Monday, April 26, 2010

Powers vs. Humean Supervenience

I’ve been interested in ontological accounts which feature dispositional properties (also known as dispositions or powers) in the starring role.  This intriguing draft paper, “Goodbye, Humean Supervenience”, by Oxford's Troy Cross, sets out to show that David LewisHumean supervenience program (which seeks to reduce everything to a mosaic of categorical properties) fails, since it can be reinterpreted as a system where Lewis’ fundamental properties are in fact identical to dispositions.  This implication can’t be rejected without abandoning the account of properties as sets of possibilia.   The possibility unattractive feature of the dispositional doppelganger of Lewis’ story is its appeal to nonlocal circumstances as the trigger for the manifestation of the powers.  My notes on the paper follow.

Cross begins by placing the debate in a historical context, discussing the role of causation (and nearby topics of counterfactuals, nomic necessity, etc.) in analytical philosophy in the 60’s and 70’s.  This was driven by naturalism and its attendant need to find “topic-neutral” concepts to describe the mental and physical non-dualistically.

But then, even causation and related concepts came to be seen by some as too much like “spooky theological vestiges”, which needed to be reduced – hence Lewis’ program.  Others, such as Shoemaker, retained causal powers and endeavored to explain all else with that toolkit.  Cross sets out to review Shoemaker’s work and uses it to provide a launching point for his attack on Lewis.

Shoemaker’s thesis in “Causality and Properties” (1980) said that properties were uniquely correlated with powers that they bestowed.  To be more precise he identified powers possessed by an object with a function from circumstances to effects, and then said every property necessarily and uniquely correlates with a function from other properties to these powers.  Cross notes a circular aspect of this system, since the function by which Shoemaker individuates properties itself takes properties as arguments, and the circumstances and effects are a matter of which properties are instantiated where.

To argue for his view, Shoemaker pointed out that if properties were untethered to powers of objects, we would never know about them.  So they must be tied to powers.  He constructed reductio-style scenarios where properties are assumed to be unrelated to powers and then odd things result (2 objects which appear identical and have the same causal powers in fact differ greatly in their properties).

But critics armed with modern epistemological theories have been able to respond to these scenarios effectively.  More devastatingly, Cross says, is that even if properties are linked to powers you can still construct analogous skeptical scenarios.  To see this, note that dispositional differences between objects need never be manifest in the actual world:  they may even contain potential manifestations alien to the actual world.  These quasi-inert powers can play the same role as the causally inert properties in Shoemaker’s scenarios.  So Shoemaker’s ontological account is not seen as panning out.

However, this notion of “alien sensitivity” – that dispositions may have radically non-actual activation conditions – gives Cross an idea for his argument that even paradigmatically categorical properties are in fact causal powers.

First, note that Lewis would admit that there are possible conditions such that for some one of his (arbitrarily perfectly) natural properties "F", F endows some power to objects in these conditions.  (He doesn’t say they can’t endow powers, just that they would endow different powers in different circumstances).  Now exactly how for Lewis the powers granted by F vary with circumstances is complicated.  But we can look at a simpler nomic theory to get the idea:  such a theory would involve laws of nature in determining when F endows the power to become a G, etc. (Some alternative theory may be better, but what matters is that there are possible conditions in which F disposes things to become G’s.)  The conditions may be states of the world, or sub-regions, but it’s not random.  On an account of Humean supervenience, it’s a matter of how powers generally supervene on the distribution of Humean facts.

Now, let “C” be this set of conditions under which F endows the disposition to become a G.
1)    1) In C, F disposes things to become G’s.
2)    2) Now compare: F disposes things to become G in C.
This second version strengthens the link between F and the disposition (compare "in England, this joke is funny” to "this joke in funny in England").  But it’s hard to resist that 1) implies 2).  Cross says these two ways of presenting things are just a general feature of power and disposition talk.  Especially easy to see in the case of habituals:  F’s G in C iff in C, F’s G.

In contrast, note this doesn’t work for subjunctive conditionals, in fact the cases where it fails for the subjunctive “are precisely those cases where dispositional ascriptions succeed and counterfactual or subjunctive analyses appear to falter.”  So called “finkish” dispositions are those which disappear in their stimulus conditions (see the work of Martin.)  Cross claims that “all counterexamples to the pattern of subjunctive inference above are also cases of finks, antidotes, or mimics.”  So, swapping the conditions in and out of the scope of the power operator doesn’t inherit the danger of the analogous operation for subjunctives.

To recap: “If there are possible conditions C, under which a “categorical” property F disposes things to become G, then throughout modal space, F disposes things to become G in C.”  This is the key move.
“Some parts of the pluriverse are poised to send the property into action.”  By the same token, the property is poised to go into action in those parts of the pluriverse and not others.   So, these categorical properties are far from the causally inert entities they’re supposed to be, seen in this light.  Properties do endow powers, in this way, and different properties will endow different powers (this distinguishes each property from another).

Now, for Lewis, all properties are sets of possiblia   Sets are individuated by their membership. So if a perfectly natural property P and a power P’ endowed by P are coextensive, then P=P’.  So Lewis is committed to (all) perfectly natural properties being powers.  So Humean supervenience “isn’t worthy of the name.”

So, you take global states of the world, such as possible laws of nature, and treat them as circumstances, which together with the original property, trigger an effect.  Then you see that the circumstances reveal the disposition in the supposedly categorical property.

Is this cheating?

A standard objection would be that these sort of global circumstances are of a different category than the standard activation conditions for a disposition such as “being put in water” or “being dropped”.  Laws concern the relations between ordinary properties, and “being in a world with such-and-such laws” is itself a property only in a highly derivative sense.

Can the critic insist you must try to run this without involvement of laws, a non-nomic version of the conditions?  Doesn’t make sense – if (as in Shoemaker) laws of nature are necessary, then every circumstance is in a sense nomic.  There aren’t any non-nomic conditions.  But before continuing down this path of trying to restrict the conditions, consider the big picture.  Compare the philosophy of Lewis to that of the dispositional doppelganger “DLewis”.  Same metaphysics, except DLewis says all the perfectly natural properties are dispositions to have certain effects, conditioned on the global state of the world.  Which global states?  Why, precisely the ones Lewis says are the conditions in which his perfectly natural properties endow the powers to bring about those effects.

So for Lewis (simplifying - leaving aside the fact that Lewis’ theory is more complicated than the straightforward nomic case) F endows the power to become G just in the circumstances in which it is nomically necessary that F’s become G’s.  For DLewis, the global states do not constitute laws; he thinks, instead, that laws reflect the dispositional essences of the properties.  Lewis sees as contingency which powers are granted by F.  DLewis sees contingency in the existence of the activation conditions for the F dispositional powers.

Lewis may not like it, but the DLewis account is just as coherent, and makes no difference.  So why object and insist on the existence of “genuinely” non-dispositional properties?  The only way to do this would be to insist properties can be non-identical even if they’re necessarily co-extensive.  (Go to some hyperintensional model, which Lewis would not wish).  The general lesson being that “the only way to avoid a reality that is cut finer than the modal is to accept a reality that is fundamentally dispositional.”

No comments: