FriSem - Rosa Cao, Assistant Professor of Philosophy, Stanford University, "The virtues and vices of computational explanations"

Date
Fri April 7th 2017, 3:15 - 4:30pm
Event Sponsor
Department of Psychology
Location
Jordan Hall (Building 420), Room 050

Abstract: Brains are paradigmatically information-processing systems, and this is often taken to mean that brains perform computation, and furthermore, that we can explain the brain's capabilities by elucidating the computations involved. Computational explanation would seem to have many virtues, allowing us to abstract from messy details while highlighting commonalities among quite different systems.

But how can we distinguish the empirical commitments of different computational hypotheses?  And conversely, what principles can we use to guide the formulation of computational hypotheses from observable facts about brain activity? A theory of computational implementation would help to address both these questions.

 

I'll briefly survey philosophical theories of computational implementation, and then look at three issues they raise for the idea of neural computation:  interpretation, indeterminacy, and systematicity. If what determine what computation a computer is performing are the conventions that dictate how inputs and outputs should be interpreted, what is doing that interpretation in the brain?  When many interpretations are available, on what basis can we privilege one of those in explanation?  And finally, should we expect systematicity of the kind exhibited by a formal language in the neural code?  Without answers to these questions, computational approaches may explain less than we had hoped.