The notion that mental states and processes intervene between stimuli and responses sometimes takes the form of a ``computational metaphor'', which is often used as the identifying mark of contemporary cognitive science. The metaphor is this: The mind is to the brain as software is to hardware; mental states and processes are like computer programs implemented (in the case of humans) in brain states and processes. Some cognitive scientists, however, make a stronger claim: Mental states and processes are expressible as algorithms: ``cognition is a type of computation'' (Pylyshyn 1985: xiii).
Thus, according to the computational view of cognitive science, (1) there are mental states and processes intervening between input stimuli and output responses, (2) these mental states and processes are algorithms (according to the strong, or literal, form)--or they are like algorithms (according to the weak, or metaphorical, form), and--hence--(3) in contrast to behaviorism, mental states and processes are capable of being investigated scientifically (even if they are not capable of being directly observed).
Insofar as the methods of investigation are taken to be computational in nature, computer science in general and artificial intelligence in particular have come to play a central role in cognitive science. It is, however, a role not without controversial philosophical implications: For if mental states and processes can be expressed as algorithms, then they are capable of being implemented in non-human computers. The philosophical issue is simply this: Are computers executing such algorithms merely simulating mental states and processes, or are they actually exhibiting them? Do such computers think?
Those who hold to computational views of the mind reject the ``extensional'' definition of cognitive science as ``nothing more than six disciplines in search of a grant-giving agency'' (Johnson-Laird 1981: 147) or ``like the proverbial blind men trying to understand the elephant ...[or] simply a political union'' (Pylyshyn 1985: xi). In contrast, others (including some computer scientists) hold the more catholic view that ```cognitive science' is a broad rubric, intended to include anyone who is concerned with phenomena related to mind'' (Winograd 1981: 261). They ``believe in the value of multiple philosophies, multiple viewpoints, multiple approaches to common issues. ...[A] virtue of Cognitive Science is that it brings together heretofore disparate disciplines to work on common themes'' (Norman 1981c: 275-276).
Even cognitive scientists who disagree about the weak vs. the strong computational view of the mind are usually willing to agree that computer programs force cognitive scientists ``to make intuitions explicit and to translate vague terminology into concrete proposals; they provide a secure test of the consistency of a theory ...; they are `working models' whose behavior can be directly compared with human performance'' (Johnson-Laird 1981: 185-186; cf.\ Pylyshyn 1985: 76). That is, the proper methodology of cognitive science is to express one's theories about (human) cognition in a computer program (rather than, say, in English or in the languages of mathematics, logic, or statistics, as other sciences do) and then to compare the program's behavior with human cognitive behavior. Note, however, that this methodology accepts the strong form of the computational view of the mind as at least a working hypothesis. Although this methodology is consistent with the denial of the strong computational view--i.e., human cognitive behavior might be simulable by a computer program without itself being computational--such a denial is a very weak (if not incoherent) claim.