Computation in the Brain: Towards a Mechanistic Theory of Cognition

University Of Missouri-Saint Louis, Saint Louis MO

Investigators

Abstract

As cognitive neuroscience continues its remarkable progress, data about the nervous system accumulate. There is increasing desire, opportunity, and need to integrate neuroscientific data with those of cognitive psychology, to reach a more sophisticated mechanistic understanding of cognition. Computation is perhaps the most fundamental notion employed in theories of cognition. Most cognitive scientists believe that cognition is computation of some sort. Their view is known as computationalism. It is important to notice that computationalists strongly disagree on how to think of cognition computationally. There are also those who reject computationalism. Anti-computationalists generally recognize the dominance of computationalism in cognitive science and construct their theories in explicit opposition to computational ones. This project outlines a strategy for taking the debate over the nature of cognition and its neural implementation to a higher level of sophistication by investigating the notion of computation and how it applies to the nervous system. The strategy adopted in the project is based on the following four working assumptions, which have been defended in the PI's previous work. First, there is not one notion of computation; rather, there are many and investigators need to be explicit as to which notion they are using. Second, there is little clarity on how notions of computation ought to be characterized and applied, and many authors whether they are in the computationalist camp or in the opposition employ the term 'computation' differently without explaining what they mean and without referring to any clear account of it; some regimentation is needed. Third, whether cognition is computation, and what kind of computation it is, is an empirical matter; but, this empirical matter cannot be resolved without first achieving clarity on how notions of computation are to be applied to concrete physical systems. Finally, since cognition is implemented by neural processes, neuroscience is the ultimate arbiter of whether cognition is any kind of computation and what kind of computation it is. The project uses the working assumptions above to motivate the development of a general and clear account of (different notions of) computation and how computation applies to concrete systems such as the nervous system. Specifically, That will involve isolating relevant notions of computation, characterizing them explicitly, and showing how they can be applied within the kind of mechanistic explanation practiced in neuroscience. Such an account can then be used to assess theories of cognition on empirical grounds.