Integrating Cognitive Models Based on Different Computational Methods

The mind integrates cognitive and perceptual processes that are currently best modeled using many different computational mechanisms. These are difficult to integrate into a single cognitive model and are thus a challenge to developing a unified model of human cognition. This paper presents two computational principles that identify common elements among these mechanisms and uses these principles, together with empirical findings about human cognition, to motivate a theory of human cognitive architecture. This theory has been embodied in the Polyscheme cognitive architecture, which enables models based on qualitatively different computational mechanisms, including Bayesian Networks, search and production systems, to be integrated into a single cognitive model and thus provides an explanation of how the mind integrates multiple cognitive structures and processes.