1. The Paradox Stated
Actualist preferentism is a theory of welfare: a theory that says what it is for someone's life to go well for her. The theory's basic idea is that getting what one wants makes one's life go better. Like any theory of welfare, this one faces problem cases: cases in which someone's desires are satisfied but, intuitively, they are not made better off (or vice versa). But in addition to these problem cases, preferentism faces the paradox of desire. In a nutshell, this objection to preferentism goes like this: I can certainly desire to be badly off. But if a desire-satisfaction theory of welfare is true, then--under certain assumptions--the hypothesis that I desire to be badly off entails a contradiction. So desire-satisfaction theories of welfare are false. (1)
But this argument does not, in fact, establish that preferentism is false. There is a way to formulate preferentism so that the hypothesis that I desire to be badly off does not entail a contradiction. My aim is to show how this version of preferentism avoids paradox. Before I proceed, though, I need to state the version of preferentism that is the target of the argument, and spell out the argument in more detail. I will start with the first task.
On the standard atomistic version of actualist preferentism, the "atoms" of welfare are episodes of intrinsic desire satisfaction: episodes (stretches of time) during which the subject has a intrinsic desire that P, and it is in fact the case that P. (2) (An intrinsic, or non-instrumental, desire is a desire that one does not have merely because satisfying it is a means to satisfying some other desire.) "Episode of intrinsic desire frustration" is defined similarly. The intrinsic value of an episode of desire satisfaction (or frustration) is equal to the intensity of the desire times the duration of the episode. Desire satisfaction has "positive" value and desire frustration has "negative" value, so the value of someone's life for her is just the sum of the values of the episodes of satisfaction, minus the sum of the values of the episodes of frustration.
For our purposes, more important than how the theory calculates the welfare level of someone's entire life is how the theory calculates someone's welfare level at a particular time. Here is what the theory says: the value of someone's life for her at a time t is just the net amount of desire satisfaction that occurs in her life at t. Since t is just one instant, we can ignore the durations of the episodes of desire satisfaction and frustration. So the value of someone's life for her at t is just the sum of the intensities of the satisfied desires she has at t, minus the sum of the intensities of the frustrated desires she has at t.
(Actualist preferentism is not the only form of preferentism. Ideal preferentism says, roughly, that it is the satisfaction of the intrinsic desires you would have, if you were to undergo some form of "cognitive psychotherapy" (you were thinking more clearly, you knew all the relevant facts ...), that contributes to your welfare. While many people accept some form of preferentism, ideal preferentism is probably more popular than actualist preferentism. (See Kagan (1998: 38) for a brief survey of the reasons.) And it looks like the paradox does not arise for ideal preferentism: it might be that if I were to undergo cognitive psychotherapy, I would not desire to be badly off. But I agree with Heathwood (2005) that none of the standard arguments for preferring ideal to actualist preferentism are any good. Since actualist preferentism is the better theory, defending it against the paradox of desire is all the more urgent.)
Now to present the paradox. The paradox of desire arises in the following kind of situation. Suppose I have several intrinsic first-order desires at t--like a desire for a cold beer, a desire for some salty peanuts, a desire for warm weather. (They are "first-order" because they are not desires about what desires I have, and are not desires about my level of well-being. …