Blog Coordinator

« Implications of the Experimental Approach to Ethics | Main | Views on free will: some unsurprising results »

11/05/2010

Comments

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

Thanks for your post, Joe.

I guess I continue to find attractive the idea, first formulated in roughly this way by Carl Ginet, that our freedom is the power to add to the given past, holding the laws of nature fixed. It is hard to prove this idea, but it seems quite plausible to me, and rejecting it appears to have implausible implications for practical reasoning. As Joe points out, one can construct an argument for the incompatibility of causal determinism and freedom to do otherwise employing this basic idea. This lies behind a formulation of the incompatibilist's argument by Carl Ginet; it also is the basis for Van Inwagen's possible-worlds version of the Consequence Argument, and my "Basic Argument".

Joe claims to produce a "reductio" of this pattern of arguentation, but I don't quite see it. Perhaps I am missing something. In any case, I'll add some roman numerals to name a few moves in Joe's crucial paragraph:

"[I]However, it is a consequence of this argument, if it is indeed an argument for incompatibilism, that any proposition that is true in all accessible worlds is such that no agent is able to render it false. If some agent were able to render At' false, for instance, then the above argument would not be a proof of incompatibilism. Reflect now on the proposition Pt.[II} It too is true in all accessible worlds. Thus, no agent is able to render Pt false. (Consider: If Pt were not true in all accessible worlds, then it would be impossible to get the result that At' is true in all accessible worlds; same holds if someone were able to render Pt false.) Yet Pt is just a proposition about the state of the world at some arbitrary time t. Thus, it is a consequence of the soundness of van Inwagen’s 2nd argument, that no person is ever able to render any proposition about the state of the world at some arbitrary time t false. And this result follows independently of the thesis of determinism."

First, it seems to me that [I] is not a consequence of the argument, but simply an independent (but, of course, relevant) point.

Second, Joe says that [II] and the associated result "follows independently of the thesis of determinism". But I don't see this; doesn't the truth of [II] depend precisely on the assumption of determinism? How else would one get [II}? This assumption appeared to be invoked in the argumentation to which Joe implicitly alludes in the parenthetical remarks above.)

As I said above, no doubt I am missing something, and Joe will enlighten me.

I should perhaps add that, although I find this version of the Consequence Argument plausible, it is no part of my doctrine of Semicompatibilism, which has it that our moral responsibility can be prescinded from traditional questions, such as the relationship between (say) causal determinism and freedom to do otherwise or (say) God's foreknowledge and human freedom to do otherwise.

I'm also having trouble seeing precisely what the argument is. Isn't the proposition you are selecting, Pt, one that is picked out as holding across all accessible worlds? In that case, it is trivially true that it is true across all accessible worlds, but hardly surprising. What you seem to have done is to rigidify Pt. It is of course true that any rigidified proposition is true across all possible worlds - that's what rigidification buys us - but this is all entirely independent of the arguments here. Why should it be a problem for the 2nd argument that it takes an arbitrary proposition and rigidifies it?

"Of course, any world that shares the past and the laws of nature of the actual world is one that shares the future of the actual world, as well, provided that the actual world is deterministic."

Hi Joe,

There is a trivial sense in which this is true, where we include in the set of true proposition in the past of W every proposition, including every true future contingent proposition in the 'past' of W about the future of W. But it's likely a mistake to take every proposition that happens to be true prior to some time t in W to be a true proposition 'in' the past of W. So supposing we can restrict the relevant propositions in the past of W to those that are actually 'about the past' of W, then we won't have a worry about two worlds W and W' sharing the same past and different futures. I take it that this is all pretty obvious and well-known. So what am I missing?

Hi Joe.

I see an argument in the vicinity in which Pt is indeed true in all accessible worlds (although I don't know if this is what you had in mind). However, it doesn't prove anything so strong.

The argument assumes that ability and accessibility are both indexed to a temporal parameter t.

