Tuesday, September 18, 2012

Against Architectural Accounts of the Attitudes: Two Thought Experiments

Most recent Anglophone philosophers appear to favor what I will call architectural accounts of the attitudes. On such accounts, what is essential to possessing an attitude is that one be in some particular type of physical or biological state or that one possess some particular piece of cognitive architecture. On such an account, to have an attitude such as belief or desire might, for example, be to possess an internal representation of a certain sort, perhaps poised to play a particular cognitive role; or it might be a matter of being in a brain state of a certain sort. On my view, in contrast, such architectural facts should be regarded as matters of implementation only, not essence. What matters to having an attitude is instead, I suggest, that one live a certain way -- that across a wide range of actual and counterfactual circumstances, one is disposed to act and react, both inwardly and outwardly, in patterns that we would folk-psychologically tend to regard as characteristic of someone who possesses that attitude.

Call whatever architectural condition is essential to having Attitude A, on one of the architecturally-based views, Architectural Condition C. Unless Architectural Condition C just is the condition of being disposed to act and react, inwardly and outwardly, in the pattern characteristic of Attitude A, then presumably it is conceivable that Architectural Condition C could be possessed by a person who lacks the such a suite of dispositions, or such a suite of dispositions could be possessed by a person who lacks Architectural Condition C. What should we say about such cases? Let’s consider two.

One: Andi, let’s suppose, is in Architectural Condition C for the belief that giraffes are born six feet tall. Colorfully, we might imagine that a 22nd century brain scanner finds in her Belief Box a slip of paper containing the sentence, in the Language of Thought, “giraffes are born six feet tall”. Or maybe the giraffe neuron is linked to the six-feet-tall neuron is linked to the birth-size neuron. Despite this architectural fact, however, Andi is not at all inclined to act and react in the usual way. She is not at all disposed, for example, to say that baby giraffes are six feet tall. If asked explicitly, she would say giraffes are probably born no more than three feet tall. If shown a picture of a giraffe as tall as an ordinary man she would assume it’s not a newborn. If a zookeeper were to tell Andi that giraffes are born six feet tall, Andi would be surprised and would say, “Really? I had thought they were born much smaller than that!” And so forth, robustly, across a wide range of actual and counterfactual circumstances. None of these facts about Andi are due to the presence of weird factors like guns to her head or manipulation of evil neuroscientists or a bizarre network of other attitudes like thinking that “three” means six. (See also my post on Mad Belief.)

Two: Tomorrow, aliens from Beta Hydri arrive. The BetaHydrians show all signs of valuing molybdenum over gold. They will trade ounce for ounce, with no apparent hesitation. When they list metal prices in their currency, they list the price of molybdenum higher than the price of gold. They learn English, and then they say things like “in BetaHydrian culture, molybdenum is more valuable than gold.” And so forth. Suppose, too, that BetaHydrians have conscious experiences. There is a kind of swelling they feel in their shoulders when they obtain things for which they have been striving. They translate this feeling into English as “the pleasure of success”. They experience this swelling feeling when they successfully trade away their gold for molybdenum. Like us, they have eyes sensitive to the visible spectrum, and like us they have visual imagery. They entertain visual imagery of returning to Beta Hydri loaded with molybdenum and of the accolades they will receive. Pleasurable feelings accompany such imagery. They plan ways to obtain molybdenum, at the cost of gold if that’s what it takes. They judge other BetaHydrians’ molybdenum-for-gold trades as wisely done. Etc. Ordinary people around Earth find it eminently natural to say that BetaHydrians value molybdenum over gold. But we know nothing yet about BetaHydrian biology or cognitive architecture, except that whatever it is can support this pattern of action, thought, and feeling. Whatever Architectural Condition C is, if we can coherently conceive its coming apart from the dispositional patterns above, the patterns characteristic of valuing molybdenum over gold, then suppose Architectural Condition C is not met. If we may conceive the physically impossible, we might even imagine that the BetaHydrians robustly, intrinsically, durably, and non-accidentally exhibit these behavioral and cognitive and phenomenological patterns, across a wide range of possible worlds, despite being made entirely of undifferentiated balsa wood. (See also my post on Betelgeusian Beeheads.)

If we are at liberty to choose an approach to the attitudes that is practically useful and that gets right what we care about in ascribing attitudes, we should choose an approach that says that BetaHydrians value molybdenum over gold and that Andi does not believe that giraffes are born six feet tall. The lived patterns are what matters, not, except derivatively, the underlying architecture.


Scott Bakker said...

