Companion Cognitive Systems: An Overview

Introduction This talk will describe work in progress on Companion Cognitive Systems, a new cognitive architecture whose goal is modeling large-scale learning of conceptual knowledge. Keywords: Analogy, cognitive architecture. Motivation Existing cognitive architectures, such as ACT-R and SOAR, tend to focus on skill learning. While important, we believe that it is equally important to understand how conceptual knowledge is learned, and organized for effective use. Moreover, models based on ACT-R and SOAR tend to be focused on specific tasks, and often are run in isolation, for short periods, rather than treated as organisms that must survive for extended periods of time. With Companions, we are exploring a very different set of issues. Our goal is to create computational systems that can be treated like sentient beings, handling conceptual knowledge in humanlike ways, without pre-built domain limitations, over extended periods of time. This includes continually adapting and learn