From Microcognition to Macrocognition: Architectural Support for Adversarial Behavior

Abstract

Asymmetric adversarial behavior is a complex naturalistic domain that involves multiple macrocognitive processes. Traditional techniques that have been applied to that domain from game theory and artificial intelligence have generalized poorly from simplified paradigms to real-world conditions. We describe a competing approach rooted in cognitive architectures and distinguished by multiple levels of processes that involve complex interactions of microcognitive constructs. The naturalistic requirements of the task impose upon the cognitive architecture additional constraints beyond those involved in modeling laboratory experiments. We describe a number of improvements to the cognitive architecture designed to boost its robustness in uncertain, unpredictable, and adaptive environments characteristic of adversarial behavior. What emerges is a symbiotic relation between macrocognitive processes that drive improvements in microcognitive constructs, which in turn provide a computational account of the realization of those processes in human cognition.