(A) an agent k is able-at-t to perform an action A at a time t', only if there is a t-accessible world w in which At' is true [PREMISS]

(B) For every world w and w' and time t, w' is t-accessible to w if, and only if, for every t' earlier than or identical with t, the total state of the world at t' is the same in w and w' (and the same laws of nature hold in both worlds) [PREMISS]

(C) Pt is the total state of the world at time t [PREMISS]

(D) Every t-accessible world w is such that Pt is true at w [from B and C]

(E) no agent k is able-at-t to render Pt false [from A and D]

If we accept the temporally indexed notions of ability and accessibility, the argument is sound and Pt is indeed true in all (t-)accessible worlds. However, what the argument proves is not fatalism, but only the necessity of the present (see footnote 3 of my 2009 Analysis paper)

For instance, From (E) it doesn't follow that there is--say--no time t' earlier than t such that some agent is/was able at t' to render Pt false.

Roberto: I think you're wrong.

In order to get the conclusion of incompatibilism, you need to generalize on the arbitrary case noted in (E). But that follows only if you generalize on the arbitrary case noted in (C). (See the post after this one.) But that is a proposition about the state of the world at some arbitrary time. I don't see how you can limit the claim in (C) to the present, for it is in fact true for any arbitrary t. If we generalize about this, we get fatalism: all propositions about the state of the world at any particular time are (relatively) necessary, that is, necessary in the sense that the proposition noted in (E) must be necessary in order to reach the incompatibilist conclusion.

Maybe my subsequent posts will help.

I have a few comments to make to John in a moment but let me make a general point about modal proofs and the thesis of determinism that should answer some of the concerns noted by him and the others.

Determinism is a conditional thesis. In the context of my version of the 2nd Argument it comes down to this:

(Pt & L) entails At'

To simplify things, I could have stressed the fact that we're thinking only of worlds that are close to the actual world, accessible to the actual world given certain specifications, and that in those worlds, L holds. That would give us:

Determinism: Pt entails At'.

From this thesis alone, nothing follows about the (relative) necessity of any proposition. If it did, then one could prove At' from determinism alone. You wouldn't need a proposition like Pt to ground the subsequent necessity. Clearly, the assumption of Pt is playing an essential role in the argument otherwise you could generate the argument without it.

It should be equally clear, to anyone who has done modal logic at least, that whatever the modal property that At' is supposed to have, it cannot get it unless Pt has that same modal property. If the modal property of Pt is "trivial," then the subsequent modal property of At' is equally trivial.

Pt entails At'
Pt is necessary in some trivial way that is irrelevant to the free will thesis.
Therefore, At' is necessary in the same trivial way.

I'm fine with this result. But if the modal property had by At' is substantive and worrisome, you are going to have to explain to me how At' could get it unless Pt had an equally substantive and worrisome modal property. Otherwise, we’re doing magic, not logic.

Nor can you suggest that the modal property had by Pt follows from the thesis of determinism alone, for then you would be able to prove incompatibilism more directly. If that is the case, just show how Pt follows from determinism and you're done. But you can't do it.

In summary, determinism is merely a conditional thesis. Garbage in, garbage out; substance it, substance out. I don't care which you choose. At most, determinism transfers the modal properties of some propositions onto others. If the necessity of Pt is trivial, then so is the necessity of At' and the 2nd argument is not a proof of incompatibilism. If Pt lacks the modal properties I claim that it has, then so does At'. If At' has those properties, then so must Pt. Garbage in, substance out is not an option. You simply cannot get a modally substantive result from a modally trivial claim together with a conditional thesis.

I should have thanked John, Roberto, and the others for their thoughtful comments. Sorry for not doing so!

John,

Hopefully the comments above will make the argument clearer.

In addition, I think (I) does follow from the 2nd argument, at least the version that I presented above. You only need to add acceptable definitions of "accessible," "render," etc. At least I can't see why one wouldn't accept (I) given those definitions. Maybe I'm missing the point here.

Keep in mind that I haven't looked closely at your basic argument in a while and I'm sure the point would play out differently there, if indeed it can be made. I'm really just talking about this rather quick, informal version of the 2nd argument.

