Wednesday, July 25, 2007

A puzzle about supervenience arguments for dualism (x-posted from T&T)

Suppose there's a qualitative duplicate of the actual world (It might be a world with haecceitistic differences from the actual one, but it doesn't have to be). Call the actual world A, and its duplicate, B.

I'm conscious in world A. Call the extension at the actual world of the things which are conscious S. There are cauliflowers in world B. Call the extension at B of the things which are cauliflowers, S*. Now consider the gruesome intension cauli-consc, which has S as its extension at world A, and S* as its extension in world B (it doesn't matter what its extension is in other worlds: maybe it applies to all and only conscious cauliflowers).

Is there a property that things have iff they are cauli-consc? So long as "property" is intended in an ultra-lightweight sense (a sense in which any old possible-worlds intension corresponds to a property) then there shouldn't be an trouble with this.

However. Cauli-consc is a property that doesn't supervene on the pattern of instantiation of fundamental physical properties. After all, A and B are alike in all physical respects. But they differ as to where cauli-consc is instantiated.

Cauli-consc is a property, instantiated in the actual world, that doesn't supervene on physical properties! Does that mean that the fact that I'm cauli-consc is a "further fact about our world, over and above the physical facts" (Chalmers 1996 p.123)? That is, do we have to say that, if there are such qualitive duplicates of the actual world, then materialism is shown to be wrong by cauli-consc?

Surely not. But the interesting question is: if some properties (like cauli-consc) can fail to supervene on the physical features of the world, what is that blocks the inference from failure of supervenience on physical features of the world, to the refutation of materialism? For what principled reason is this property "bad", such that we can safely ignore its failure to supervene?

Here's a way to put the general worry I'm having. Supervenience physicalism is often formulated as follows (from Lewis, I believe): any physical duplicate of the actual world is a duplicate simpliciter. But if duplication is understood (again following Lewis) as the sharing of natural properties by corresponding parts, then to get a counterexample to physicalism you'd need not only to demonstrate that a certain property fails to supervene on the physical features of the world, but also that some natural property fails to supervene: otherwise you won't get a failure of duplication among physical duplicates. The case of cauli-consc is supposed to dramatize the gap here. Sometimes it looks like you can get properties which fail to supervene, but which don't seem to threaten materialism.

However, when you look at the failure-to-supervene arguments for dualism, you find that people stop once they take themselves to establish that a given property fails to supervene, and not, in addition, that some natural property does so (For example, Chalmers 1996 p132 assumes that it's enough to show that the 1-intension of "consciousness" fails to supervene, without also arguing that it's a natural property) .

Now, I think in particular cases I can see how to run the arguments to address this issue. Add as a premise that e.g. the 1-intensions of the words of our language supervene on the total qualitative character of the world, so that we're guaranteed that if there's a world in which "1-consciousness" is instantiated and another where it isn't, those can't be qualitative duplicates. If now we find a failure of 1-consciousness to supervene on physical features of the world, we'll be able to argue for the existence of physical duplicate worlds differing over 1-consciousness, we now know can't be qualitative duplicates. (In effect, the suggestion is that the sense in which cauli-consc is bad is exactly that it fails to supervene on the total qualitative state of the world).

That all seems reasonable to me, but it does start to add potentially deniable premises to the argument against materialism. (For example, I'm not sure it should be uncontroversial that consciousness supervenes on the total qualitative state of the world. Is it really so clear, for example, that there are no haecceitistic elements to consciousness: that a world containing me might contain a conscious being, but a qualitiative duplicate containing some other individual doesn't?)

So I'm not sure whether the elaboration of the Zombie argument for dualism I've just sketched is the way Chalmers et al want to go. I'd be interested to know how they have/would respond (references welcome, as ever).


Carrie Jenkins said...

Hmmmm ... how compelled am I to believe in a possible world which is a qualititative physical duplicate of the a.w. but distinct from it (so that in particular it has a different distribution of some property)? Can't I blame the assumption that there is one of those for the problem that cauli-conscness looks like a property that doesn't supervene on the physical?

Robbie said...

There's a bunch of reasons why you might believe in the starting point.

