John Fischer calls the “flicker” strategy in dealing with PAP. So, once you go “internal” as Jon puts it, if there is no genuine alternate possibility with respect to the primary internal action (of the will, whatever) which gives rise to the action A which is being Frankfurtered, then A is not performed freely. But it is easy to be grilled at this point and have nothing but intuition to fall back on. So, let’s set aside PAP for now.
    2) The main intuitive thrust behind the principal argument for incompatibilism is that the laws of nature together with the world-state at some past time entails that you will do what you do (if determinism is true). So, one might argue that there is an instance where entailment is showing its importance in explaining why we lack free will.
    But, suppose that it were only causally necessary that one will perform an action given the fixity of the past and the laws of nature. So weaken the modal force of the conditional in the premise having to do with the connection between the laws and the past, and any future events. I think the argument for incompatibilism loses no force. (One still has no control over the causal necessity of (past+laws ->I perform some action A). Beta will change a bit, of course, but one might argue it loses none of its plausibility.
    On the other hand, if you weaken the conditional as above, the similarities between K and beta diminish. This may or may not be a good thing for the person giving the incompatibilist argument. Some people point to K and say “You’re confusing beta and K somehow; that’s why beta looks plausible to you.” “Others will say that beta’s similarities to K gives us reason to think it’s true.
    Anyway, it seems to me that the fact that the main argument for incompatibilism retains force with a conditional weaker than strict implication is telling with respect to the question of entailment’s relevance to free will.
    3) It seems to me that any good Molinist will take Jon’s theological claim to be a soft fact, assuming that the proposition God believes has elements to ensure the rigidity of “tomorrow.”

    June 8, 2004 — 18:27
  • jon kvanvig

    Three quick (or not so quick) points here:
    1. Matt’s right that Molinists will want to take God’s beliefs about the past as soft facts–in fact, I’m one of them! But the road is hard: especially, Fischer’s argument that there has to be something about God’s state of mind that is a hard fact about the past (assuming, as we do for the f/f problem that God is a temporal being), even if we reject the idea that the state of mind in question has the particular content that it does in a hard way. I deny this claim, but am no longer as comfortable as I once was doing so.
    2. Suppose we take the argument from laws of nature and initial conditions to imply, with nomological necessity, the entirety of the future. There is a problem here, in that classical theists will insist that the laws of nature are “oaken”: hard to break but not logically impossible to contravene (so miracles are not logically impossible). But then []p -> p is not a theorem, when the box is interpreted as nomological necessity; and even though the conclusion follows of nomological necessity from the premises, it is logically possible for it to be false while the premises are true. Yet, if necessity doesn’t logically imply actuality for the kind of necessity in question, it is less compelling in an argument for incompatibilism.
    3. So, Matt, you think that the argument with mixed necessities, nomological and accidental, still preserves the principle “from necessary premises necessary conclusions follow.” I guess I think so too, but I can’t explain why. Do you have an explanation?

    June 8, 2004 — 20:22
  • Jon,
    I think you’re right about nomological necessity. I was trying to weaken the conditional and get an argument. You could try weakening it to a counterfactual, or some cooked-up conditional which is stronger than a material conditional but weaker than strict implication. Anyway, you don’t need entailment, I don’t think–the beta-analogue with something weaker than entailment between the laws and history, and an action might be less convincing, perhaps.
    On #1, I don’t think that Fischer’s argument works because God was in a mental relation to a “soft” proposition. So the fact that God had the mental state God did istelf is a soft fact.

    June 8, 2004 — 22:10
  • Mike

    _Suppose we take the argument from laws of nature and initial conditions to imply, with nomological necessity, the entirety of the future. There is a problem here, in that classical theists will insist that the laws of nature are “oaken”: hard to break but not logically impossible to contravene (so miracles are not logically impossible). But then []p -> p is not a theorem, when the box is interpreted as nomological necessity…_
    Very puzzling. If []p does not entail p, where [] is a causal necessity operator, then we can have a case where history H and causal laws L nomologically necessitate e, H and L occur and e doesn’t occur. But how could this happen? It could happen only if our world is not causally accessible from itself. I can see how []p-> p is not a theorem in (say) deontic logics, since our world is not among the morally ideal worlds. But what could it mean to say that our world is not causally ideal? It could mean only that *our* causal laws do not exceptionlessly govern our world (as our moral laws do not exceptionlessly govern our world). But then our causal laws are not afterall laws. Bizarre causal talk.

