COLIN KLEIN: "COMPUTATION + REPRESENTATIONAL STRUCTURE = COGNITION"
The format in which data is represented affects which algorithms are most efficient to run on that data. Work in cognitive ontology has spent a lot of time distinguishing computational processes, and comparatively little time considering the format of representation. I discuss several examples demonstrating the importance of broad representational concerns, including an extended discussion of the representational bases for peripersonal space. The choice of representation ultimately affects how we individuate processes. I also discuss work showing some of the pitfalls of using dimensionality reduction algorithms to characterize representational spaces, which suggests that representational space cannot be characterized wholly independently of a theory of the underlying computation. Hence distinguishing computations and exploring common representational structure must be intertwined scientific problems.
Return to videos page.