As I noted in my comments of above, if Pt follows from determinism, you should be able to produce that result directly. But then you wouldn't need to prove that At' follows from Pt and determinism since Pt, given that t is some arbitrary time, would give you the needed result.

Clearly incompatibilism is a modal thesis. Maybe you could explain to me, without resorting to the basic argument or the 2nd argument, how a robust modal thesis like this could possibly follow from a non-modal assumption about the actual world together with a conditional thesis like determinism. How is such a result even remotely possible?

I can see how one reaches that result in the 1st argument and the 3rd argument, since in those arguments there are assumptions about the necessity of the past and the laws. But of course the past does not seem to be essentially necessary in the way that those arguments suppose (this is the result of my 2007 Analysis paper). The question is: How could you produce the result without making assumptions about the necessity of the past?

Joe, I'm confused (I warn you), and it's probably my fault. I know you know a lot more about these issues than me. So help me. In the post, you claimed that it was a reductio of the 2nd argument that it entailed that no agent can render Pt false. I think that it is true, for any arbitrary Pt, that no agent can render it false, but this is because for the argument rigidifies that proposition. Your clarification focuses on the transfer of modal properties to At, but the original claim was about Pt. It is a further and independent question  - on which I take no stand - whether an agent's  being unable to render Pt false entails that At is necessarily true in all worlds, or necessary in the same sense as Pt. 

Neil,

Thanks for pressing me further. I know that it sounds as if I'm more confident about the point than I am. Part of that is because I'm slammed with other work and I'm trying to be quick! I’ve only been thinking about the 2nd argument for a week. And I need to make the point more formally, but I won't be able to do that for a while. I thought in the meantime I'd post it here and see what kind of feedback I get. That you, and John, and Roberto, and others are not convinced worries me!

Let me start with a question: How is it that At' is shown to be true in all (close) worlds if it isn't also the case that Pt is assumed to be true in all (close) worlds? That is the question behind my point. Thus, I don't see it as something I'm doing -- rigidifying Pt, for instance. I'm merely observing that the result of incompatibilism cannot be reached unless Pt is in fact true in every possible (accessible) world. Another way to make the point, perhaps, is that this is just the assumption of global fatalism and it isn’t surprising that incompatibilism follows from this assumption.

Think of it on a case-by-case basis. You start with one (close) world (where L is true) and from the truth of Pt you deduce At'. Then you move to another (close) world and from the truth of Pt you deduce At' in that second world, as well. At the end, you get incompatibilism: At' is true in all (close) worlds. But I don't see how the result follows UNLESS Pt is assumed to be true in all (close) worlds, as well. What besides the assumptions of determinism and Pt and L is doing the work? Nor can you get the result without the continual assumption of Pt. If there is something else, I don’t see it.

And if the property of being true in all (close) worlds is what is yielding the unfortunate modal property that At' seems to have, the one that makes us think incompatibilism is true, it must be the case that Pt has that unfortunate modal property, too. The proponent of the 2nd argument never explicitly says this but how could it be otherwise?

Joe, if van Inwagen's 2nd argument really allows that Pt can be about any time whatsoever, then I, like you, am puzzled about how the argument can support incompatibilism. However, I'm rather doubtful that the argument does in fact allow this. If it did, how would it be an expression of the idea that our freedom is a freedom to add to the actual past?

I looked briefly at the relevant section of vI's book. Though perhaps I didn't look carefully enough, it seems to me as follows.

A key notion of the 2nd argument is that of worlds that share a slice with each other (are indistinguishable at some instant). The MAIN IDEA is, roughly, that we have access to a world only if it shares the laws and at least one slice with ours. Since, strictly speaking, the MAIN IDEA doesn't specify that the shared slice must lie in the past, the argument doesn't strictly express the idea that our freedom is a freedom to add to the actual past.

That point aside, determinism is so defined that, if our world is deterministic, then any world that shares our laws and a slice of our world is our world. Add the premise that our world is deterministic to the MAIN IDEA, and you get the conclusion that we have access only to the actual world. But how do you get that conclusion just from the MAIN IDEA, without the premise asserting that our world is deterministic?

Two great posts, Randy. Let me deal with the second one first.

