Tuesday, July 14, 2009

The Smallish Difference Between Belief and Desire

In the usual taxonomy of mental states (usual, that is, among contemporary analytic philosophers of mind) belief is one thing, desire quite another. They play very different roles in the economy of the mind: Desires determine our goals, and beliefs determine the perceived means to achieve them, with action generally requiring the activation of a matched belief-desire pair (e.g., the belief that there is beer in the fridge plus the desire for beer). I confess I’m not much enamored of this picture.

Surely this much at least is true: The belief that P is the case (say, that my illness is gone) and the desire that P be the case are very different mental states -- the possession of one without the other explaining much human dissatisfaction. Less cleanly distinct, however, are the desire that P (or for X) and the belief that P (or having X) would be good.

I don't insist that the desires and believings-good are utterly inseparable. Maybe we sometimes believe that things are good apathetically, without desiring them; surely we sometimes desire things that we don’t believe are, all things considered, good. But I’m suspicious of the existence of utter apathy. And if believing good requires believing good all things considered, perhaps we should think genuine desiring, too, is desiring all things considered; or conversely if we allow for conflicting and competing desires that pick up on individual desirable aspects of a thing or state of affairs then perhaps also we should allow for conflicting and competing believings good that also track individual aspects – believing that the desired object has a certain good quality (the very quality in virtue of which it is desired). With these considerations in mind, there may be no clear and indisputable case in which desiring and believing good come cleanly apart.

If the mind works by the manipulation of discrete representations with discrete functional roles – inner sentences, say, in the language of thought, with specific linguistic contents – then the desire that P and the belief that P would be good are surely different representational states, despite whatever difficulty there may be in prizing them apart. (Perhaps they’re closely causally related.) But if the best ontology of belief and desire, as I think, treats as basic the dispositional profiles associated with those states – that is, if mental states are best individuated in terms of how the people possessing those states are prone to act and react in various situations – and if dispositional profiles can overlap and be partly fulfilled, then there may be no sharp distinction between the desire that P and the belief that P would be good. The person who believes that Obama’s winning would be good and the person who wants Obama to win act and react – behaviorally, cognitively, emotionally – very similarly: Their dispositional profiles are much the same. The patterns of action and reaction characteristic of the two states largely overlap, even if they don’t do so completely.

This point of view casts in a very different light a variety of issues in philosophy of mind and action, such as the debate about whether beliefs can, by themselves, motivate action or whether they must be accompanied by desires; characterizations of belief and desire as having neatly different "directions of fit"; and functional architectures of the mind that turn centrally on the distinction between representations in the "belief box" and those in the "desire box".


  1. Another way to generate the kind of dispositional overlap you describe in your penultimate pgraph is with certain meta-attitudes. Compare S's believing that he desires that P with S's desiring P. Or compare S's desiring that he believe P with S's believing P. Here the overlap's to be expected given the similarity in the satisfaction conditions of the *particular* attitudes in question. But it's not clear that this can serve as much of a basis for skepticism about differentiating beliefs and desires *in general*.

    I'm sympathetic to this kind of skepticism, just not sure that these sorts of overlaps constitute much of a case for it.

  2. A very helpful post. And I am in agreement - but for one point.

    If the best ontology of belief treats as basic the dispositional profiles, then we no longer have reference to any particular Belief at all - so I continue to think. On the dispositional account, it is no surprise we distinguish with difficulty a belief from a desire, since we have lost what we treat as 'basic' about Belief in the first place: the disposition to assert 'that p', where 'that p' becomes the pre-existing standard we use to check our behavioral dispositions - what we 'really belief', what I call our D-Beliefs. For example: I sincerely 'believe that' all people are equal, yet I unwittingly desire to discriminate, consistent with my possession of D-Beliefs that accord with the conflicting Belief - 'belief that' - all people are not equal. I do not believe that people are not equal. I rather very ardently believe that people are equal. Yet, I do appear to D-Believe that people are not equal. As for what I take my desires to be: haven't we always assumed the human heart to be a dark mystery to its owner? This is perhaps one psychological point the introspective puritans got right.

    Perhaps this gives us one way to distinguish a belief from a desire: a belief gains an identity through the linguistic expression 'that p', whereas desire is something we associate with images, sensations, urges, emotion, mood, or biological need. Michelle Montague ('Against Propositionalism' - was forthcoming in Nous couple years back) wants to say that objectual attitudes hold "simply between thinkers and non-propositional objects, rather than between thinkers and propositions," but I do not mind merely replacing the belief box full of linguistic-like contents with these other mediating sensual mental states and simulations.

