Wednesday, September 29, 2010

Belief and Desire as Commitment

According to a common view, beliefs suffer a coherence constraint that desires do not. If I believe that p, then I'm very unlikely, at the very same time, to believe that not-p -- and if I do, that's a clear rational failing. But desiring various contradictory things is commonplace.

I don't want to dispute that the English statement of the common view just given can often express a truth. But I think it's a mistake to infer from this that there's something interestingly different between the natures of belief and desire. The words 'belief' and 'desire' are used loosely to refer to a few different kinds of things. And the ones that are of most interest in a lot of philosophy, I think, are structurally much more similar than the common view would lead one to think.

Notice, for instance, that there's a way of talking about 'belief' that does not require the coherence constraint. The way in question is a less fully-committed one than the one that is required for knowledge, or explains action. I may be said to 'believe' that p merely by virtue of finding p rather likely. But of course, there are sets of propositions, each of which I (rationally) find rather likely, but which are logically inconsistent. There's another, more committal, sense of 'belief', which does seem to enjoy something like the coherence constraint. It's this one that figures most prominently in epistemology. It's also, I think, this stronger notion that is particularly interesting in action explanation. To act is to commit oneself.

What of desire? In ordinary, loose speech, 'desires' might not amount to much. I'm said to 'desire' a third pint merely by virtue of having a felt inclination towards it. In this weak sense, we often face conflicting desires. But this might not reach the level of commitment needed to explain a decision; to explain actions, we cite 'desire' in the stronger sense of an all-things-considered desire. ("If he'd really wanted another pint, he wouldn't have gone home.") In this sense, it is a rational failure -- and a reasonably unusual situation -- to have contradictory desires. Contradictory desires commit one to preferring an inconsistent situation, just as contradictory beliefs, in the strong sense, commit one to being in an inconsistent situation.

The two weak senses described don't seem to have all that much of interest in common with one another. (The belief-like analogue of the felt-inclination sense of 'desire' seems more like intuition than like the finding-likely sense of 'belief'.)

But the two stronger senses, I think, are interestingly related. They're the things, for instance, that figure into belief-desire explanations for action.

When we discuss 'boxological' approaches to cognitive architecture, it's 'belief' and 'desire' in these strong senses that constitute interesting boxes worth theorizing about.


  1. jonathan weinberg10/04/2010 12:39:00 PM

    As one might say: there is some truth in that! But not, I think, enough truth to get you what you want. What's right is that there are surely some transient desire-ish thingies that are best understood not as being of the same sort of thing as desires. Tamar's "besires" are that sort of thing.

    But, first and foremost, that's not going to work for the class of problem cases discussed in the prior thread. In the case of the McSmartypants chair, neither the apparent desire that I have that chair, nor the apparent desire that I live in Tucson, are good candidates for that treatment. The easiest way to see this is to go off your penultimate point, namely, they are going to figure in rational explanations of my behaviors -- the former, to explain my sending off an application for the chair, my kissing up to Prof Ruddygore who it is rumored will be on the selection committee, and so on; the latter, to explain my eagerly perusing the real estate listings in Tucson, my contacting an agent there, putting my current house up for sale without looking for housing anywhere other than Tucson, and so on. Yet, again, I positively do not desire that both ((I hold the McSmartypants chair)&(I live in Tucson)).

    I do think that many shallow and transient conative states will still be well-understood as just weak passing desires, while others will be, as you say, well-understood as some other creature entirely. If I happen to be walking by the fridge, and the possibility of having a beer comes to mind, and I pause, quickly check my watch calendar for any even vaguely pressing commitments, shrug, and then pull out a beer -- it might have been that even the slightest countervailing desire would have kept me going past the fridge, but lacking such, I got a beer, and my doing so is eminently explicable in propositional-attitude terms.

    Finally, though this is orthogonal to the particular debate here, I do think it is a mistake to run together the PA-as-commitment model with the PA-as-symbol-in-box model. For well-ordered, non-wanton minds, they should overwhelmingly tend to line up. But they can and do come apart from each other, as can perhaps most easily be seen in cases of strongly felt but reflectively disowned desires. (E.g., as in a homosexual person who has the severe misfortune to have been brought up a fervent Baptist.)