Keep in mind that, as I've admitted, I haven't had time to look at van Inwagen's version of the 2nd argument as carefully as I should in order to work out an adequate response. Thus, my version of the 2nd argument is based on versions offered to me by Ish and Andrew, neither of which is published yet. I think you're correct that van Inwagen does not need the concept of the "freedom to add to the actual past" in order to run the argument. Even in my informal version, the "t" in Pt could be any time at all: past, present, or future. It is interesting that van Inwagen does say, about one of his principles: "MAA may be regarded as a statement of the familiar principle that no one can change the past" (1983, 92). Ish and Andrew claim that you can run the argument without making assumptions about the remote past and you're noting here that you can do it with talking about the past at all, and I think you're all correct. But that's also why it is hard to formulate my point. It might be formulated differently depending on which version of the 2nd argument we're talking about.

Second, I agree that the way I put it initially -- "one can prove that no one is ever able to do otherwise without the assumption of determinism" -- is not the correct way to make the point. I like the way I formulate the point in the step-by-step version I gave in response to Neil's concerns. That said, if you start with a specification of the state of the actual world at some arbitrary time t (Pt), then at every (close) possible world you consider, Pt will be true. And this is the case whether or not determinism is true. That is the sense in which the result that Pt has the exact same modal property that At' has is independent of the truth of determinism. It is built into the very procedure that the argument uses. This is not my doing, it is an essential part of the 2nd argument.

But I don't really need to say anything about the status of Pt independent of its use in the 2nd argument for incompatibilism. It is still the case that at the end of that proof, once you show that At' is true in all (close) possible worlds and try to use that as a basis for incompatibilism, Pt is going to have exactly the same modal property that At' has. If the modal property is a trivial one, then the result is trivial and it is not really a proof of incompatibilism. If the modal property that At' has is non-trivial, one that will support incompatibilism, then Pt must have had that property, too. Perhaps the point is not that you've now got a proof of global fatalism but, rather, you've tacitly assumed global fatalism in order to get the incompatibilism conclusion. No surprise there! Either way, the argument is far less than it appears to be and should not be regarded as a proof of incompatibilism.

I still want to say something about the "freedom to add to the actual past," since even if it is not essential to the formulation of the 2nd argument, one might still derive a version of that argument (or Fischer's basic argument or Ginet's argument) in an attempt to flesh out the consequences of thinking of our freedom in this way. Ish, Andrew, and Fischer all seem to have this in mind, for instance.

Remember that I'm a classical compatibilist. I think that S has free will iff S has or had the ability to do otherwise AND that free will is compatible with determinism. I also share with Fischer the belief that the relevant kind of freedom is the freedom to add to the actual past BUT my understanding of what that amounts to is different than his understanding. In short, I do not think that we should understand the freedom to add to the actual past as the freedom to do otherwise holding fixed the past and the laws. It is just the freedom to do otherwise given the past and the laws, the freedom to do otherwise now, the power to do otherwise. My comments about the 2nd argument are one way of illustrating the problems associated with understanding the concept in the “strong” way (as John Perry puts it) that Fischer understands it. Another problem is the Mind argument, or van Inwagen's "rollback" argument, which suggests that this type of freedom is mere randomness. Here is a third reason for thinking that the strong way of understanding the freedom to add to the actual past is the wrong way to understand it.

I have a friend with Tourette's syndrome. Let's contrast his shouting "F**k" on a given occasion and my doing so. Intuitively, my performance is something that I freely do, something for which I might be blameworthy (suppose I said it with children around and was rightly scolded by a parent) whereas his is not. His behavior is uncontrollable, not something over which he has control. On a given occasion when I said "F**k," I had the freedom to add to the actual past. I was able to say it and able to refrain from saying it. He lacked this very freedom. This is just to say that Tourette’s syndrome is a kind of compulsive disorder.

