Superpositional belief : sub-symbolic processing as a platform for a folk conception of mind

This thesis defends connectionism against the charge of propositional-attitude eliminativism. Recently a restricted class of connectionist models-those employing superpositional data-storage techniques-have been criticised for the support they lend to the elimination of the contentful states of mind recognised by folk psychology. In response, I argue that superpositional networks can ground a realistic construal of the central commitments of folk psychology. I begin by arguing that superpositional networks are fundamentally different from traditional cognitive models. Superpositional models, in contrast to traditional symbolic models, employ sub-symbolic processing. A system is sub-symbolic if the features responsible for the system's dynamics are at a lower level of organisation than the features that encode content within the system. This style of processing produces features-such as the acquisition of the initial representational economy-which are fundamentally incompatible with symbolic representation. It is precisely this style of processing, however, that has provoked the recent eliminative concerns. Folk psychological explanation necessarily involves the ascription of intentional mental states. As such, folk psychology is held hostage to a naturalistic account of representation. I argue that the critical task confronting any theory of representation proposed as a basis for the vindication of folk psychology is an explanation of the normative dimension of representations which does not jeopardise naturalism. I argue that this can be successfully achieved by a teleosemantic account of content. Furthermore, I argue that the representational nature of superpositional networks can be legitimated according to teleosemantic principles. I conclude that, since teleosemantics is self-consciousness naturalistic, superpositional networks may offer a metaphysically respectable platform upon which to model intentional belief states. Indeed, I claim that superpositional networks can preserve the central intuitions of folk psychology