Consequentialists are comfortable evaluating whether one's character is "morally fortunate" (beneficial). But people commonly evaluate characters on other dimensions also. To see this, imagine that a Trickster God sets things up so that the best consequences result from vicious desires (e.g. to gratuitously harm others). So it might be good to make ourselves bad. For this claim to be coherent, we must be talking about two different kinds of evaluation: it might be fortunate (rather than harmful) to make ourselves vicious (rather than virtuous). This disambiguation resolves the apparent contradiction. But it raises a new question: how should consequentialists analyze virtue and vice, as distinct from un/fortunate character? I think of virtue as "moral rationality", but this requires some unpacking.
As a first pass, we might think that virtue consists in aiming at or desiring the good. But do we mean 'the good' de dicto or de re? These interpretations come apart if someone has false normative beliefs (consider Huck Finn, who thinks he's morally required to report runaway slave Jim). I'm with Arpaly in thinking that it reflects better on the agent to act against their false normative beliefs ('crazy notions') in such cases. Since Huck Finn is moved by what is in fact morally important (namely, Jim's humanity), he thereby shows himself to be of virtuous (admirable) character, and his acts performed from this disposition of character are thereby praiseworthy.
So, as our second pass: virtue consists in concern for the [actual, de re] right-making features (or what I long ago called 'partial reasons'). This still isn't quite precise enough, since (as Liz points out) we can distinguish 'surface' and 'deep' right-making features. Most moral theories will agree that the "surface right-making features" include, say, respecting autonomy, keeping promises, helping others in need, etc. These are all things that can - all else equal, or in the appropriate combination - make an act right. But different moral theories will diverge in their underlying explanations of why these features are right-making: whether it is because they tend to promote utility, or respect the Categorical Imperative, or whatever.
This raises the question: does virtue require a concern for the ultimate right-making feature (so that, if utilitarianism is true, whole-hearted Kantians can't be virtuous)? Or is it enough to get the surface features (including their relative weightings, etc.) right? The former option seems excessively strict. But it is tricky to come up with a principled reason to prefer the latter.
One possibility is that we should think of the deep feature ("being utility-maximizing") as really a kind of higher-order property: it is merely alerting us to the fact that the act in question has other (surface) features in virtue of which it maximizes utility. The claim that φ-ing maximizes utility doesn't add anything further to the claims that φ-ing helps Bob, harms no-one, and has no significant opportunity costs. If surface features have explanatory priority in this way, that could explain why it is that proportionate concern for them, rather than the deep feature as such, is what's constitutive of good will (virtue).
I'm not sure how plausible that particular attempt is (feedback welcome!). But in my next post I'll explain how utilitarians had better be able to establish this result somehow or other, because it's essential to their theory's ability to get the right results on questions of praise/blameworthiness.