Computational Theory of Mind

The computational theory of mind (CTM) holds that the mind is a digital computer: a discrete-state device that stores symbolic representations and manipulates them according to syntactic rules; that thoughts are mental representations -- more specifically, symbolic representations in a LANGUAGE OF THOUGHT; and that mental processes are causal sequences driven by the syntactic, but not the semantic, properties of the symbols. Putnam (1975) was perhaps the first to articulate CTM , but it has found many proponents, the most influential being Fodor (1975, 1981, 1987, 1990, 1993) and Pylyshyn (1980, 1984).

CTM's proponents view the theory as an extension of the much older idea that thought is MENTAL REPRESENTATION -- an extension that shows us how a commitment to mental states can be compatible with a causal account of mental processes and with a commitment to materialism and the generality of physics. Older breeds of representationalism were unable to explain how mental processes could be semantically coherent -- how thoughts could follow one another in a fashion appropriate to their meanings, while also being bona fide causal processes that did not depend on an inner homunculus who understood the meanings of the representations. Using formalization and digital computers, however, we can explain how this occurs. Formalization shows us how to link semantics to syntax. For any formalizable symbol system, it is possible to develop a set of formal derivation rules, based wholly on syntactic properties, that license all and only the inferences permissible on semantic grounds. Computers show us how to link syntax to causation. For any finite formal system, it is possible to construct a digital computer that automates the derivations of that system. Thus, together, formalization and computation show us how to link semantics to causation in a material system like a digital computer: design a set of syntactic rules that "track" the semantic properties of the symbols (i.e., formalize the system), and then implement those rules in a computer. Because digital computers are purely physical systems, this shows us it is possible for a purely physical system to carry out symbolic inferences that respect the semantics of the symbols without recourse to a homunculus or to any other nonphysical agency. Syntactic properties are the causal determinants of reasoning, syntax tracks semantics, and syntactic properties can be implemented in a physical system.

CTM has been touted both for its connections to successful empirical research in cognitive science and for its promise in resolving philosophical problems. The main argument in favor of the language of thought hypothesis and CTM has been the "only game in town" argument: cognitive theories of language, learning, and other psychological phenomena are the only viable theories we possess, and these theories presuppose an inner representational system. Therefore we have a prima facie commitment to the existence of such a representational system (Fodor 1975). Some have claimed that CTM also explains the INTENTIONALITY of mental states and that it reconciles mentalism with materialism. The meanings and intentionality of mental states are "inherited from" the meanings and intentionality of the "mentalese" symbols (Fodor 1981). And because symbols, the ultimate bearers of semantic properties and intentionality, can both have meaning and be physical objects, there is not even a prima facie conflict between a commitment to semantics and intentionality and a commitment to materialism. Finally, CTM has been held to explain the generative and creative powers of thought that result from the COMPOSITIONALITY of the language of thought. Chomskian linguistics shows us how an infinite number of possible sentences can be generated out of a finite number of atomic lexical units, syntactic structures, and transformation rules. If the basis of thought is a symbolic language, these same resources can be applied directly to explain the compositionality of thought.

Although CTM gained a great deal of currency in the late 1970s and 1980s, it has since been criticized on a number of fronts. First, with philosophers' rediscovery in the late 1980s of alternative approaches to psychological modeling, represented in NEURAL NETWORKS and dynamic adaptive systems, the empirical premise of the "only game in town" argument has been brought into question. Indeed, the main thrust of philosophical debate about neural networks and connectionism has been over whether their models of psychological phenomena are viable alternatives to rule-and-representation models.

Second, writers such as Dreyfus (1972, 1992) and Winograd and Flores (1986) have claimed that much human thought and behavior cannot be reduced to explicit rules, and hence cannot be formalized or reduced to a computer program. Thus, even if CTM does say something significant about the parts of human cognition that can be formalized, there are large portions of human mental life about which it can say nothing. Dreyfus and others have attempted to argue that this includes all expert knowledge and such simple skills as knowing how to drive a car or order in a restaurant.

