Reconstructing Moore's Open Question

Andy Altman (an old professor of mine at Georgia State University!) wrote a paper some years ago in which he proposed a new construction of Moore's open question argument (OQA). The argument confuses me, so I wanted to work through it a bit here.

Moore's open question argument seeks to demonstrate that reductive analyses (what Moore calls “definitions”) of GOOD (I'll use caps to designate properties) are bad. Suppose a suggested definition of GOOD has it that x is good iff x is f. It will always be an open question whether or not something which is f is good. For instance, if you define GOOD as PLEASANT, then you are committed to every question of the form “Is x, which is pleasant, also good?” being a closed question, decidable strictly in virtue of GOOD's definition. The question is open, so the definition must be flawed.

The OQA famously has four problems. I only understand three of them, so I'll give those three here. First, it begs the question. The claim that any question of the form “is x, which is f, also good?” will be a closed question for any possible f is only plausible on the assumption that no definition of GOOD is possible. Second, Moore generalizes hastily. Moore considers a few candidate definitions of GOOD, shows that those definitions do not succeed in yielding a closed question when plugged into “is x, which is f, also good?”, and concludes that all other candidate definitions will suffer the same fate. Third, Moore hadn't yet learned a lesson that we've all learned from Kripke and Putnam: sometimes identity statements can be informative. Superman is identical to Clark Kent; nonetheless, the question, “is s, who is Superman, also Clark Kent?” is plausibly open. Ergo, GOOD may in fact be identical with some other property to which we could then reduce it, and this wouldn't be ruled out by Moore's ability to generate an open question.

Andy's solution is to recast the open question as follows: properties to which GOOD can allegedly be reduced, or by reference to which GOOD can allegedly be defined, can all instantiate GOOD. For instance (it may be that) BEING PRODUCTIVE OF PLEASURE is a good property. GOOD, however, cannot instantiate itself. Therefore GOOD can not be identical to any property which can instantiate GOOD. Therefore, GOOD cannot be identical to any of those other properties. QED.

This recasting of OQA avoids the above objections. It doesn't obviously beg any questions (I am not assuming that GOOD isn't identical to some other property when I observe that it can't self-instantiate, or when I observe that PLEASANT can be good.) It does not hastily generalize. The incapacity of GOOD to self-instantiate is unrelated to the capacity of other properties to be good, and if I think that any property (except GOOD) could (intelligibly) be good, and that GOOD can't, I need not get to that conclusion by generalizing from a few cases. And I don't need to deny that identity claims can be informative in order to get from GOOD's not being self-instantiable and other properties' being potentially good to the conclusion that GOOD is not identical with any of the latter properties.

We have a distinction between properties that self-instantiate and properties that cannot. STRONG is not itself strong, SQUARE is not itself square, BEING A LAWYER is not a lawyer. BEING A PROPERTY, however, is a property. NOT BEING A NUMBER isn't a number. Good is like STRONG, not BEING A PROPERTY.

Plato seems to have thought that the form of the Good was good. I asked a few people (without adornment) whether GOOD is good. They said yes. So what Andy's saying isn't obvious. Andy's argument for the claim that GOOD doesn't self-instantiate is that in general we just know that most properties can't self-instantiate. There are some interesting exceptions, but they are exceptions to a fairly ubiquitous rule. Unless we have reason to think that GOOD is such an exception, we should continue to assume it doesn't self-instantiate.

But here's my confusion. I would have thought that the reason most properties can't self-instantiate is that most properties can't be properties of properties. Some properties can be properties of properties, so they can self-instantiate (and some, not all, do). It's not only STRONG that isn't strong; no concept is strong. STRONG isn't the kind of property that a property can instantiate. Properties can instantiate most negative properties, so they can instantiate NOT BEING A NUMBER or NOT BEING A UNICORN. The general, ubiquitous rule that most properties can't self-instantiate has less to do with the funny relationship that properties have with themselves, and more to do with the fact that some properties (such as STRONG, HUMAN, ANGRY) can't be instantiated by properties. Since they, themselves, are properties, ipso facto they cannot instantiate themselves.

This brings us back to GOOD. I'm inclined to agree that GOOD isn't good. But that's because it's not clear to me that any property can be good. It may be that some properties are such that anything that instantiates that property is good, or it can be good for something to instantiate that property, but that doesn't amount to the property itself being good. So I think we can provide what Andy's analysis doesn't: a firm reason that good can't self-instantiate. But that explanation comes at a cost to Andy's argument. Because if I'm right, and GOOD can't self-instantiate because no property can instantiate GOOD, then neither can any of the properties by reference to which GOOD is allegedly being defined. So, being unable to instantiate GOOD is something that GOOD and (for instance) BEING PRODUCTIVE OF PLEASURE have in common. This strategy does not, finally, reveal a property that GOOD has and which its alleged analysandum lacks.

So suppose we wanted to preserve the intuition (which I must admit I don't feel at all) that the property BEING PRODUCTIVE OF PLEASURE can be a good property. What on earth could it mean for such a property to be good? I can think of a few candidates. It could mean it would be good if more things had that property, or having the property is good, or all good things have that property. These all feel a little forced to me, but let us countenance them. The problem is that if this is what it means for a property to have the property of GOOD, then GOOD has that property par excellence! The only alternative I can think of is that for a property to be good is for it to be good that the property exists. On some immanent theories of properties, for a property to exist is to have an instantiation, so this is no more than to say it is good that something has the property which is clearly a property that GOOD has. On other theories of properties, properties are Platonic necessary existents or sets of possible individuals. Strange indeed to think it good that necessarily existing abstracta exist, or that sets of possible individuals exist, but (more to the point) if you are comfortable with such strangeness then there is no reason for you to become suddenly uncomfortable when we consider the possibility of it being good that GOOD exists. And, finally, for the nominalists, if properties are linguistic items akin to predicates or mental entities of some kind, then it can certainly be good that properties exist (perhaps because their existence requires languages and thinkers, respectively, and it might very well be good that such things exist), but then it is as good that GOOD exists as any other property. So any sense in which any property can instantiate GOOD, I'm inclined to think, is a sense in which GOOD can instantiate GOOD as well.

Altman's reconstruction of Moore's argument may well save OQA from canonical objections. But it invites a whole host of new, more damning objections. The standard reading of Moore, in spite of the hurdles it faces, is better equipped to conquer its obstacles than is this newer and paler imitation.