I've been following Bryan Caplan's posts on analytical egalitarianism with interest, and I agree with him: the point of a model is to track truth and enforcing moral and epistemic norms is the function of institutions and culture. I'd like to point out an additional worry I have about the same bits of Sandra Peart's reply that Bryan excerpted:
Without AE, we wonder whether truth-seeking is incentive compatible…
Consider models with agents of different fixed types. Suppose a modeler proposes to pick who is in the “better” and who in the “worse” class. If the modeler can do this and policies follow from the exercise, the modeler may benefit. That’s one incentive issue. We consider rewards from both material sources and applause.
[O]nce we allow for difference to creep into the analysis, the incentives are asymmetric: the theorist gains more by showing difference than similarity.
Isn't there a weird kind of question-begging going on here? The incentive-compatibility of truth-seeking depends on how you model it. If the model recognizes the heterogeneity of motivation, then one can say that some people will opportunistically abuse the heterogeneity of models that allow it, but others won't. Peart's argument for AE turns on assuming it in her implicit model of motivation. If AE assumptions about motivation are false, the moral and/or epistemic argument for assuming them anyway falls apart. That said, even assuming rational choice homogeneity of motivation, institutions and the internalization of social norms can raise the price of defecting from truth-oriented epistemic norms making truth-seeking incentive compatible.
In one of their papers, Peart and Levy point out the anti-AE views of Edgeworth in Mathematical Psychics, which fascinated me, as it illustrates the fallacy of thinking utilitarianism is an egalitarian doctrine in which everyone is equal because everyone's pains and pleasures count equally. What counts equally are equal units of pleasure and pain. If I remember from my later reading of Edgeworth, his fundamental unit of analysis is the smallest discenible duration of experience. Each such moment of experience will have an intensity on a pain/pleasure scale. Now, Edgeworth claimed that some races have a narrower range than others. This inequality in hedonic range can easily justify all kinds of odious injustice. As Edgeworth himself says, if you've got lamps that shine bright and lamps that shine dim, and a limited amount of energy, you'll mazimize light by reserving it all for the bright.
Now, as I see it, the problem isn't Edgeworth's violation of AE, but his vulgar utilitarianism. It is an empirical matter whether inequality in hedonic capacity is true or not. It may be. But a satisfactory moral theory or theory of justice ought to be indifferent to this particular empirical contingency. That Edgeworth's theory isn't indifferent–indeed, that massive morally relevant consequences turn on it–refutes his normative theory. But if heterogenity of hedonic capacity is a fact, it's a fact, and a good model of hedonic capacity will say so.