  2. Thanks, Jonathan. I agree that my way of putting things isn't decisive against what you've just said, but I do think there is considerable latitude, on a approach like mine, to avoid these results. Here's another way of describing the situation. (I actually think the one I'll ultimately want to endorse isn't this, but rather a more complicated one. But this one illustrates the flexibility.) The desire box is about all-things-considered desire; sometimes we use 'desire' more loosely to mean something like inclination. These inclinations are sometimes, but not always, shallow and transient.
    What do you all-things-considered desire? Let's suppose what you really ultimately want is to hold the McSmartypants chair. So that goes in the desire box, and goes into explaining your kissing up, etc. On my view, then, the unqualified 'live in Arizona' is not in the desire box, although of course there's some sense in which you desire that. The thought could be something like, you desire [you live in Arizona if you don't get the McS chair]. This desire combines with your beliefs about the likelihood of the chair, and how to go about living in Arizona, explains your real estate decisions.
    Another way to go -- probably my preferred one when all is said and done -- will make all desires (and beliefs, for that matter) relative to background assumptions. So it's not just propositions that go in the boxes, but something like ordered pairs of propositions and sets of worlds, or properties picking out sets of worlds, or something like that. Then you'll need to define coherence for these ordered pairs, but that's probably ok.

  3. I think the last paragraph of your comment is interesting, Jonathan, but I don't quite get it. What are you supposing is right about the unfortunate repressed individual? I see four possibilities, none of which seems obviously right or wrong, depending on the case. (i) the sexual desires felt and the chastity desires are endorsed and thus constitute commitment; the desire box contains chastity desires; (ii) just the reverse; the desire box contains sexual desires; (iii) both desires are felt but neither is endorsed; the subject hasn't made up his mind which way to go, and neither desire gets in the box; (iv) both desires represent commitments -- incompatible ones -- so the subject exhibits practical irrationality.
    All of these look possible on my view, and each seems like a plausible case. But it sounds like you think I'm forced into a particular treatment, and that its untenable. Can you say why?

  4. jonathan weinberg10/05/2010 11:35:00 AM

    re: 2 - I don't find the idea of a desire that if-not-McSmartypants-then-Tucson particularly intelligible. What would make sense to me would be something like: if(f) I believe that I am not going to get the McSmartypants chair, then I will desire that I live in Tucson. But that's constructed from two different PAs, not one PA with a conditional structure, as you've suggested. (As noted on a previous thread, it's rather more part-and-parcel of what desires do that they change in response to our beliefs, as opposed to having a particularly robust internal dynamic of their own.)

    I also suspect in general that this kind of conditional structure will get totally computationally intractable once I have more than about 5 or so different goals in any sort of complex tension with each other. I mean, my total desire set also can be called upon to answer hypotheticals like: what would I do if I don't get the McSmartypants chair, and the AZ job falls through? Probably happily stay here at IU, but then what would I do if _that_ ceased to be a possibility? And so on. We can _work out_ those scenarios based on our current desires, but your approach seems to require them to be pre-worked out, and on what basis, it is hard to see. We have in our noggins lots of local assignments of different amounts of value to different outcomes, without generally having them represented in the kind of nested-within-nested-within-nested conditionals that your approach will require. Remember, although the "box" is a basically a functionalist metaphor, the internal structure of the representations is supposed to be real.

    re: 3 - what I had in mind is someone who clearly has desires that they just as clearly are committed to not having. I have trouble seeing how a person who has strong homosexual desires (not just fleeting weak impulses), but who fervently wishes he did not, is maybe taking active steps that he thinks will reduce those desires, etc. can count as either (i) not having those first-order desires, or (ii) being _committed_ to those first-order desires. They are anti-committed to those desires! So we need our cognitive psychology to allow for desires that are not commitments.

  5. On the first point: I don't see the problem. You say you don't know how to make sense of desires with conditional contents, but I don't see why. Most of us think that the sentence I used, "you live in Arizona if you don't get the McS chair" expressed a proposition. Now stick it in the desire box; what's the problem?
    On the second point: 'commitment' as I use it is something of a technical term. I don't mind the way you've described the situation, but I don't think it maps on exactly to the way I'd be using those words in the framework gestured at above. Remember, I think there's a perfectly good sense of 'desire' that doesn't entail the all-things-considered desires that figure into psychological explanations. I don't see why the homosexual desires in question couldn't be among this sort.

  6. jonathan weinberg11/20/2010 05:13:00 PM

    (Sorry, I had forgotten about this thread!)
    The problem isn't with the very idea of a logically structured member of the desire box. Rather, I'm having trouble concocting any sort of even vaguely plausible mental life in which _that_ conditional is one that would be desired. The overall conative state that it most closely resembles is one that is more naturally represented in terms of a mix of belief states and desire states. To restate a point from my previous comment: having to construct these complex conditionals is just not a good way to implement keeping track of our rankings of various possible outcomes for the world.

    I guess I just don't know quite what this technical notion of "commitment" is. We agree that some very transient, impulse-like, "besire" states may be usefully distinguished from desires proper. But the man in my story's desire for other men (let's stipulate that it's Larry Craig) isn't like that. Note that this state will play exactly the role in psychological explanations that you have in mind. I.e., he arranged his flights to connect through Minneapolis so that he would have a chance of having sex in the men's room (even though he hates himself for not just wanting to do so but acting on it). Sure looks like a desire in that more robust sense.