    June 9, 2004 — 19:24
  • Mike says:
    “Very puzzling. If []p does not entail p, where [] is a causal necessity operator, then we can have a case where history H and causal laws L nomologically necessitate e, H and L occur and e doesn’t occur. But how could this happen? It could happen only if our world is not causally accessible from itself. I can see how []p-> p is not a theorem in (say) deontic logics, since our world is not among the morally ideal worlds. But what could it mean to say that our world is not causally ideal? It could mean only that *our* causal laws do not exceptionlessly govern our world (as our moral laws do not exceptionlessly govern our world). But then our causal laws are not afterall laws. Bizarre causal talk.”
    The standard line is this: all law-like statements have to have ceteris paribus qualifiers in order to have any hope of being true. This is most obviously true in the special sciences, but can be extended to the most basic sciences since no science can be responsible for “interference” from outside the system it is describing. So psychological laws can be responsible to phenomena such as bullets penetrating grey matter, and the laws of physics can’t be responsible to phenomena such as divine intervention. The law is contravened because the ceteris paribus clause doesn’t hold.
    So, the law holds, and the antecedent (minus c.p.) holds. But since the c.p. clause doesn’t hold, the conclusion is false. To say that the (fundamental) laws plus the initial conditions (nomologically)implies the entirety of the future implies, logically, that the future will obtain as long as the c.p. clause is satisfied.

    June 9, 2004 — 21:46
  • Mike

    I don’t follow you. The special sciences do not have any laws to speak of. There are no psychological laws or sociological laws, and there aren’t even biological laws, since all of these so-called laws lack universality. J.J.C. Smart made much of this (Phil. and Scientfic Realism, 1963(?)). But more recently N. Cartwright is pretty convincing that laws containing ceteris parabis clauses might be exceptionless (and so meet that particular criterion) but they are useless in explanation. She has in mind that cp clauses make explanations possible only under the idealized conditions specified in those clauses. So I guess I can’t see how the appeal to cp clauses has nonetheless emerged as a standard line.

    June 10, 2004 — 9:34
  • John Fischer

    I strongly agree with Jon Kvanvig that the “Frankfurt rebutters” don’t get anywhere really by “going internal”. It is surprising that they seem unaware of the literature on the Frankfurt-examples which goes back quite awhile. Recently I had an exchange with Hasker in which he graciously admitted that he had not understood the force of my own defense of the Frankfurt-style examples. Bill Hasker still isn’t convinced, but I think he now sees that it is not so easy to “rebut” them.
    Second, I believe that, properly understood, Molinism itself is not a defense of compatibilism about God’s omniscience and human freedom in the sense that involves alternative possibilities. Rather, it is an account of HOW God knows what He knows. As Jon points out, the Molinist will need to adopt some other stance on the compatibility problem, and typically they adopt Ockhamism. Despite what many seem to believe, Molinism is not an answer to the free will problem, and seems to be dependent on Ockhamism, which I (and others) believe has fatal problems.

    June 10, 2004 — 18:50
  • John–good to see you outside the confines of forking paths! I like your characterization of Molinism, and agree on its connection, at least historical, to Ockhamism. Part of what my post was angling toward is whether there is a way for a Molinist to abandon Ockhamism entirely by acknowledging that God’s beliefs (or states of mind) are strictly about the past, and entail their contents, but that such is irrelevant to libertarian freedom. I can’t quite bring myself to this position, but McCann seems to be quite comfortable with it (not the Molinism part, but the part about severing the link between (non-causal) necessitation and loss of freedom).
    So, I remain undecided on whether Molinism needs Ockhamism as I have assumed till recently…

    June 10, 2004 — 20:47
  • Molinism doesn’t need Ockhamism, because Molinism is consistent with compatibilism. Now whether libertarian Molinism needs Ockhamism is another matter. Even though I’m a compatibilist, I remain convinced that philosophy of time is driving the foreknowledge problem, and if you accept as unproblematic the existence of truths about future contingents then you shouldn’t have a problem with God’s knowledge of contingents as long as God is atemporal. I think you need to say both those things, though. I haven’t seen another libertarian strategy that I’m convinced will succeed.

