"Emotion and the computational theory of mind"

Computational functionalism allows for cognitive processes to be implementation independent; the representational level of description is how we describe cognition, and different sub-symbolic systems could be used to instantiate any one kind of representational system. This is one of the advantages of the theory for those who believe that very different types of structures could be minds, just as there are different kinds of ways to instantiate a computer program. This view, however, is much less plausible as an account of human cognition when we consider processes that are intimately involved with cognition, but which are sub-symbolic. Emotion and other affective states are such processes. Although potentially independent of cognition, emotions are normally cognitive, and both affect and are affected by cognition. Processes like emotional congruence in perception show that emotion affects the very kinds of processes that a computational theory of mind wants to explain. But the failure of theories of emotion which attempt to reduce emotion to judgments show that these effects cannot be assimilated to the representational processes themselves. In fact, the extended body and the effects of emotion upon it are essential to some of the effects emotion has upon cognition. By examining these kinds of emotional processes, I show that a computational functionalist cannot ultimately defend, as a complete account of human cognition, a representational level of description that is implementation independent.

In Two Sciences of Mind: Readings in Cognitive Science and Consciousness, Sean O'Nuallain, Paul McKevitt and Eoghan MacAogain (eds). Philadelphia, PA: John Benjamins, 1997.