Abstract
We propose 'Cognitive Bootstrapping' as a blanket term for all instances of the process by which the perceptual apparatus of an autonomous agent is conceptually extended and experimentally validated. Bootstrap techniques are necessary to transcend the paradox inherent in validating perceptual categorisations via the objects of perception. Perception may thus become self-founding only within certain crucial a priori limits required to maintain referentiality and provide a validation criterion for the proposed perceptual updates. We hence survey the subject areas in which this mechanism occurs, ultimately advocating a hierarchically open-ended perception/action approach to artificial cognition in order to objectively ground perceptual updating