I'm attracted to a view of well-being with (roughly) the following structure: x is intrinsically good for y if and only if y believes that x is intrinsically good for y (under the right conditions, such as a coherent belief set). In other words, there is nothing more to something's being good for x than for x to believe that it is good for her. The closest cousin of this view is a desire-satisfaction view, which holds (again, very roughly, and leaving out all sorts of bells and whistles) that x is intrinsically good for y if and only if y desires x (under the right conditions, such as full-information, say). I think the belief formulation can plausibly respond to several worries that plague desire accounts (such as the so-called "paradox of desire" and the problem of welfare-irrelevant desires, etc.). But the view to which I'm attracted (call it, for lack of anything better, "belief-ism") faces an immediate and substantial hurdle: semantic circularity. It looks as though the analysis of "intrinsically good" or "prudentially valuable", etc., is given in terms of the concept it purports to be analyzing. Below the fold, I want to try out one possible response to this worry.
Strictly speaking, the way I stated the circularity problem above is not quite right. After all, on a belief-ist analysis, something is prudentially valuable if and only if it is believed to be so by the person for whom it is prudentially valuable. Hence the analysis of "prudential value" is not given in terms of "prudential value," but rather in terms of beliefs about prudential value. But this doesn't clear up the problem. We need to know what makes a belief about prudential value a belief about prudential value. Plausibly, this will be given by the semantic analysis of prudential value. Hence the circle reappears: the analysis of prudential value is given in terms of beliefs about prudential value, which are analyzed in terms of prudential value, which are analyzed in terms of beliefs about prudential value, etc., etc. David Brink writes: "It is true that, on this view, we analyze X, not in terms of X, but in terms of beliefs about X. But if we accept the not unreasonable assumption that any story about what makes a belief a belief about X must ultimately advert to X, then it appears that this sort of analysis is ultimately circular."
One possibility is to accept a form of primitivism. What it means to be a belief about prudential value is just primitive. (Sharon Street tries out a response like this in the normative domain in a nice essay "Constructivism about Reasons," which I recommend.) This might work, but it is unsatisfying. Let's call this Plan B. Furthermore, one surely can't give an analysis of "prudential value" in terms of some belief-independent or objective property. This would seem to defeat belief-ism, insofar as such a move would appear to imply that x is good for y if and only if x possesses the relevant belief-independent or objective property.
So what to do? I've been thinking about this problem for a long time, and only recently did I realize that the answer was staring me straight in the face. The basic idea is this. Belief-ism is a theory about the truth conditions of claims about well-being. What makes it true that x is prudentially valuable for y is that y believes that x is prudentially valuable for y. So why, then, do we believe that the semantic analysis of "prudentially valuable" must be given in terms of what is believed to be prudentially valuable? Simple: because our best theory of truth tells us that in order for a sentence to be true, the terms must successfully refer–very roughly, the meaning of the sentence must "correspond" to the world. Any other semantic analysis, and you're not going to end up with truth conditions that are compatible with belief-ism. Solution: reject the "correspondence" theory of truth for talk of prudential value (I put that in quotes because many theories beyond the simple correspondence theory will accept that truth conditions are determined by semantics) in favor of a "coherence" theory of truth for talk of prudential value. Such a view yields belief-ism: roughly speaking, “x is prudentially valuable for y" is true if and only if “x is prudentially valuable for y" is part of y's coherent set of beliefs about prudential value (or, assuming that y's belief set is not coherent, y's set of beliefs about prudential value rendered coherent). (I give a fuller specification of a coherence theory of truth in the moral domain here.) Furthermore, it does not require that the semantics of "prudential value" or "well-being" or "intrinsically good" have anything to do at all with the truth conditions of these claims.
Hence, my solution runs as follows: give any non-circular semantic analysis you want of "prudential value". I prefer a semantic analysis that refers to non-natural properties, say "property q". So when I believe that cherry pie is intrinsically good for me, I believe that cherry pie possesses property q. But the existence of non-existence or presence or non-presence of property q has nothing at all to do with whether cherry pie is intrinsically good for me, i.e., whether "cherry pie is intrinsically good for me" is true. The truth of these sentences is determined instead by their being coherently believed.
Anyway, that's my approach. I'm suspicious that what I'm arguing commits me to p and not-p somewhere down the line, but I'm not sure that I see where or how it does so yet. Also it would be helpful to be schooled those who are better versed in the philosophy of language; I'm sure I've violated a bunch of standard conventions.