    You Write: "Desires determine our goals, and beliefs determine the perceived means to achieve them, with action generally requiring the activation of a matched belief-desire pair."

    Corollary: So, beliefs would be perceived means to achieve our goals, whereas knowledge would be our actual means (capacity) to achieve them?

    You write: "are the desire that P (or for X) and the belief that P (or having X) would be good."

    This alternative language seems revealing: Desire 'for X' versus Belief 'that having X would be good'. Our reference to Belief is fixed in the linguistic statement that P whereas our reference to Desire is not.


  3. Pete: With the meta-attitudes there may especially be dispositional overlap between believing P and believing that one believes P -- enough, I think, that if one believes that one believes that P, one necessarily is at least in an in-between believing state with respect to P (though sometimes *only* in-between).

    I'm less clear, Pete, about your final remark. What kinds of cases do you think constitute a good case for skepticism about the differentiation of belief and desire?

  4. Thanks for the comment, Michael. Your criterion for belief in terms of a disposition to assent to (if I'm understanding you aright) a linguistic expression of the form "that P" seems to me too cognitive, too linguistic. I don't want to say that the implicit racist determinately believes that all the races are intellectually equal, regardless of what he may linguistically avow. He is, I'd say, in an "in-between" state. You can cover such cases by saying he believes one thing and D-believes another, so in some sense the dispute is linguistic. But I don't think that makes it unimportant: It matters what set of phenomena we privilege with the important philosophical and folk label "belief"; and I suggest that we should privilege the overall arc of action and reaction rather than simply the linguistic avowal part.

  5. Thanks Eric. To me this is not a matter of privileging one set of phenomena over another. Rather, I think we must begin - and I argue that we do begin, however unwittingly - with a semantics of Belief as intrinsically linguistic before moving on to address the overall arc of action. As with all the terms we use in philosophy and science, I find an important distinction between the original natural language and the developed ad hoc technical language. And I claim that the natural language of 'belief' is grounded in linguistic statements of fact. We would not be talking about 'belief' to begin with, but something else altogether, if we were not first talking about belief 'that p'. We do not have a 'belief' - in prototypical, natural usage - unless we have a belief 'that p', where 'that p' is a statement of fact. And without a belief in prototypical natural usage, then we have lost any way to reference D-Belief (including Belief as reference to an overall arc of action). Without the semantically basic, linguistic belief 'that p' there is nothing for a given D-Belief to accord or not accord with. For example, there is no in-between believing about the color of grass unless we locate first the linguistic belief 'that grass is green' and then locate within the arc of action a D-Belief that we can reference only by the statement of fact 'grass is not green.' But as we seem to agree, the D-Belief is not linguistic in nature, which means that our reference to this D-Belief depends on the conception of the linguistic belief 'that grass is not green'.

    This is my attempt at illuminating just how dependent we are on our natural language when addressing the unknown sophistication of the unconscious mind, which I can only suspect will help carve off mistakes that would otherwise be made down stream. In fact, I wonder if we are left with discussing either the literal linguistic beliefs that p of S or the metaphorical beliefs of S's unconscious mind. Just as my unconscious mind 'binds', locates', and 'tracks' an Object (language grasped through metaphorical comprehension), so does my unconscious mind 'really believe' that all men are not equal.

    Yes? Or . . .

  6. Eric, I wouldn't hang it on just a few cases, but on larger explanatory and theoretical considerations. I tend to be convinced by the general Churchland/Stich/Dennett network of considerations re eliminative materialism.

  7. Michael: I'm not so sure you're right about natural language. After all, we naturally say that the dog thinks the cat ran up the tree though the dog cannot express herself linguistically. I believe natural language here is complex and multi-textured, admitting of rigorization in a variety of different ways -- and it then becomes a pragmatic or political choice what rigorization to prefer. But even if I'm wrong about this, I don't see the need to hew closely to ordinary language. Why adopt that principle?

  8. I see, Pete. Although I'm not hostile to eliminativism, I think certain terms are more valuable than others and that "belief" and "desire" might be salvageable with some rigorization -- at least for the foreseeable future. So then my case for overlap turns on the dispositional nature of the rigorization and the overlap in the plausibly associated dispositions between desires and certain corresponding beliefs.

