February 27, 2007

New paper online: "From Folk Psychology to Folk Epistemology: The Status of Radical Simulation"

This is a copy of the extended version of a paper that was recently accepted for presentation at the Pitt/CMU Graduate Philosophy conference:

In this paper I consider one of the leading philosophic-psychological theories of "folk psychology," the simulation theory of Robert Gordon. According to Gordon, we attribute mental states to others not by representing those states or by applying the generalizations of theory, but by imagining ourselves in the position of a target to be interpreted and exploiting our own decision-making skills to make assertions which we then attribute to others as "beliefs." I describe two leading objections to Gordon's theory--the problem of pretense and the problem of adjustment--and show how a charitably interpreted Gordon could answer these objections. I conclude, however, that the best case for Gordon's position still runs into a new problem concerning the epistemological presuppositions of belief-attribution. These presuppositions are themselves explicit and theoretical, and seeing how they operate shows how simulation theory lacks the elegant simplicity it seemed at first to have.

Posted by Ben at 10:10 PM | Comments (0)

January 06, 2006

Comments on Weiskopf, "Mental Mirroring as the Origin of Attributions"

I'm going to start the practice of occasionally blogging comments on papers I read, especially when the papers are new and available online. Since a paper in the latest issue of Mind and Language by Daniel Weiskopf ("Mental Mirroring as the Origin of Attributions") is directly relevant to the dissertation chapter I'm working on, I figured this was the perfect way to start.

First a little background on my dissertation. My dissertation is on the role of a naturalistic account of belief in the project of naturalizing epistemology. Basically, if naturalized epistemologists want to naturalize knowledge all the way down, it seems this should include a naturalistic account of belief, since it is the reliability of beliefs that is usually investigated in naturalized epistemologies. If the same standards used to naturalize other elements of epistemology (e.g., justification) do not deliver a naturalization of belief or some acceptable substitute, then the whole project is in trouble. I argue that those standards don't deliver, and that the project is in trouble.

So what does this have to do with Weiskopf? Weiskopf offers considerate but critical evaluation of Robert Gordon's "radical simulation" theory of belief attribution. Gordon's view is a subspecies of the wider category of "simulation theories" which are offered in opposition to "theory-theories" of folk psychology. Whereas theory-theories urge that our understanding of minds is based on stored information, simulation theories hold that this understanding instead derives from an ability we have to mimic others' mental states using our own mental apparatus.

It might seem that folk psychology is utterly irrelevant to the question of naturalizing "belief." After all, what "the folk" think about beliefs is not necessarily reflective of what beliefs actually are. But I think examining folk psychology does come to play a role when other routes of naturalization have been cut off. If there are naturalistic difficulties with identifying how mental states might reduce to or supervene upon physical states (and in earlier chapters of my dissertation I argue that there are), then a last resort could be to deflate "belief" via a pragmatic account of the use of the term "belief."

This approach seems to be exactly what Quine, the granddaddy of naturalized epistemology, recommends. As early as Word and Object he talks about belief-attribution as involving a kind of "dramatic" projection of oneself into another's place. In his later works he says this works through a process of "empathy." Perhaps if Quine can naturalize "empathy," he can naturalize a deflationary account of "belief." Following Quine's lead, philosophers like Jane Heal, Robert Gordon, and Alvin Goldman have attempted to defend simulation theory as an account of folk psychology. The question is whether these accounts, however faithful to our belief-attribution practices, are fully naturalistic.

Gordon's version of simulation theory is the best candidate for assessing whether or not "simulation theory" can succeed in naturalizing empathy, and this is made especially clear by Weiskopf's account. Gordon's theory is the "radical" version of simulation theory because he attempts to explain belief-attribution without use of the attributor's concept of "belief." He attempts to explain simulation using nothing but the resources of the attributor's powers of imaginative projection. If such an account could succeed, and presuming that some account of imagination would itself be naturalistically acceptable, then we would have a fully naturalistic, and non-circular account of belief-attribution, and a deflationary theory of "belief" that could be compatible with the naturalistic approach to epistemology.

Weiskopf means to critique Gordon, mainly on the grounds that evidence from psychology does little to support the radical simulation view. But I think Gordon's position is actually in bigger trouble than Weiskopf does, and Weiskopf's lucid presentation makes this even clearer to me. This comes from examining Weiskopf's presentation of the "problem of adjustment" objection to simulation theory.

The "problem of adjustment" runs as follows. If belief-attribution is really a process of simulation, then it needs to explain how we can interpret others whose mental states differ from our own. Weiskopf asks us to think about trying to understand someone else's taste in music, perhaps to predict what they might do if they like the Beatles, supposing that we ourselves dislike the Beatles. If we could imagine ourselves liking the Beatles, then we might imagine ourselves buying their records, going to conferences, etc., and we could then use this imagination to predict the same of the other. The trouble is that we don't like the Beatles, and can't seem to imagine this of ourselves. Yet we are able to understand the other's appreciation, and make predictions accordingly.

The problem of adjustment is easier to handle on other versions of the simulation theory than it is on Gordon's. On Goldman's theory, for example, there is no "radical" restriction that forbids our explanation of belief-attribution from making reference to the attributor's mentalistic concepts. One can simply imagine that one likes the Beatles, and see what follows from this. For this reason, of course, Goldman's theory does little good as an attempt to provide a deflationary naturalization of "belief" (or, in this case, "desire") since it utilizes the very concepts it seeks to explain. Gordon's theory, by contrast, has to take more extreme measures to avoid the reference to mentalistic concepts, Goldman-style.

