Naturalness vs. "Arbitrariness" and "Simplicity" in Mereology

There are three answers one can give to the question: "When does composition occur?"

(1) Always.
(2) Sometimes.
(3) Never.

The first view is something like David Lewis's view: Any plurality of objects composes a third. Hence, for example, there is an object consisting of Barack Obama, my left leg, an orange, and half of the beach in Santa Monica. (We could call it the BLOB.) This is the view encompassed in classical mereology.

The third view is something like Peter Van Inwagen's view in 'Material Beings'. This view holds that (with maybe a few very specific exceptions), there are not, literally, any composite objects. There are just "simples" -- atoms in the void, physically proximate to each other and arranged in various ways.

The second view encompasses all other possibilities. One of these possibilities is "common sense" ontology, or something like it. One such view might hold that things like physical organisms, tables and chairs, rocks, planets, stars, maybe even galaxies, etc. are composite objects. But not just any plurality of things constitutes an object on this view; for instance, there is definitely no object such as the BLOB.

One argument (I think due to Van Inwagen) says that (2) can be ruled out rather easily because it is arbitrary and/or overly complicated. Hence, we must choose between (1) and (3).

However, it seems to me that people who hold to (2) might argue that their ontology only encompasses what is natural; just as there is a distinction between natural properties (like 'having mass') and gerrymandered properties (like 'being Barack Obama-or-my leg-or-an orange-or-half of Santa Monica Beach'), there may well be a distinction between natural composites and gerrymandered composites. And just as one might choose to privilege the natural properties by saying they are the only ones that exist (as D.M. Armstrong does), so one might choose to privilege the natural composites by saying they are the only ones that exist.

Obviously more needs to be said than this and this view would need to be fleshed out. But I'm more interested in the methodological question, and all I need granted is that it is a distinction one could coherently use so as to avoid (1) or (3).

Now, people like Van Inwagen might (probably, would) respond to this view by claiming that it is arbitrary, that it multiplies distinctions, that the notion of "naturalness" is mysterious and vague, and so on.

But I think it is worth noting here an "arbitrariness" in this objection: Claiming that some entity is more natural than another (or, by extension, that one's theory is more natural) is no more mysterious than claims that that (2) is arbitrary and complex, and that (1) and (3) are non-arbitrary and more simple. Defining the sense in which (1) and (3) are "non-arbitrary" and "simpler" is no easier than defining the sense in which (2) is "more natural."

Frankly, simplicity and arbitrariness, as used in this way, seem to me to be just as bad off as the other notions that anti-hyper-intensionalists use as criteria of theory choice; they are themselves hyperintensional notions in fact. That's not to say that they are bad off -- I do think there is an intuitive sense in which theories can be "simpler" and "less arbitrary" than other theories. But it is arbitrary to use "arbitrariness" and "simplicity" as criteria for selecting between metaphysical theories, and then pretend you don't know what it means when one says that his theory is more "natural" than others or, relatedly, posits entities that are "more natural."

Modal Realism and the Serviceability Argument

Here's a quote from David Lewis: "Why believe in a plurality of worlds? -- Because the hypothesis is serviceable, and that is a reason to think that it is true."

Question for David Lewis and other modal realists: Lots of worlds are serviceable, not just the metaphysically possible ones. Many times when we do semantics, discuss language, give thought experiments, etc., worlds which are strictly logically possible but not metaphysically possible are helpful. For example, one of the ways that intensional semantics deals with oblique transitive verbs, control verbs, etc. is by invoking worlds where, for instance, water is not H2O, or where Hesperus is not Phosphorus. Presumably these are not metaphysically possible worlds, but rather 'logically' possible worlds. (Sometimes metaphysically possible worlds are called 'broadly' logically possible worlds; by 'logically' possible worlds I mean what are sometimes called 'strictly' logically possible worlds.)

Do these exist too, in exactly the same way as the metaphysically possible ones? If yes, then we run into problems. After all, isn't it only the metaphysically possible worlds which can exist? If not, then what is the distinction between metaphysical possibility and mere logical possibility supposed to mean? In fact, if merely logically possible worlds exist just like the metaphysically possible ones then there is no distinction. But there is, of course, a distinction.

At the very least, aren't the metaphysically possible worlds the only ones which could be actual? But if 'actual' is indexical as Lewis thinks, and the logically possible worlds exist on a par with the metaphysically possible ones, then any of these worlds could be actual.

Personally, I think there's just as good reason to admit the existence of logically impossible worlds as there is to admit the existence of possible worlds (though I don't think there's much reason to admit the existence of either).  If we really needed possible worlds, I think we'd need impossible ones too. But if logically impossible worlds are serviceable too then that makes things even worse for the modal realist. After all, what would it mean to say that a logical contradiction actually holds true in a concrete world just like ours? Clearly there are no such concrete worlds, since whatever concretely exists must at least be possible. But even if one resists the need for impossible worlds, the metaphysically possible worlds are a proper subset of the strictly logically possible ones, and it should be clear that these latter are "serviceable" too.

