Distributed Cognition (DCOG): Foundations for a Computational Associative Memory Model

摘要

Computational models of human behavior and performance continue to gain in importance for the development and use of tools to support Air Force personnel in warfighting operations. In this report, we describe the foundations of a different type of computational architecture; one that we believe will be less susceptible to cognitive brittleness and can better scale to complex and ill-structured work domains. More specifically, it describes an investigation into the design and software implementation of an associative style memory model. The memory model expresses knowledge at three fundamental levels of granularity as opposed to one found in other machine representations of cognition. The report describes the theoretical constructs used in the memory model; illustrates how they can be used to account for certain cognitive phenomena such as over generalization of categories by children and base rate neglect in decision making; develops these concepts mathematically and algorithmically; and illustrates how they capture cognition in a concept leaning task. It concludes with a software implementation plan that embeds the memory model into a software actor Distributed Cognition (DCOG) that employs a distributive software agent architecture.