Monday, December 8, 2008

Concept Empiricism

Representing and Doing: Two Faces of Concepts

In a recent paper, Jerry Fodor (2004) has suggested that rationalists about concepts can be distinguished from defenders of other theories by their view of what concepts are for. Rationalists claim that concepts are for thinking. More precisely, to have a concept is to be able to think about something. Having the concept DOG is being able to think about dogs. A better word for thinking, in this context, might be representing. Concepts are primarily in the business of representing. For opponents of rationalism, including empiricists, having a concept is being able to do something. For example, having the concept DOG might be construed as the ability to categorize or interact with dogs. Fodor’s way of setting things up distinguishes two broad functions for concepts: rationalists say that concepts are primarily in the business of representing, and opponents of rationalism say that concepts are primarily in the business of doing. This distinction should not be regarded as a disjoint dichotomy. Empiricists do not deny that concepts representing. Rather, they claim that concepts have other equally important functions. Empiricists say that concepts must be able to representing things in a way that facilitates interaction with those things. Representing must be in the service of doing. The difference between Fodor and the empiricists is captured by a difference in the nature of the mental representations that they postulate. Fodor thinks concepts are like words. They are arbitrary symbols in a language of thought. There is little one can do with an arbitrary symbol.

A symbol does not include any instructions for how to interact with the category that it represents. For the empiricist, concepts are more like mental images, or inner models. Representations of that kind can be used to guide action. We can read features of a category off of our concepts if empiricism is right. I will discuss these issues of representational format below. In this section, I want to consider another question about the distinction between representing and doing. Fodor attempts to separate these two functions by developing a theory of representation that would allow concepts to represent things without encoding the kinds of features that would allow them to do anything. That theory constitutes one of the best current explanations of how concepts represent. If the theory supports Fodor’s rationalism, then empiricists have reason for concern. My goal here is to show that Fodor’s theory of reference is actually better suited for empiricism. Fodor (1990) develops a theory of representation that promises to explain how concepts represent without making any mention of what they do. This allows him to secure the conclusion that representing is prior to doing, which is a central tenet of rationalism. I begin with a summary of the theory.

According to Fodor, a concept represents that which would reliably cause the concept to be activated. A concept represents dogs if encounters with dogs would ordinarily cause that concept to activate. This is only a first approximation. Formulated in this way, the account faces an obvious objection. Our DOG concepts are activated when we encounter dogs, but they are also activated when we encounter things that merely look like dogs, e.g. foxes in bad lighting. If concepts represent anything that activates them, any concept that represents dogs would also represent foxes. To solve this problem, Fodor notes that there is an asymmetric dependency relation between dogs and foxes with respect to our DOG concepts: foxes would not cause our DOG concepts to activate were it not for the fact that dogs do, but the converse is not true: the fact that dogs cause our DOG concepts to activate is not a consequence of the fact that foxes do. Fodor construes this asymmetry synchronically, in terms of counterfactual dependencies. I will not argue the point here, but I think the best way to make sense of it, is diachronically. A DOG concept is one that was created in the context of dog encounters. Fox encounters would not cause the concept to activate were it not for dogs having done so in the past, but not conversely. On this reading, a concept represents a category when two conditions are met:

Nomological causation: the concepts is disposed to be reliably activated by encounters with members of the category, and

Etiological causation: encounters with members of the category played a role in the acquisition of the concept.

Fodor thinks that concepts represent in roughly this way (with a synchronic condition in place of the etiological causation clause). He also thinks that this story favors the hypothesis that concepts are primary in the business of representing, not doing. To see why, it is important to consider two other alternatives to this causal theory of reference. According to one view, concepts refer by resemblance. They are mental images that are structurally isomorphic with the things they represent. According to another view concepts are feature sets that refer via description. The concept DOG refers to dogs, because the concept dog contains a collection of features describing dogs, and dogs are the only things that satisfy the description. DOG contains FURRY, BARKS, QUADRUPEDAL, and so on. By denying these two theories of reference, Fodor is able to defend the view that concepts are unstructured arbitrary symbols (Fodor, 1998). They are words in a language of thought. Fodor can defend the language of thought story only be arguing that concepts do not depend on description or resemblance to refer. An individual word does not describe anything, and it does not look like what it refers to. By embracing a causal theory of reference, Fodor explains how word-like mental representations can refer. Without this, it would be difficult to maintain that concepts are couched in an arbitrary code. It would also be hard to maintain that concepts are primarily in the business of representing. An arbitrary symbol cannot be used, on its own to recognize dogs or draw inferences about dogs.