Warming up for another debate with Carruthers? ;)

The big defence for ‘architectural accounts’ is that this is simply the way cognitive business is done in the life sciences. Explanation in the life sciences is mechanistic explanation (Bechtel’s book on this is good), and given that many of the cognitive sciences are also life sciences, it seems strange to demand that an entirely different explanatory paradigm be used. What you really need is an explicit argument for why philosophical explanatory practice should not conform with scientific explanatory practice.

You might think that you have a strong case for ‘mechanism skepticism,’ and that epistemic scruples warrant abandoning mechanistic approaches, but note how different this argument sounds depending on *whom* you are talking to. So for instance, it has to sound plausible to those, such as yourself, who are concerned with the strength of their epistemic commitments relative to the rational and evidential warrant for those commitments. But what about a functional neuroanatomist? As the life science of neural architecture, functional neuroanatomy *obviously* falls outside the purview of your argument, begging the question: Just who are you arguing to?

It can’t be “Anglophone philosophers” in general, since a good number of them see themselves as ‘prescientific prospectors,’ you could say, moving out in advance of ongoing research in ongoing sciences. As the saying goes, ‘experiment without theory is blind,’ and to this extent their contributions are plausibly warranted. I’m not sure a ‘cognitive architect’ like Carruthers couldn’t just shrug their shoulders, and say that, as you put it, “an approach to the attitudes that is practically useful and that gets right what we care about in ascribing attitudes” is just not what his bunch are interested it. The dispositional domain you outline is an important data resource, certainly, and should doubtless trump ‘architectural fables’ when engaged in daily problem-solving, but it is incapable of providing scientific, as opposed to merely pragmatic/interpretative, explanations. And producing the former requires formulating mechanistic hypotheses - which is just to say, being wrong more often than right.

You gotta break some cognitive eggs to make a scientific omelet!

(Just a note, Eric. Machines Like Us posted the wrong source link for their repost...)

Elliot Carter said...

Interesting post!

I'm wondering whether it is question-begging in this context to assume that architectural conditions and dispositions to act (inward and outward) can come apart in the way you describe. In the Andi example, you claim that she has a mentalese sentence in her belief box, "giraffes are born six feet tall" but that she acts in all ways and all circumstances as if she doesn't believe giraffes are six feet tall. You stipulate that nothing "weird" accounts for these facts about her. But presumably the proponent of one kind of architectural view will insist that there is something weird going on in this case: namely, that Andi's cognitive architecture is weird. Her belief box isn't connected in the appropriate way to her thoughts and behavior. So, depending on the architectural view in question, Andi may fail to satisfy the architectural conditions for having the belief that giraffes are born six feet tall.

It seems to me that one can tie dispositions into the architectural conditions -- i.e., the belief must be related in the appropriate way to the faculties centers responsible for acting, introspecting, etc. This view seems to me to have the advantage of preserving the intuition that an attitude is a kind of state that a mind can be in (rather than a set of dispositions) but allowing for the possibility that the dispositions are still necessary for having the attitude.

Eric Schwitzgebel said...

Thanks for the thoughtful comments, Scott and Elliot!

@ Scott: Indeed, I do have Carruthers specifically in mind. We are both working up contributions to an edited volume on belief and I recently read his draft contribution which is critical of my work on broadly the grounds you describe. My general answer is: It's terrific for scientists to do this sort of thing, but it's too early to treat scientific models as getting at what is humanistically important in attitude ascription.

@ Elliott: If it is not conceptually possible for the underlying mechanism to come apart from the actual and counterfactual outwardly lived pattern, then I'm happy enough with identifying the belief with the mechanism and see doing so as a kind of metaphysical sleight-of-hand that might serve certain purposes (such as fitting better with certain claims about causation). However, if it is conceptually possible (even if nomologically impossible!) for the two to come apart, then we can consider cases like the ones above as diagnostic of what is essential. At least that's my thought, with a nod to the complexity of issues around conceivability, modality, and essentiality.

Scott Bakker said...

I know I was roundly dissappointed by his criticisms of your arguments regarding introspection in Perplexities. Is that what you're referring to?

"My general answer is: It's terrific for scientists to do this sort of thing, but it's too early to treat scientific models as getting at what is humanistically important in attitude ascription."

But this *seems* to be a far different argument than the one you make in your post. And I'm still not sure he'll be able to make hash of your claim, short of specific examples where he and others 'over-reach' in this manner. Or failing that, some general prescriptive argument for practicing 'humanistic' philosophy - something which strikes me as a tall order.

For my part, 'subpersonalism' troubles me whenever it assumes (as behaviourism once did) that it has somehow immunized itself against the problems pertaining conscious experience. You find - or at least I find - that 'global broadcasting' gets parachuted in at awfully convenient junctures.

Given the 'liberal' in your brand of dispositionalism, it strikes me that you're pressing in the same general direction here. I'm wondering if the 'subpersonal' isn't what you want to focus on as opposed to architectural/mechanistic. The latter as so basic that you'll find yourself mired in clarifications. 'Subpersonalism,' on the other hand, really marks something strange: Persons, informatically encapsulated in all the ways your liberal dispositionalism describes, postulating subpersonal mechanisms, and claiming that their inevitable reference to the personal constitutes 'facon de parler.' There does seem to be a problematic fiction involved, a wilful denial of the actual informatic provenance of their models.

Suddenly this feels unclear to me. Do you see what I'm saying? Maybe instead of framing it as an 'appropriate spheres of discourse' argument, pitch it as a 'full epistemic disclosure' argument. The problem with the former is that your charge of 'too early' is too easily blocked by simply scaling back on epistemic commitment. I can almost hear him! "What? Too early too dare dream, Eric?"

Charles T. Wolverton said...

Eric -

I agree with your "live a certain way" view of a PA (AKA, a family of context-dependent behavioral dispositions). I also subscribe to the view that notwithstanding its problems, the intentional idiom is a useful shorthand when treating an organism as human-like. And given that PAs seem typically undefined (although apparently often thought of as being free-floating entities something like the contents of Andi's "belief box", I'm all for defining them explicitly along the lines you suggest. But I don't understand why you frame these positions in terms of an opposition to architectural accounts and implementations. Could you elaborate just a bit on your motivation for doing so?

I seem to agree with Scott on some related issues, so perhaps I'm just ineptly restating his last comment. He uses a vocabulary that - due to my sparse background in this material - is often over my head. If so, apologies.

Charles T. Wolverton said...

On further reflection, for my question to be answerable probably requires some elaboration.

I view the "mental" from a system engineering perspective, which starts at the top and works down. So, my first question is "what is the system (here, the organism) trying to do?" And my Rortian answer is "Cope with the environment", which can be elaborated as "develop dispositions to respond to various combinations of internal state and external stimuli" - AKA, to "live a certain way" in a dynamic environment.

As a next step, one could group those responses according to common features of internal states, external stimuli, or combinations thereof. Eg, one could define classes of responses - call them "bleefs" - the members of which have in common that each is consistent with the organism having a specific assumed relationship with the external environment. Eg, the organism might respond to the presence of some other type of organism by trying to escape from it in various ways depending on the environmental context. That bleef might be labeled "avoid X's". (But see note below.)

In any event, a natural next question is "where and how are these dispositions implemented?" This seems the point at which architecture enters the picture. Presumably, it is generally agreed that the answer to the "where" part is "in the brain", giving us a first baby step in defining an architectural structure. Consensus on detailed answers to the "how" part remain elusive.

Laid out this way, I don't see your (our, since I agree in essence) top level view as being in conflict with any particular architectural view since any detailed version of the latter is at a lower level of "implementation detail". So, my question is "why do you?" Thinking your way has led me to a decidedly non-computational way of thinking about implementation, but I don't see any obvious conflict with other ways.

Note: I've come to question the value of such groupings when addressing the "scientific image". So, I agree with Elliot on that point except that I would restrict the intentional idiom to the "manifest image". See Ramberg's essay in "Rorty and His Critics".

Eric Schwitzgebel said...

Thanks for your thoughtful comments, Scott & Charles! I was camping with the family, hence the slow reply.

@ Scott: In the current draft of the paper I'm working on about this, I have chosen to recast the distinction as "deep" vs. "superficial". Relative to a set of "surface" features (behavior, phenomenology, folk psychologically recognizable cognitive states), my approach is superficial in that it doesn't probe beneath them. I'm not sure if this is better or clearer.

@ Charles & Scott: There's nothing wrong with probing the mechanisms, I think, and "daring to dream" or embracing some particular architectural account of how attitudes are implemented in the human case. I just want to suggest that such efforts concern only the *implementation* of something that should be conceptualized (at least for now) as not essentially a matter of underlying architecture. At least, that's my thought.

Charles T. Wolverton said...

Well, "superficial" seems to discount your (IMO, important) approach as well as the significance of the distinction you're making. The Ramberg essay referenced above distinguishes descriptions of an organism as a person, for which the "mental" vocabulary is well-suited, and descriptions of an organism as just an organism, for which it isn't. Descriptions of implementation are clearly the latter. Both your project and the implementation project are necessary, but they require their own vocabularies. In reading descriptions that are really about implementation but still use mentalese, I now more-or-less automatically purge the mentalese. Doing so almost always helps me in getting the point and sometimes even highlights what seem to be conceptual errors.