One is if you're a haecceitist. E.g. suppose the actual world has a reflective symmetry along some axis. Maybe there's a world where I and my mirror twin are in each other's locations (and that's not just a matter of funky business with counterparts). That appears to be a qualitative duplicate of the actual world with merely haecceitistic differences.

Another reason for worrying about such scenarios is if you have no reason to rule them out given your metaphysics of worlds. I guess this was Lewis's view: as I recall, he thought he had to be agnostic about how many concrete space-times duplicating our own there were (from 1 to infinitely many!). Since any of these pluriverses look to sustain the same modal truths, it doesn't look like there's much to choose between them (I dunno: maybe quantitative parsimony?).

A last thing. Consider a world where some random device is about to land H or T, and then the world will be destroyed. Suppose that there are only two nomically possible qualitative ways for the future to go, so that nomically possible worlds relative to the present time divide into two equivalence classes under duplication: worlds with H outcomes and worlds with non-H outcomes.

That doesn't yet guarantee that there is more than one world in either class. But suppose in addition the following: if there's an objective chance 2/3 that a flipped coin lands heads at t, then there proportion of nomically possible worlds in which the coin lands heads to those where it doesn't has to be 2/3. In that case, if the random device has an objective chance of 2/3 of having the H outcome, then we need twice as many H-worlds as non-H worlds. But that means there's at least two H-worlds, but they're qualitative duplicates.

I actually would endorse most of this argument: except I think there's a dodgy slide between the "proportion" of H-worlds to non-H worlds being 2:1, to there being two distinct H-worlds to every non-H-world (measures over a space of possibilities needn't take that form). But in the finite case at least, there's at least some appeal to allowing the move. (Notice that nothing here requires there to be haecceitistic differences between the qualitative duplicate futures.)

Anyway, none of this commits *you* to the possibility, but I hope it indicates why a bunch of people may well be committed to allowing it.

Louis said...

Hi Robbie,

I don't know the consciousness debates well, but I don't think you need to get into the difficult independent metaphysical disputes over haecceitism, etc., to answer Carrie's question. Isn't the right way with her question to say: the dualist supervenience arguments are committed to the existence of worlds that are qualitative physical duplicates of the actual world: zombie worlds or what have you. Suppose that there is a zombie world w. Specify cauli-consc with respect to the actual world and w, just as in your main post. Then there is another property, cauli-consc, that also doesn't supervene, but anti-physicalism doesn't follow from the failure of cauli-consc to supervene. So the dualist conclusion requires more support.

Your example is really intriguing, btw, and I am very sympathetic to the idea that supervenience (or its absence) has no interesting metaphysical upshot. But can't the dualist cry foul on the grounds that you've pulled a logician's trick with a world-indexed property like cauli-consc. So the restriction isn't to natural properties, but to non-world-indexed properties. And consciousness does not seem to be world-indexed (even if a little logical gerrymandering can make it look that way).

- Louis deRosset

Carrie Jenkins said...

Hi Robbie,

I guess I remain sceptical (on my own behalf - of course other physicalists would have to speak for themselves, but the sensible ones should agree with me!). I've never been a fan of world-counting approaches to understanding chance. And the haecceitist you describe has already given up on supervenience, so a few more failures should come as no surprise to her. As for having no reason to rule out these worlds, I would have thought believing in supervenience on the physical is a reason to rule them out ...

Robbie said...

Hi Louis,

I guess one thing I thought was interesting about the original setting was that cauli-consc was demonstrably consistent with physicalism, even in supervenience formulations (e.g. "every physical duplicate of the actual world is a duplicate simpliciter"). And that allowed the strong conclusion: a failure of a property instantiated at the actual world to supervene is consistent with physicalism being true at the actual world. If we ran things from the Zombie world case in the way you suggest, then we can't quite do that.

I suppose, though, that one might run your thought the other way around. Suppose that the actual world lacks ectoplasm, but some physical duplicate has it. Now consider an analogous cauli-ectoplasm property. This property instantiated in the actual world (by cauliflowers). But it doesn't supervene on the physical facts, since in the ectoplasm-enhanced physical duplicate of actuality it's differently instantiated. So again the failure to supervene of this-worldly properties looks consistent with physicalism.