A third line of criticism has been directed at CTM's use of symbolic meaning to explain the semantics of thought, on the grounds that symbolic meaning is derivative from the intentionality of thought, either causally (Searle 1980; Haugeland 1978; Sayre 1986) or conceptually (Horst 1996). Thus the attempt to explain intentionality by appeal to symbols is circular and regressive. Searle (1990) and Horst (1996) have taken this line of argument even further, claiming that the "representations" in computers are not even symbolic or syntactic in their own right, but possess these properties by virtue of the intentions and conventions of computer users: a digital machine not connected to our interpretive practices has a "syntax" only in a metaphorical sense of that word. Horst's version of these criticisms also yields an argument against the claim to reconcile mentalism with materialism: what digital computers show us how to do is to link convention-laden symbolic meaning with CAUSATION by way of convention-laden syntax, not to link the sense of "meaning" attributed to mental states with causation.

A fourth line of criticism has come from advocates of externalist theories of meaning. For many years, advocates of CTM tended also to be advocates of a "methodological solipsism" (Fodor 1980) or INDIVIDUALISM who held that the typing of mental states needed to be insensitive to features outside of the cognizer because the computational processes that determined thought have access only to mental representations. At the same time, CTM required that the typing of mental states reflect their semantic properties. These two commitments together seemed to be incompatible with externalist theories of content, which hold that the meanings of many terms are at least partially determined by factors that lie outside of the cognizer, such as its physical (Putnam 1975) and linguistic (Burge 1979, 1986) environment. This was used by some externalists (e.g., Baker 1987) as an argument against computationalism, and was used at least at one time by Fodor (1980) as a reason to reject externalism. Nevertheless, at least some computationalists, including Fodor (1993), have now embraced strategies for reconciling computational theories of mental processes with externalist theories of meaning for mental representations.

Additional links

-- Steven Horst

References

Baker, L. R. (1987). Saving Belief: A Critique of Physicalism. Princeton: Princeton University Press.

Burge, T. (1979). Individualism and the mental. In P. French, T. Euhling, and H. Wettstein, Eds., Studies in Epistemology, Midwest Studies in Philosophy, vol. 4. Minneapolis: University of Minnesota Press.

Burge, T. (1986). Individualism and psychology. Philosophical Review 95(1):3-45.

Dreyfus, H. (1972). What Computers Can't Do. New York: Harper and Row.

Dreyfus, H. (1992). What Computers Still Can't Do. Cambridge, MA: MIT Press.

Fodor, J. (1975). The Language of Thought. New York: Crowell.

Fodor, J. (1980). Methodological solipsism considered as a research strategy in cognitive science. Behavioral and Brain Sciences 3:63-73.

Fodor, J. (1981). Representations. Cambridge, MA: MIT Press.

Fodor, J. (1987). Psychosemantics. Cambridge, MA: MIT Press.

Fodor, J. (1990). A Theory of Content and Other Essays. Cambridge, MA: MIT Press.

Fodor, J. (1993). The Elm and the Expert. Cambridge, MA: MIT Press.

Haugeland, J. (1978). The nature and plausibility of cognitivism. Behavioral and Brain Sciences 2:215-226.

Haugeland, J., Ed. (1981). Mind Design. Cambridge, MA: MIT Press.

Horst, S. (1996). Symbols, Computation and Intentionality: A Critique of the Computational Theory of Mind. Berkeley and Los Angeles: University of California Press.

Putnam, H. (1975). The Meaning of "Meaning." In K. Gunderson, Ed., Language, Mind and Knowledge. Minnesota Studies in the Philosophy of Science, vol. 7. Minneapolis: University of Minnesota Press.

Pylyshyn, Z. (1980). Computation and cognition: Issues in the foundation of cognitive science. Behavioral and Brain Sciences 3:111-132.

Pylyshyn, Z. (1984). Computation and Cognition: Toward a Foundation for Cognitive Science. Cambridge, MA: MIT Press.

Sayre, K. (1986). Intentionality and information processing: An alternative model for cognitive science. Behavioral and Brain Sciences 9(1):121-138.

Searle, J. (1980). Minds, brains and programs. Behavioral and Brain Sciences 3:417-424.

Searle, J. (1984). Minds, brains and science. Cambridge, MA: Harvard University Press.

Searle, J. (1990). Presidential address. Proceedings of the American Philosophical Association.

Searle, J. (1992). The Rediscovery of the Mind. Cambridge, MA: MIT Press.

Winograd, T., and F. Flores. (1986). Understanding Computers and Cognition. Norwood, NJ: Ablex.

Computational Theory of Mind

See also

Additional links

References

Further Readings