It seems to me to be a mistake to try to understand the freedom at issue in this way, that what I had and he lacked was the power to say "F**k" holding fixed the past and the laws. It could very well be that the way Tourette's syndrome works is nondeterministic. We might subject my friend to a rollback experiment, placing him back in the very situation he was in prior to yelling “F**k,” holding the past and laws fixed. Were we to let him proceed further he might fail to yell “F**k” a second time, he might yell it again a third, fail again a fourth and fifth time, etc. And it might be that, if we were to rollback time on a particular occasion of my utterance of "F**k," I might say it over and over again, each time we rolled time back, precisely because, given the situation, given my level of anger or joy, given that I grew up in New Jersey, etc. it is a natural consequence that I say it. And this might be true of me even if determinism is false and my saying it is not an inevitable consequence of the past and the laws. Maybe I only fail to say it one in a million times, given the specifications of the situation.

Since I don't see the rollback method as an effective way of distinguishing my utterances of "F**k" from my friend's, I don't think that the strong view of the freedom to add to the actual past is the correct one. And I can say this with confidence even though I do not have a better theory to put in its place (which I don't). I don’t need the right way to understand the freedom to add to the actual past in order to see that this is the wrong way to think about it.

(I use the problem of Tourette's syndrome not because I like repeating “F**k” but because it is VERY possible that this particular compulsive disorder might have a nondeterministic and even random manifestation, unlike say kleptomania. And because I say “F**k” a lot yet I don’t have a compulsive disorder. And yes I have a friend who does have this compulsive disorder.)

Joe,

This is interesting stuff; thanks for posting.

You've hinted at a worry as to whether the 2nd CA is indeed an argument for incompatibilism. I'm a little puzzled here, though, for isn't your reply supposed to show that the 2nd CA is an argument for fatalism? And doesn't fatalism (necessarily, no one is free) imply incompatibilism (necessarily, if determinism is true, no one is free)?

I understand your argument as the following (Let Pt express the complete state of the world at an arbitrary time t):

1. For any x, someone is free with respect to x only if there's an accessible ~x world.
2. Pt is true at every accessible world.
3. Therefore, there are no accessible ~Pt worlds (from 2).
4. Therefore, no one is free with respect to Pt (from 1 and 3).
5. Since t was chosen arbitrarily: for any t, no one is free with respect to the state of the world at t (from 4).

And if all this is right, then (1) and (2) jointly imply fatalism. Since (1) and (2) just are the premises of the 2nd CA, the 2nd CA is an argument for fatalism. Before I offer a reply, I want to make sure I've got your point right here. Does the above capture your argument?

Thanks for the comments and questions, Andrew!

I'm split about which way to go. Originally, I offered the criticism along the lines that you suggest above. But I have doubts that this is the best way to make the point. Suppose I agree that it is? Why won't that work?

In later responses to Neil and Randy, I suggested that the point is better understood as saying that there is a tacit presupposition of global fatalism, and since global fatalism entails incompatibilism it is not too surprising that the conclusion follows. Nor too worrisome. The reason is that as arguments for global fatalism goes, this argument is not so good, for on this understanding global fatalism is a presupposition of the argument.

Likely the presupposition is stronger than mere global fatalism (no one can ever do otherwise). Maybe it is what Kadri Vihvelin calls "impossibilism" (the ability to do otherwise is impossible).

A point here about fatalism. Unavoidability is not equivalent to fatalism. Causality, for example, is a form of conditional unavoidability, but causality does not entail fatalism. That is because unavoidability is a relative term, and requires context for claims of how something is not avoidable. Fatalism is a particular form of unavoidability, and involves either foreknowledge of an unavoidable event and/or control of such an event. I'd say that Andrew's (1) and (2) above might be compatible with unspecified unavoidability of Pt, but in the absence of any account of how Pt might be foreknown or controlled in some way, I can't fathom why the word "fate" applies here. I see Joe as simply asking if certain assumptions about fixed conditions across close possible worlds in a CA-style argument are not just question-begging in supposedly implying conclusions about freedom.

Joe,

I think I have a clearer idea of what's going on now. I'll quickly state how I think the 2nd CA goes and then comment on your first reply. Then I'll take up your second reply strategy.