    June 10, 2004 — 21:30
  • Jeremy–I agree with David Hunt on this point: adopting an atemporal view of God doesn’t solve the freedom/foreknowledge problem. Whether or not God is temporal, the direction of explanation should be from facts or truths to knowledge (and hence foreknowledge), not the other way around. So there shouldn’t be a good argument from foreknowledge to a loss of freedom, whether or not there is a being that has complete foreknowledge. Even if God is atemporal, the philosophical problem remains.
    The intuition I cited is meant to imply that we couldn’t find out that freedom and foreknowledge are in fact incompatible. The intuition is meant only to show that no one should facilely assume incompatibilism. The reason Molinism-cum-Ockhamism is something to be excited about is that it claims to be able to preserve complete sovreignty together with a libertarian account of freedom.

    June 10, 2004 — 21:52
  • John Fischer

    Jon: I agree that strictly speaking Molinism does not need to be accompanied by Ockhamism. But of course it needs something to address the compatibility problem, or else it is simply a way of understanding how God knows what He knows.
    Also, and I don’t want to be presumptuous or obnoxious, for any of your readers interested in overviews of some of the literature on Frankfurt-type examples, I have the temerity to suggest two of my pieces: “Recent Work on Moral Responsibility,” *Ethics* Oct. 1999, pp. 93-139; and “Responsibility and Alternative Possibilities”, in the Widerker/McKenna anthology published by Ashgate Press.
    It is a pleasure to see your blog developing. We have a link to it and a post about it at “Forking Paths”. I hope you and your readers will join us too!

    June 11, 2004 — 1:47
  • Right. I didn’t say atemporality solves the problem. You need to say what you just said about causation (though that point is so obvious to me that it amazes me how anyone denies it). I remember John Hawthorne convincing me that even that won’t solve the problem if God is in time, but I can’t remember his argument. Maybe I’d disagree with it now.

    June 11, 2004 — 7:23
  • John Fischer

    By the way, Jeremy and Jon: David Widerker has a number of published pieces in which he argues that atemporality doesn’t solve the free will problem–that in fact it leads to similar problems about the fixity of the past.

    June 11, 2004 — 8:02
  • John Fischer

    Jeremy says, “Molinism doesn’t need Ockhamism, because Molinism is consistent with compatibilism.” Hmm… This is a bit confusing to me, since Molinism might be “consistent with compatibilism” only if it assumes something like Ockhamism!
    In my view, Ockahmism is fatally flawed, and (as David Widerker has shown) the atemporalist solution is not much more attractive.

    June 11, 2004 — 8:49
  • John,
    I haven’t kept up on your objections to Ockhamism for seven or so years now. Is your main concern still the one that Jon mentioned–the problem about God’s having a particular belief state being a hard fact?

    June 11, 2004 — 12:02
  • John,
    If Molinism is roughly the view that God knows counterfactuals whose consequents involve the actions of creatures who act in the sort of way a libertarian would deem sufficient and necessary for free actions, it still could be consistent with compatibilism. It would be a strange view, but one could hold that God knows these sorts of counterfactuals, but compatibilism still is true. One might think that being able to do otherwise (together with the other required libertarian sort of stuff) was sufficient but not necessary for acting freely. (There then are questions, of course, as to why God would actualize a world that didn’t have free creatures who could be causally determined.)
    Or suppose that some compatibilist analysis of free will is correct, and what the libertarian provides as being sufficient for acting freely isn’t. If we consider a world with a bunch of compatibilist-free creatures there still might be all sorts of true counterfactuals (about these beings, and other “possible” beings) of the following sort:
    If (S satisfied libertarian sufficient conditions for freedom in C and) it were the case that S were in circumstances C, then S would perform A (where A is an action a libertarian would deem free).
    Or, if the claim is that Molinism is compatible with God’s having foreknowledge and our acting freely, there is the Molinist move of denying that accidental necessity is closed under entailment (just mentioning, not endorsing this).

    June 11, 2004 — 13:11
  • Molinism, as I understand it, is just that God knows what free beings would do in different circumstances. Compatibilists believe we have freedom, and compatibilists believe there are truths about what people would do if circumstances were different. Why would it be a problem to add that God knows those truths?
    As for Ockhamism, isn’t that just the soft fact/hard fact distinction? Is it the distinction that you think is fatally flawed or its use to solve the foreknowledge problem?

