The Embodied Computational Trap

From Ben Shirt-Ediss

A term paper submitted as coursework for the Intelligence in Animals and Machines module of EASy MSc, Sussex University, January 2008.

Abstract

This paper is an argument that, in light of recent dynamical intuition, a so-called “embodied computational” perspective is insufficient to explain the behaviour of complex, embodied biological systems. Several biological studies which each report to offer semi neural-based explanations of behaviour are shown to commit to this flawed perspective. Rather than trying to decipher what elusive computational mapping an animal’s nervous system might implement, it may be better to divert effort into building and experimenting with simple models from the ground-up, formulated in terms of the more rigorous dynamical systems framework. Work has already begun in this direction.

Files

• Term Paper: The Embodied Computational Trap (PDF)