"The facts", let's say, are those things over which we have no control; they are true in all the worlds to which we have access (all the worlds we have the power to make actual). We have the power to make actual only those worlds that share certain minimal features in common with the actual world. The 2nd CA says that the laws are among the facts. And so is the state of the world at some time or other (whether past, present, or future). We have the power to make actual only those worlds that share at least these two minimal features with the actual world: being exactly like the actual world with respect to the laws, and being exactly like the actual world with respect to some time or other ("sharing a time slice" with the actual world).

The premises of the 2nd CA are logically weaker than those of the 3rd. The 3rd requires that the *past* makes its way into the facts (and this is how Campbell-style objections get traction; they show that the existence of a past is contingent). The 2nd CA requires only that *some* time or other (whether past, present, or future) makes its way into the facts. The 2nd CA appeals, then, not to the fixity of the past, but to the fixity of some time or other.

Now we can see, I think, where your first reply to the 2nd CA goes wrong. It's not an assumption of the 2nd CA that the state of the world at *every* time makes its way into the facts (that would be fatalism!). For *some* time t, Pt is true at all the accessible worlds. So for *some* time t, we are unfree with respect to the state of the world at t (for all the 2nd CA says). The generalization step of your argument (as I reconstructed it) does not follow from the assumptions made by the 2nd CA, since t was *not* chosen arbitrarily (t was chosen, rather, for this peculiar feature: it is among the facts).

I think your other strategy of reply goes wrong in similar ways. Your claim is that the premises of the 2nd CA presuppose fatalism. Fatalism is the thesis that necessarily, no one is free (or perhaps just that no one is free, if you want to distinguish fatalism from impossibilism). But the thesis that for *some* time or other, no one has any choice over the state of the world at at that time isn't fatalism; for it's compatibile with that thesis that someone has a choice over the states of the world at *other* times. Since this assumption of the 2nd CA is compatible with the denial of fatalism, it does not imply fatalism (and so, I suppose, does not presuppose it).

Distinguish:

1. There's some slice s such that every world accessible to any of us shares s with our world.

2. For every world that is accessible to any of us, there is some slice s that that world shares with our world.

(2) is a premise of the 2nd argument. (1) isn't.

Andrew,

you say: "It's not an assumption of the 2nd CA that the state of the world at *every* time makes its way into the facts (that would be fatalism!). For *some* time t, Pt is true at all the accessible worlds."

I'd put things differently. I'd say that the two main premises of the 2nd CA are

CA1) the actual LAWS are true in every accessible world
CA2) every accessible world w is such that, for some time t, the t-slice in w is shared with the actual world

From CA1 and CA2 surely doesn't follow that for *every* time t Pt is true in all worlds; but it also doesn't follow that for *some* time t, Pt is true in all worlds.

The main problem with Joe's argument is, therefore, understanding why (chosen an arbitrary time t and its complete description Pt) Pt should be true in all worlds.

If we presuppose determinism, then Pt holds indeed in all worlds, but then there is seemingly nothing problematic with the 2nd CA as an argument for incompatibilism.

If we don't presuppose determinism, then we need some additional ground to include Pt in the set of "the facts" (as you call them).
My argument (A)-(E) above was meant to provide an answer to this question: if we index the notions of ability and accessibility to a temporal parameter as I suggested, then it follows that, in order to be able *at t* to perform a certain action A (at some time t'), the worlds that matter are those who are t-accessible and (given my definition of t-accessibility above) in all those worlds Pt is true. However, my point was that if we generalize the result we obtain only that:

(G) For every time t, no one is able-at-t to make Pt false

which is what I call the principle of the necessity of the present (in the sense that from the standpoint of any time t we have no power with respect to what happens at t itself [the "present"] but only--perhaps--with respect to what happens at later times [the "future"]), and the necessity of the present alone is insufficient to derive fatalism.

It seems to me, therefore, that pressed with the question "why should Pt be true in all worlds?" Joe is faced with a dilemma: either he is presupposing determinism (but then the 2nd CA is OK) or he doesn't (but then either we have no reason to hold the truth-value of Pt fixed across worlds, or (for those who endorse my argument) what we can conclude is--at best--only the "necessity of the present")

An important distinction, Randy. What I say above suggests that I think your (1) is a premise of the 2nd CA, which is a mistake. While I think an argument taking (1) as a premise is plausible, your (2) is intended here.

