Smith goes on to suggest that the truth-maker for a priori beliefs is merely constructive, i.e. the set of a priori beliefs that "we would all converge on if we were to subject our initial beliefs about what is a priori true to a reflective equilibrium process and so came up with a maximally informed and coherent and unified set of beliefs about what is a priori true."
If you then replace 'a priori belief' with 'desire', and 'a priori true' with 'valuable', then that pretty much yields Smith's analysis of normative facts. We have just as much reason to think that maximally coherent desire sets will converge, as we do to think that maximally coherent a priori belief-sets will.
It's widely believed that there need not be any such convergence of desires. Two perfectly informed and fully rational people might nevertheless retain different values and desires. The thought is that what values a fully rational person brings out of reflective equilibrium will be "radically relative" on those values he began with. But this skeptical argument will also extend to a priori beliefs:
Consider two people who have rather different beliefs about what is a priori true from each other to begin with, prior to engaging in the reflective process. Shouldn't we suppose that what one of them ends up believing to be a priori true, after subjecting their beliefs to a reflective equilibrium process, will be radically relative to the beliefs that they had to begin with? Absent the causal regulation of their various beliefs by the one set of apriori truths - that is, given the constructivism - the answer must be yes. (p.294)
Smith goes on to explain how we may preserve a priori truths nonetheless:
Given that, on this view, people have to converge in their beliefs about what is a priori true for any claim about what is a priori true to be true at all so it follows that, if we wish to interpret people as having beliefs about what is a priori true where some such beliefs are indeed true, then we simply have to so interpret them that they would converge if they had a maximally informed and coherent and unified set of such beliefs. (p.295)
The same is then said of desires: if we wish to interpret people so that it's possible to have true beliefs about what they have normative reason to do, then "we have no choice but to interpret them in such a way that the differences between them would slowly cancel out as they came closer and closer to having a maximally informed and coherent and unified set of desires."
That all sounds rather question-begging to me, and indeed Smith admits that "none of this is to say that there is good reason" to interpret people in such a way. Nevertheless, he concludes:
The crucial point is simply that worries about convergence are in both cases irrelevant to that decision. Convergence can be secured if it needs to be. Arguments from skepticism must therefore come from elsewhere.
I must admit I do not see the force of his argument here. If the only way to "secure convergence" is to radically reinterpret people's beliefs and desires, then isn't that problematic? (My problem here may be that I don't really understand what sort of "interpretation" he has in mind. Does anyone else have any idea?)
So, rather than reassuring me about the convergence of values, I'd have to say that Smith's argument has merely made me more skeptical about the possiblility of a priori truth!