    June 11, 2004 — 14:25
  • jon kvanvig

    Molinism is the view that God’s knowledge occurs in three sequential logical moments: first, he has natural knowledge (which is knowledge of necessary truths, next he has middle knowledge, which is knowledge of what would occur under various conditions, including what free individuals would do under those conditions, and last, he has free knowledge, which is knowledge of contingencies that result from what he has created (this is a bit sloppy, but I won’t bother to fix it here).
    As such, Molinism disputes any view according to which Go has to know contingent facts about the actual world in order to know what a free individual would do in various circumstances, since God uses this middle knowledge in order deciding what kind of a world to create.
    In light of this, compatibilism would be explanatorily superfluous for Molinists. God doesn’t need to know either the laws of nature or the initial conditions of the universe in order to know what a person would freely do in a set of circumstances.

    June 11, 2004 — 14:40
  • Jon,
    I’m not sure if that was directed to me or not. My only claim was that the two were compatible. The event in the consequent of the contingent prevolitionally-known counterfactuals need not consist in a free action. I could be anything indeterministic. So, you might think that no plausible theory of agency can be made out for a libertarian, and thus you might opt for some sort of compatibilist analysis. But you still might think that God knows some contingent counterfactuals prevolitionally because there are true propositions of this sort. You’re right, a main motivation for going with Molinism is to account for providence with libertarian-free creatures. I just was trying to examine a view in logical space.
    I take it the sorts of deliberative counterfactuals God would look at prevoitionally if one denies middle knowledge and embraces compatibilism would look similar to counterfactuals of freedom. (There is, as you know, some debate on how richly one is to take the antecedents of CFFs.) They’d be something like, “If I actualize a world with initial conditions C, and it is deterministic with laws L, then if S were in C*, S would freely (in a compatibilist sense) perform A.” This is know via his natural knowledge; no free knowledge required here. There’s no foreknowledge here, either, of course.

    June 11, 2004 — 16:19
  • jon kvanvig

    Matt–I agree with what you’ve said, and wasn’t trying to undermine it. It is an interesting proposal to think of middle knowledge as independent of libertarianism. I don’t think it helps the Molinist much (though this is a completely different matter from your remarks). In particular, if the grounding objection undermines Molinism with libertarian freedom in the counterfactuals, it will undermine Molinism-cum-compatibilist counterfactuals. Unless, of course, the counterfactuals are logically necessary…

    June 11, 2004 — 19:31
  • Al Plantinga’s position on the grounding objection is that it’s not a big deal if there is no ground–as he might say, the objection isn’t very impressive; and the last time I talked to Tom Flint about this and we hashed things out, he told me that had he to do things over again, he’d say that CCFs are ungrounded. (The argument I pushed was a relatively simple one–very roughly it was this: He wants to ground the truth of the counterfactuals in relations between worlds (in the Lewis semantics). But if you take libertarian freedom seriously, you think something to the effect that for any free action A, there is a world just like the actual world up to A (add caveats about future contingents, non-transient propositions, etc.), but in which I don’t perform A. What happens after the action shouldn’t count for evaluation of similarity relations in giving truth conditions for the counterfactual. So, for any free action, you’ll have a “tie;” there will be two worlds, equally close to the actual, no other ones where the antecedent holds and is closer to the actual world, where the consequent holds in one and doesn’t hold in the other (and the antecedent holds in both). So, similarity relations can’t ground the things. So, Flint said (this was a couple years ago), that he thought it was better to think of them as ungrounded. But he may have changed his mind, and the argument I gave him was much more formal than this sketch here.

    June 11, 2004 — 21:42
  • Jon,
    I had missed a sentence in what you said. Yes, I had in mind that the conditional I mentioned is a necessary truth–so known via God’s natural knowledge (and we’re not Molinists at all, at least not so far).

    June 11, 2004 — 21:54
  • jon kvanvig

    Matt–the sentence about the counterfactuals being necessary was meant just for trying to sustain Molinism, i.e., to preserve knowledge of the counterfactuals as counting as middle knowledge.
    I’m inclined, too, to say that ccf’s are ungrounded, and I think your version of the grounding objection in terms of the standard semantics is exactly the right one to push. After all, the point of a semantical theory is to tell us in what the truth of a sentence consists…

    June 12, 2004 — 10:51