P.S.: Sorry! R. Clarke's comment appeared right after I submitted mine...

I have a bit of time this morning to try to respond to some of the new, interesting, and complex posts. I doubt I'll be able to get to all of them but I'll try. My schedule gets considerably lighter next week.

I want to begin with Andrew's long post and just make a quick point about his opening comments. He writes:

"'The facts', let's say, are those things over which we have no control; they are true in all the worlds to which we have access (all the worlds we have the power to make actual). We have the power to make actual only those worlds that share certain minimal features in common with the actual world."

I disagree with these comments in their entirety. Part of the point of the Tourette's syndrome post was to convey some of the reasons why I think this way of thinking is incorrect.

Consider another example. We do not cease to call a man who plays piano for a living a "pianist" merely because he rides a train and the train has no piano. But what is a pianist other than a person who has control, ability, and power over a piano that I lack? Standard usage suggests that we simply do not use all ability terms in the way suggested by Andrew's passage.

One has to first do a bit of philosophy before he can become convinced that Andrew's way of speaking is the correct way to speak about the matter. But arguments like the 2nd argument, or the other formal versions of the consequence argument, are precisely the philosophy one must do to reach this conclusion. How on earth can one appeal to that way of speaking as a kind of intuitive support to get the 2nd argument going? This is a clear case of putting the cart before the horse.

Again, I have no problem with the phrase "the power to add to the actual past." I only have a problem with the claim that we MUST interpret "power" in the way that Andrew, John, etc. say we must (that is, in what Perry would call the "strong" sense of the term). Of course, if I were convinced that this way of speaking were the correct way of speaking, independent of considerations of the 2nd argument, etc., then I would think that the 2nd argument was sound. I can see how things go from that point forward. How can you get me to that point without presupposing the argument itself?

Another comment on Andrew's longer post. (I have not teased out the distinction that Randy notes and that might play some role in the matter but there are some other issues worth noting first.)

Andrew writes: "Now we can see, I think, where your first reply to the 2nd CA goes wrong. It's not an assumption of the 2nd CA that the state of the world at *every* time makes its way into the facts (that would be fatalism!). For *some* time t, Pt is true at all the accessible worlds. So for *some* time t, we are unfree with respect to the state of the world at t (for all the 2nd CA says). The generalization step of your argument (as I reconstructed it) does not follow from the assumptions made by the 2nd CA, since t was *not* chosen arbitrarily (t was chosen, rather, for this peculiar feature: it is among the facts)."

What I meant by saying that t is "arbitrary" is not that any old t will do. Certainly Pt must be "among the facts." The point is that t can be any old time among the times of the actual world, not necessarily past or present or future. This itself is an interesting presupposition since it presupposes that there are future facts. If there are no future facts, then in fact t cannot be any old time. It must be past or present and the argument starts to look more like the 1st or 3rd argument. In any event, if Pt is not arbitrary in this sense, if t in fact is a past fact, then the 2nd argument is VERY much like the 1st or 3rd argument, so much like it that I can use the Analysis trick on the 2nd argument since it in fact does, or would have to, presuppose that there is a remote past.

Let me make one final comment on the phrase "we are unfree with respect to the state of the world at t." This strikes me as a bizarre claim, even if t is a past or present time. Among the facts about the present time is the fact that I am married. Do I wish to say that I am "unfree with respect to" this fact? Well, I could change it, though I don't wish to do so. Further, it isn't as if the fact were thrust upon me, against my will. It is a fact that I am now married because of something I did previously, namely, I got married. And that was something that I was not "unfree with respect to." In short, what strikes me as relevant about the state of the world NOW is not that I can't NOW do anything about it. What strikes me as relevant about the state of the world NOW is that I was free at various times in the past to do things differently, to help make the world the way that it is now or to help make it some different way than it is now. In short, in 1991 (when I did get married) I could have not gotten married. If I didn't get married in 1991, then the state of the world now would be very different. Again, what is important is NOT that I can't NOW do anything about the state of the world NOW but whether I could have done something BEFORE about the state of the world NOW. That is the relevant issue wrt my freedom.