In sum, if Lewis's argument works for the existence of concrete metaphysically possible worlds, then it works for the existence of metaphysically impossible worlds too. But these can't exist concretely; that's the whole point of making some metaphysically possible and others not. Hence, Lewis's argument does not work. This can be taken as either reason to abandon the 'serviceability' criterion of existence, or as reason for rejecting concrete possible worlds. I'm inclined to reject both.

Naturalness in the World

I'm currently taking a course on a newly concocted discipline in analytic philosophy called metametaphysics. For Aristotle or Aquinas the topics grouped under this heading would probably just fall under the science of metaphysics. We are primarily reading from the recent Chalmers volume and Ted Sider's new book Writing the Book of the World. In the latter book quite a bit turns on a distinction which Sider calls variously 'naturalness', 'fundamentality', 'structure', and 'carving at the joints'. My professors claim to not understand what Sider is talking about. Admittedly, Sider could be quite a bit more clear. However, I think that an Aristotelian would want to agree at least in some respects with his general point insofar as we'd admit some boundaries in reality to be privileged over others.

Consider an example Sider gives which is represented in the figure above (I'll paraphrase these examples a little bit). Suppose there is a world filled with red and blue liquid. There are many true ways that we can divide this world up when we describe it. We can divide it into those parts satisfying the normal understanding of our predicates 'red' and 'blue' as in the first figure. We could also divide it into those parts satisfying different predicates, call them 'bled' and 'rue', which correspond respectively to the portions left and right of the diagonal in the second figure. Both of these ways of describing the world are true. There are indeed bled and rue portions of the world just as much as there are red and blue ones; or, to put it another way, there are divisions of the world along the lines of both the first and the second figure. Yet the first way of speaking seems in some way to be natural while the second seems bizarre. Sider wants to assert that the first way of dividing up the world, i.e. dividing it up along the lines of our 'red' and 'blue' predicates, is better because it describes those features of reality which are in some way privileged; to put it in his terms, it describes those features which are natural/ fundamental/carve at the joints/are part of the structure of reality.

It seems like Sider gets a lot of his account from David Lewis's work. It is at least similar to the natural vs. non-natural distinction Lewis makes in his paper "New Work for a Theory of Universals." Consider two properties: being green and being grue, where grue is defined here as the property of being green and observed before 3000 A.D. or blue and not examined before 3000 A.D. This property which we have called grue appears "gerrymandered" in a way the property green is not. In Lewis's term grue is a "less natural" property than green is. Sider agrees on this point with Lewis. It should be noted that what makes grue a less natural property is not the syntactic complexity of its definition--after all, we can just give it a simple name like grue--but rather that it is in some way privileged over these other properties as being a more fundamental feature of reality. (Lewis may actually disagree here but I think his view is highly implausible and Sider does not seem to endorse it.)

Take a final example. Consider two classes of things: the electrons, and the electron-or-cows (EoC's), the latter class consisting of everything which has the property of being an electron or a cow. The things in the first group seem to go together quite well in a way which the objects in the latter group do not. The electrons do not go together better simply because they share many properties. For one, the EoC's share many properties. In fact, the EoC's share infinitely many properties: they each have the property of being an EoC or four feet long, the property of being an EoC or five feet long, the property of being an EoC or six feet long...and so on.  So it's not simply the number of properties shared which distinguishes the two. It's the fact that a grouping of things into electrons and cows gets the way reality fundamentally is, whereas a grouping of things into electron-or-cows does not.

So these are just a few examples where we can contrast fundamental/natural features of reality in opposition to gerrymandered or non-natural ones. It is necessary to use such illustrations since fundamentality is taken to be a primitive distinction which is likely not definable in more basic terms. There are a lot of questions that can be asked here: is fundamentality/naturalness/carving at the joints/structure of reality/etc. itself fundamental? Is fundamentality supposed to be a property? Is fundamentality a feature of our thoughts and concepts, of entities, or both? What is fundamental? How can we know? Is the same notion at work in each case? These are all good questions, some of which are discussed in the book.

However my primary concern is this: Is Sider really getting at some objective distinction or not? What I mean by 'objective' is whether his distinction between the fundamental/non-fundamental really corresponds to something in the external world. Personally I think he is onto something and makes an interesting case regardless as to the connections with Aristotelianism. However, I think an Aristotelian would want to admit the distinction even if he might disagree about what the fundamental things are. I will try to explain why in later posts.

If anyone is reading, I'd especially like to hear your thoughts on (1) whether you think Sider's distinction makes any sense and (2) whether the examples illustrate the point.