Conditions for Belief and Knowledge

In the previous chapter we were led to the paradoxical conclusion that the computational theory of mind (CTM) may result in a position that is at once dualistic and behaviourist. Whether it does so or not depends on how dogmatically its proponents interpret the strong AI program (see Chapter 3). If, as Pylyshyn counsels, they abandon their faith in the computational ‘metaphor’ of the mind as a kind of salvation-bringing new paradigm,1 interpret their claims literally (for example, thought as formal symbol manipulation) and then ensure that programs are constrained by empirical validation of their choice of functional architecture and algorithms (that is, that they stay close to the evidence) such consequences may be avoided. But, as I said, it is not at all clear what form these constraints should take.