  9. Eric, I would say that the dog 'thinking' is an example of taking the intentional stance. I have spent some time attempting to demonstrate that Dennett's intentional stance is actually discreet acts of explanation via metaphor. Although, the dog example is difficult, since dogs are so much like us. A science of dog cognition, dog behavior, and dog brains would need to have some level of self-consciousness about our natural language about what dog's 'think' before offering explanations for what, precisely, this thinking is. At the end of the day we might be best served with D-Think, E-Think, and N-think modules, all having little to nothing to do with conscious, human thinking.

    As for preference: my position is that what we are really after is a science of D-Belief, which is not linguistic in nature. So we agree: away with the belief box and language of thought. I want to understand, however, why we are talking about 'belief' in the first place, as well as how we are indeed trapped in employing natural language of linguistic belief 'that p' in referencing non-linguistic D-Belief. In your essays, you are forced to hew closely to linguistic Belief 'that p' in order to reference D-Belief as well as create scenarios of in-between believing.

    I would be interested to see how you might characterize differently the analysis above of the in-between believing that grass is green.

  10. Michael: Okay, I think I understand better. I'm not sure I understand your last question. Dogs are not disposed to say "grass is green", but they might be disposed to respond A-ishly to grass if they are in general rewarded for responding A-ishly to green things. Is that the issue you have in mind?

  11. Eric, This is not exactly what I had in mind, but this is perhaps better. Since the dog is not linguistic, could we not just consider conflicting behavior with respect to the color of grass as indeterminate, rather than an in-between state? Perhaps some modules are tracking the greeness of grass and some simply are not; I assume there are no dog faculties 'aimed at truth'. Without the conception of linguistic belief 'that p', it is not clear that we will discover any overall 'commitment' of the dog's faculties to green information pertaining to the grassy stuff as might be salient to an Arc of action. Without the belief 'that p' association, then it seems we would be up the same creek with humans too.

  12. Michael: In-betweenness, in my view, is a form of indeterminacy between "believes that P" and "does not believe that P" -- so do we disagree? I do think that there's compelling *practical* reason to ascribe propositional attitudes to non-human animals if one is interested in characterizing their behavioral patterns, and thus to resolve the indeterminacy, as it were, in favor of the dog.

  13. Eric, I think we agree much more than we disagree. But: While an intentional stance can be a helpful 'way to talk', digging into a more scientific, explanatory practice seems to require something different. For example, I argue that the behavioral patterns of Dennett's virus are explained only through discreet metaphors, regardless of whatever global intentional realism motivates the research of the scientist.

    Indeterminacy might have been a poor word choice. What I mean to say is that there is no question to begin with as to whether the dog's cognitive system tracks the grassy stuff as green in a believing way since there is no question as to the dog's linguistic stance to the statement 'the grass is green'. In-betweenness arises from a tension between behavior we would expect from a rational, human agent that linguistically believes 'that grass is green' and the behavior we would expect from an agent that linguistically believes the opposite. If there is no linguistic belief 'that P' to begin with, then there cannot be a D-Belief tension between two different beliefs that p. Belief is linguistic at the semantic root (whereas desire is not). So it still seems to me . . .

  14. Be careful of the semantics here.
    A ‘belief’ is not the same as a ‘thought’. We are not aware of most of our beliefs while, on the other hand, our thoughts are subject to logical analysis. Aboriginals (for example) may not have heard of Newton but believe in the concept that we call gravity, i.e. that an arrow will fall and probably not drift upwards to become part of a cloud.
    Most of our beliefs haven’t been expressed with language. Pre-verbal children have beliefs. Most of us don’t know what we believe about ourselves, i.e. are we fundamentally a good person?
    Often, beliefs are nothing more than affirmations. such as a statement such as “I believe in God” (you may, but that is different than this affirmation).
    Beliefs are primarily the product of experience; thought may or may not be logical, but is subject to analysis and is dependent on language (whether spoken or not)Reference ”How Expectancies Shape Experience” by Kirsch, 1999.

  15. I need to correct my wording - I meant to say that beliefs are usually thought of as nothing more than affirmations such as “I believe in God”.

    A child’s pre-verbal mental ‘internal working model’ (IWM) of the world which he could not put in words (at least not by the child) is an approximation of a belief – but not a thought because it is not consciously ‘thought’ about.

    This forum seems to consider the philosophical perspectives when ‘science’ has caught up and surpassed much of this thinking – such as philosophizing about whether the earth is flat.

  16. Thanks for the comment, Jack/anon! I agree with you about that. (See for example my earlier post on whether people really believe in God, here.)