To solve the problem of adjustment, Gordon argues that evaluative properties (such as liking) supervene on descriptive properties, and that this suggests a simulation strategy exploiting nothing but the attributor's imaginative resources. One need only imagine different evaluative properties while holding descriptive properties fixed. So, for example, one could imagine that the Beatles sound like Miles Davis (if one happens to like Miles Davis). Now Weiskopf thinks that this solution comes at the cost of further problems: why, in interpreting another's like for the Beatles, do we not then attribute other Miles Davis-properties to them? (This is the "problem of inferential promiscuity"). Weiskopf goes on to suggest various ways Gordon can avoid this problem, but I think this is unnecessary. I think there is a problem with this "Miles Davis" strategy that stops these further solutions from ever getting off the ground.

Recall that Gordon is trying to explain our mental state attributions in accordance with his "radical" restriction. On this restriction, interpreting another's desire to listen to the Beatles cannot involve anything like thinking to oneself "imagine that I like the Beatles." The Miles Davis strategy is supposed to solve this because one can simply substitute the Beatles for Miles Davis mentally, without having to use the concept of "like." The question I wish to raise is, what guarantees that the concrete example one uses as a substitute will be something that one in fact likes? For the substitution to work as a simulation, it does need to be something one likes in fact. The question is how this similarity can be guaranteed without the attributor's use of the concept of "like."

Normally, it seems that if we want to imagine what Gordon bids us imagine, we would approach it as follows: "What is something I like? Oh, Miles Davis. So I can imagine that the Beatles sound like Miles Davis." In other words, we use the concept "like" to achieve memory recall of the relevantly similar example. But this strategy is not available to Gordon, given his "radical" restriction. For the Miles Davis strategy to work, some other strategy that does not involve the use of the concept "like" would need to be involved. Is there such a strategy?

There are of course non-conceptual means of achieving memory recall. An obvious one is through perceptual association. We taste the pastry, and a la Proust it brings back memories of times past. We can revel in these memories and use them for any imaginative function, without having to use any particular concepts to recall them. If the pastry brings back memories of an old lover, we can imagine the lover.

The trouble is that there is no guarantee, or even any special tendency, for perceptually associated memories to involve objects that bear the relevant similarity we need for mental simulation to work. In fact perceptually associations are notorious for relating things willy-nilly. A taste reminds us of a lover, a song reminds us of hallways in our old high school, etc.

For Gordon's strategy to work, there would need to be some way in which thoughts of the Beatles would bring up the association of Miles Davis. Of course since the attributor in question does not even like the Beatles, there is no chance that the association could be in relation to things one likes. Perhaps one could observe the other person acting on his desires, going out an buying records, etc. And then one could be reminded that one has done the same on account of one's own preference for Miles Davis. But this would presuppose the very understanding of the connection between mental states and actions that one is here trying to explain. If one already understands that the other's actions are connected to this particular mental state, there is nothing left for simulation to accomplish.

In short, Gordon's strategy of imagining different supervenient properties, through the use of example substitution, does no good unless there can be some guarantee that the object imagined bears the relevant similarities to the object of the others' preferences. Since such a guarantee would need to come from the use of the very concept that his theory is supposed to explain, it seems that his theory fails—and with it, the prospects for a deflationary naturalization of "belief."

Posted by Ben at 07:46 PM | Comments (0)

December 20, 2005

Letter to The Atlantic Monthly on modularity theory

I sent the following letter to The Atlantic in response to an article by Yale psychologist Paul Bloom.

I agree with Paul Bloom (December Atlantic) that the basic error in supernatural belief is the undue separation of mental from physical properties. Ghosts, immortal souls and God do not exist, because there are no souls without bodies: the mental depends on the physical for its existence. I also agree that it is our capacity to separate mental and physical properties that makes this error possible. However I think that Professor Bloom's explanation for that capacity is dubious.

Bloom believes that infants possess separate neurological systems (what psychologists call "modules") for processing thoughts about the mental and the physical. Since the systems are separate and confined to their own subject matter, this is supposed to explain why their outputs needn't be correlated in the way they ought to be.

But this explanation seems exceedingly baroque. Why do we need separate modules for the mental and the physical to explain the possibility of imagining the separation of the two properties? If someone erroneously believes that men can fly, and separates the
"flight" property from birds while projecting it onto men, does this mean there is a special, separately evolved "flight" module?

Why not just chalk up the possibility of this separation to the general human capacity for abstraction, which permits us to imagine just about any two properties in isolation? It makes far more sense to say that our general power of abstraction evolved independently, owing to its obvious survival advantage as the source of rational thought, leaving in its wake the tolerable side-effect of the possibility of erroneous abstraction.

My explanation above does nothing, of course, to explain the pervasiveness of the religion error. But, I think, neither does the Bloom's theory. Why, after all, are he and his friends able to resist the innate lure of Cartesianism, while others cannot? Granted, social pressure cannot explain the possibility of belief in an error, but it can explain its motivation—and that can explain the pervasiveness. People believe in religion, I think, not primarily because it is comforting (often it is not), but because they fear social disapproval. God isn't a biological accident; He's a result of peer pressure!

To prepare the letter, I actually read most of Bloom's book, Descartes' Baby. It's a good example of a modularity theory of folk psychology, which my dissertation project is quite opposed to. This is a good topic to start my blog with, because it shows why one needn't resist the supernatural by resorting to naturalization. That is exactly what is happening with Bloom. He wants to explain away theistic belief, so he naturalizes it via a biological mechanism. The trouble is that he also would need to naturalize his own atheism. And why won't that explain it away?

Oh, and by the way, Bloom says that he thinks his six-year old son is wrong to think that thinking is what he does, while his brain just "helps him out." I think his son is right. We don't need to reject the commonsense view of the mental in order to reject the Cartesian dualist view that the mind is separate from the body. And we shouldn't. This is a general theme of my dissertation: first person observation trumps naturalization proposals any day.

Posted by Ben at 04:45 AM | Comments (0)