Given the way that Andrew is describing the 2nd argument NOW, I'm not at all sure that it doesn't presuppose the necessity of the past, or more importantly that it does not presuppose the necessity of the remote past. It isn't clear to me that my Oscillating Adam example won't work to defeat it. In which case, even if it did prove something the something it proved wouldn't be incompatibilism. In any event, I'm convinced that part of the allure of the 2nd argument is that it is a moving target.

Thanks, Roberto.

I've actually addressed all of the concerns you've restated in your new post. But you seem to have missed them, so let me restate some of them with the new post in mind.

First, I'm talking about MY (version of the 2nd) argument, not YOUR argument, which I might respond to in a different way (though I am NOT discussing at the moment). In MY argument, t is an arbitrary time. That t is arbitrary is an essential feature of the 2nd argument, as I see it. For if t is a past time, then the 2nd argument depends on a premise about the remote past and I've already dismissed arguments of this type.

Forget for the moment that we disagree about that. It is irrelevant to MY argument. If you want to say that MY argument is not a version of the 2nd argument that is fine. But that isn't what you're saying. Given your dismissal of the 2nd argument you seem to agree with me that it is problematic. That is the entire point of my post.

Consider now the main principles of MY argument, as you've formulated them:

CA1) the actual LAWS are true in every accessible world


CA2) every accessible world w is such that, for some time t, the t-slice in w is shared with the actual world

You then write: "The main problem with Joe's argument is, therefore, understanding why (chosen an arbitrary time t and its complete description Pt) Pt should be true in all worlds."

But there is a VERY good reason for this, and I have said it in many different ways. Unless Pt is true in ALL (close) possible worlds, one cannot derive At' in ALL (close) possible worlds. And without the latter derivation, there simply is no proof of incompatibilism. If you can derive At', or some similar proposition, from determinism alone without the use of a proposition like Pt, then do it. But you can't, at least not while giving a version of the 2nd argument.

And it matters not whether YOUR argument is a simpler argument that achieves this goal. Your argument does not seem to be a version of the 2nd argument and my comments are restricted to versions of the 2nd argument, not ALL possible arguments for incompatibilism. I'd like to comment on one complex argument for incompatibilism at a time.

Joe says:

"Again, I have no problem with the phrase "the power to add to the actual past." I only have a problem with the claim that we MUST interpret "power" in the way that Andrew, John, etc. say we must (that is, in what Perry would call the "strong" sense of the term). Of course, if I were convinced that this way of speaking were the correct way of speaking, independent of considerations of the 2nd argument, etc., then I would think that the 2nd argument was sound. I can see how things go from that point forward. How can you get me to that point without presupposing the argument itself?"

I simply find it plausible that our freedom is the power to extend the past, holding the laws fixed. Additionally, I have tried to argue (in particular, in The Metaphysics of Free Will (and other places) that denying this leads to problematic results for practical reasoning. But I do not think there is a knockdown argument in the offing. I agree that one need not adopt this conceptualization of our freedom, and very smart philosophers such as Lehrer, Lewis, and Perry have been able to resist it.

As I have said, Semicompatibilism is the doctrine that moral responsibility does not require freedom to do otherwise; thus, we could be morally resposible, even in the absence of freedom to do otherwise. Thus, we could be morally responsible, even if causal determinism obtains and the Consequence Argument (in some form or other) is sound. But Semicompatiblism is consistent with also denying the soundness of the Consequence Argument.

Now John Fischer's Total Package of Views About Free Will is a bit more expansive than Semicompatibilism, and it includes the view that the Consequence Argument is indeed sound. But this is no part of Semicompatibilism, and I reiterate that I do not think that the CA must be accepted or that there is a decisive, airtight argument for it.

The comments to this entry are closed.

Books about Agency


3QD Prize 2014: Marcus Arvan