It is a dog symbol in the purest sense: it represents dogs and does nothing else. A mental image of a dog represents dogs and can also be used to recognize them. A dog description represents dogs and can also be used to draw inferences about them. Fodor’s causal theory of reference secures his hypothesis that concepts are primarily in the business of representing, not doing. Fodor’s mental word theory is radically different from the way most psychologists think about concepts. Psychologists emphasize the role that concepts play in categorization. If concepts are tools for categorizing, they cannot be unstructured word-like entities. They must be built up from features. Some psychologists say that DOG is a prototype; others say it is a mini-theory; and still others say it is a set of exemplar representations. As a rationalist, Fodor thinks that a theory of concepts need not explain how we categorize. His mental word theory is ideally suited for rationalism. After all, words in public languages represent, but we cannot categorize with them; they are arbitrary symbols. On Fodor’s view categorization is achieved by independent mechanisms. He doesn’t offer an account, but he might say that we have complex mental databases containing perceptual information, theories, prototypes, memory traces, and any number of features and facts. That is to say, categorization is achieved using the kinds of mental mechanisms that psychologists postulate. Think of a concept as a label on a large mental file. Information in that file, and information stored elsewhere can play a role in categorization.

The concept hovers safely above the overflowing sheets and scraps in the file. Items in the file represent what category members look like, the ontological domain they belong to, the attributes of specific instances, and so forth. But only the label represents the category itself. On the face of it, Fodor seems to have what he wants. He has a theory of how mental representation works that is consistent with rationalism. Concepts are arbitrary symbols that can be used for nothing other than representing categories. They cannot be used to draw inferences, to plan actions, or to categorize. All of those functions are handled by the contents of our mental files. But this picture is very odd. It renders concepts needlessly anaemic. Why should we say that concepts are arbitrary labels, rather than identifying concepts with the contents of our mental files? After all, the contents of those files do much more work. They allow us to categorize and act. Moreover, these files are absolutely essential for Fodor’s own theory of representation. A mental label represents a category by being reliably activated by instances of that category. But the label can be activated by category instances only if we have mechanisms that allow us to recognize those instances. An arbitrary DOG symbol can be triggered by dogs only if we have resources for recognizing dogs. Fodor assumes that all of the necessary resources are contained in our mental files. But once he makes that concession, the arbitrary labels begin to look unnecessary. It seems we should identity concepts with the file contents, rather than the file labels. We should say that concepts are the mechanisms that allow us to recognize categories rather than arbitrary mental words that flash on in the head when a category has been recognized. The labels are entirely unnecessary.

The moral is that Fodor’s theory of representation may not favour rationalism after all. Once we adopt a causal theory, we are forced to postulate mechanisms that allow us to reliably detect category instances. Once we postulate such mechanisms, we might as well identify them with concepts. We do not need to postulate arbitrary labels. Concepts can be complex databases. Such databases allow us to represent, but they also allow us to do things; they allow us to interact successfully with the world. So representing and doing are not disjoint functions, on this picture. They are intimately linked. On Fodor’s theory, we think using unstructured symbols. DOG is just an arbitrary word in the language of thought. On the view I am recommending, DOG is constituted instead by the representations used to identify dog. Thus, DOG is constituted by features that tell us what dogs look like and what their behavioural dispositions and affordances for interaction are. These features allow us to represent dogs, by securing reliable casual relations with then, but they also allow us to recognize dogs, and do things with dogs. Rather than saying that concepts are for representing, I would say that representations are for doing. The only reason we represent the world is to make our way through it. If concepts are not guides to possibilities for action, they are not useful. Concepts that merely represent belong to the fictional realm of pure Cartesian egos. Conceptual capacities that evolved in the real world allow us to run for cover or play fetch. But concepts also represent. The mechanisms that allow us to identify objects and interact with them also, thereby, establish reliably causal relations with those objects. Fodor himself shows how such causal relations can be used to establish reference. Ironically, his theory of reference fits perfectly with the anti-rationalist program.

Note: This article is based on the latest book that I am reading currently. I know that it is a bit technical (sorry I am a left-handed!!), but I really hope you guys can digest it. Cause it helps you to understand certain things that are really difficult to comprehend in our daily life.

No comments: