Observation-based expectation generation and response for behavior-based artificial creatures

Abstract

This thesis seeks to address the incorporation of a low-level cognitive ability into reactive, behavior-based artificial intelligence architectures. Specifically, it addresses the need to generate short-term, observation-based expectations about the world and react appropriately to the violation of those expectations. In it I discuss the motivation for incorporating expectations into a reactive behavior-based architecture, outline the qualitative properties of expectations and the conditions under which they may be violated, propose a model for generating expectations and responding to their violation, detail one implementation of such a model, and finally propose this work as a starting point from which future work on higher-order cognition and behavior might begin.