I'd need to check carefully through the various definitions of "supervenience" at work to make sure that fancy footwork doesn't get someone out of this one. But again, the wider significance might be to consider whether consciousness might be a non-supervening property of this general form. This reminds me of ideas that Stephan Leuenberger has explored in his paper "Ceterus Absentibus Physicalism" (it's coming out in the next Ox Studies in Metaphysics I believe: it won the Young Scholar's prize last year). But I'd need to check. Stephan's coming to Leeds as a postdoc this Sept, so I'll be able to ask him in person soon!

Robbie said...

Hi Carrie,

Thanks for this. I agree that if the only examples of qualitative twins of the actual world differ on haecceities, then we could shortcut the discussion by pointing out that haecceitistic properties don't supervene on the physical qualitative facts: and of course we could move straight to the suggestion that consciousness might be object-involving (though I've no idea if that bit will lead us anywhere useful). I think it's useful to have the general recipe, though.

I don't want to suggest that we analyze chance by counting worlds. the direction of explanation might be in the other direction. Maybe there are primitive propensity facts, but that when constructing an account of what possible worlds there are, we get a smoother theory if we duplicate some of the worlds, so that the connection between counting worlds and chance facts is sustained.

I wouldn't want to rest on this. In fact, I'm not strongly committed to there being qualitatively duplicate worlds. It just strikes me as a bad thing if the effectiveness of supevenience arguments had to be committed to the *impossibility* of this: it just looks like the sort of hostage to fortune that a theory shouldn't be making.

(I've been thinking recently that one theoretical virtue that I take really seriously in philosophy is what you might call "modularity": a theory of X's being self-contained and neutral on matters apparently unrelated to X's. And it's partly in that spirit that I'd like to see a formulation of arguments for dualism which are unaffected by delicate issues in the metaphysics of worlds).

Louis said...

Hi Robbie,

Thanks for clearing me up. I think, then, you have two arguments: (i) an argument for the conclusion that a failure of a property instantiated at the actual world to supervene is consistent with physicalism being true at the actual world; and (ii) an argument that a failure of a property instantiated at the actual world to supervene is consistent with that property-instantiation failing to be a "further fact about our world, over and above the physical facts". (i) depends on defending the idea that there are non-actual worlds that are physical duplicates of the actual world; but (ii) does not.

I don't think going to ectoplasm will help save (i) for all forms of physicalism, by the way, since many would claim that you have to throw a capstone "that's all!" claim into the subvenience base in formulating physicalism. (This is all over Chalmers's and Jacksons's work). So cauli-plaz differences between the actual world and the ectoplasm world don't show lack of supervenience, since there are also differences with respect to the subvening properties.

Here's an obvious generalization of your argument for (ii) to all supervenience arguments for dualist conclusions. A P-Q-dualist is one who claims that the Q-facts do not obtain in virtue of congeries of P-facts. Suppose we are given the premise: There is a world W which differs from the actual world with respect to the Q-properties but not with respect to the P-properties. Pick any property p_i from among the P-properties. Specify a property NS (the "non-supervening property") so that all things which have p_i in the actual world have NS, and all things which do not have p_i in w have NS. NS fails to supervene on the P-properties. A certain thing's being NS is still not a "further fact about our world, over and above the P-facts." So the premise does not by itself yield the P-Q-dualist conclusion. QED.


Robbie said...

Hi Louis,

That all sounds right to me. And what you say a about the cauli-plaz argument sounds right: I had forgotten about the "that's all" clause.

Actually, why doesn't that work against your original proposal too? Essentially all I did structurally was to switch around which world we were supposing to be actual. My head hurts when I try to think about "that's all" facts, though, so I might be missing something here.

I didn't answer your original point about world-indexed properties. But could you clarify for me what a world indexed property is? I know I was gerrymandering things in describing the property, and that sort of looks ineliminable. But, for example, if we're into Lewis's picture of properties as sets of possibilia, then what I described was just one more set. Do you think there's an objective line here to be drawn? Does it depend on what your metaphysics of worlds and possibilia (and abundant properties) is?

Louis said...

