I'm curious: Do you think that my argument shows that consequentialists need to deny that value can depend on immorality? If so, that alone is an interesting result. For one thing, it may mean that consequentialists are committed to deny that a person can value immorality for its own sake. I wouldn't have expected that result.
I'm finding the entire discussion very interesting! But I agree that this result, if valid, sounds very surprising.
Firstly, I should point out that the problem only gets off the ground if we accept an objective desire-fulfillment theory of value. More subjective accounts, such as hedonism, would avoid these problems. So denying DF is an easy way out for consequentialists. To make things interesting, I will make two assumptions for the sake of this post: (1) that value can be identified with desire-fulfillment; and (2) that the morally right action is that which maximises value (desire fulfillment). But for the record, I do not subscribe to (2) exactly as stated. Now...
Thinking back to the liar paradox, it seems to have precise rather than general consequences. That is, there are certain self-referential sentences we must not allow, but other forms of second-order statements - even claims of falsehood - are unproblematic (e.g. "The statement 'grass is blue' is false").
I think the same applies to desires. It's fine to desire that other desires be thwarted, so long as you don't include the present desire. Recall the 'Adolf' example I presented as a (desire-based) variant of the liar paradox:
Adolf desires that more desires be thwarted than fulfilled. [A 'desire that P' is 'fulfilled' if P is in fact true, 'thwarted' if P is false.] Further suppose that - apart from this desire of Adolf's - there are in total an equal number of fulfilled and thwarted desires in the world. It follows that Adolf's desire is fulfilled if and only if it is not fulfilled.
The problem with Adolf is that he was put in a context where his desire became self-referentially paradoxical. This situation could potentially arise whenever you have a desire that refers to 'all desires'. But it's logically impossible that a desire could have its own thwarting as a fulfillment condition - that's equivalent to a statement having its own falsity as its truth condition! So I think we must hold all such desires to be impossible. (It would seem odd to say they are possible some times but not others, depending on the external context. Why should the existence of an internal desire depend on the external world? But maybe this could be argued as an alternative solution. [Update: I do so here.])
One might plausibly claim that morality is about all desires. If this were so, then the above considerations suggest that desiring immorality would be impossible. But perhaps this is not so. We can avoid self-referentiality by having a desire that refers not to 'all desires', but rather to 'all other desires'. So we might refine our understanding of morality so that, if appealed to in the content of a desire, it is instead understood to be about all other desires. This would solve our problems. According to this definition, we can value immorality without fear of logical contradiction.
If you don't like to redefine morality in such a way, then let us refer to my new (restricted) concept as 'schmorality'. I would then conclude that it is possible to value schimmorality, but logically impossible to value unrestricted immorality. This latter result sounds surprising, but I think it is actually fairly insignificant, because schimmorality is so similar that it can do all the work we previously required of immorality.