Hi Robbie,

"That's all" clauses also cause me pain; and there are an awful lot of evidently different proposals for articulating the "that's all" idea in the literature. But I think I have enough of a grip on the general idea, to say, intuitively, why the "that's all" response doesn't work against my original proposed response to Carrie's worry. Ectoplasm worlds are worlds in which there's a new kind of thing (or stuff), ectoplasm, in addition to all the kinds of things that there are in the actual world. That's why ectoplasm worlds will falsify any adequate articulation of the "that's all" idea. Zombie worlds, if there are any, aren't worlds in which some new kind of thing is added to the actual kinds; it's a world in which some actual kind of thing is missing. (Chalmers also talks about inverted spectrum worlds, right? These don't involve new kinds, either.)

WRT world-indexed properties, I guess I do think there's an objective line to be drawn, indicated by the ineliminability of world-mentions from the specification. This distinction is something that adequate theories of properties should capture, not deny. (Compare Lewis's attitude to the gruesome/non-gruesome distinction: "If I have to, I'm happy to accept it as primitive. But I think I can explain it.") I don't really work in metaphysics of properties and I'm writing right off the top of my head, though, so you shouldn't consider this an informed opinion.

Lewis himself acknowledges that we may require structure in properties (for some purposes), and supplies resources for explaining that structure in "Plurality". I'm thinking here of his discussion of triangularity vs. trilaterality. If so, then Lewis's own theory can represent the difference.

Here's a rejoinder on your behalf: it is easy enough to show that (on a Lewis-style view), to the structured property cauli-consc, there corresponds a non-structured property that also doesn't supervene on the P-properties: it's the extension of cauli-consc. This property can (in principle) be specified by enumeration. (If there's worries about enumerating an infinite set, make sure that cauli-consc has finite extension.) I think I'll leave matters there, since I am unsure whether you want to go down the rabbit-hole on this.

At bottom, what makes your example go is that, if the assumption Carrie resists is granted, the identity of the world inhabited by a physical thing does not supervene on the distribution of physical properties in that world, even though it is not a "further fact about the world, over and above the physical facts about the world".


djc said...

Hi Robbie,

Interesting question. I'd definitely take the line you suggest at the end. Physicalism should be construed as the thesis that all qualitative properties supervene on physical properties, where qualitative properties are those that supervene on fundamental properties (or on qualitative properties in some narrower sense, or something like that).

(I'm not sure that I discuss this issue anywhere in print, but it's come up in conversation from time to time, and I mention the issue briefly in my powerpoint "65,536 definitions of physicalism".)

Regarding your worry at the end: yes, I suppose room is opened up in philosophical space for a view that denies that phenomenal properties are qualitative properties. e.g., maybe they are identical to certain object-involving properties. I don't think this a very attractive view, though. Obviously there's a strong intuition that the difference between conscious beings and zombies is quite different in kind from the difference between qualitative twins -- intuitively, it's a qualitative difference. Less flat-footedly, the hypothesis that phenomenal properties are broadly physical object-involving properties seems vulnerable to the intuition that there's a zombie world with all the same broadly physical objects as in our world.

Another relevant difference is that arguably, object-involving properties (and any other non-qualitative properties) don't yield epistemic gaps. E.g., once I know the distribution of qualitative properties in this world (and indexical truths), I don't seem to have ignorance of which individuals are which. at least, there's no thought I can entertain of whose truth-value I am ignorant (contrast this with the consciousness case). In the 2D framework, this point corresponds to the claim that even if there are qualitatively identical worlds, 1-intensions don't discriminate between them. That premise, combined with the 1-conceivability of zombie worlds and the thesis that 1-conceivability entails 1-possibility, seems enough to rule out the view in question. But of course some philosopher could deny the premise -- and on we go.

Robbie said...

Hi Dave,

I'm going to have think some more about Zombie intuitions and the epistemic gap. I guess what'd help is if there are some objects identifying knowledge of which requires knowledge of their phenomenal properties (these need to be physical objects here: it won't help the physicalist if Cartesian egos are the only things satisfying that description). If so, then the epistemic gap wouldn't open: for to know (in the ideal limit) which objects are around, you also have to know some phenomenal properties.

I'm not sure what I think of that: the whole notion of "identifying knowledge" scares me a little. But it strikes me that the action is likely to be in this area.