The regulation of control: Two computational studies.

摘要

The term cognitive control refers to the set of processes responsible for configuring and coordinating basic information processing mechanisms in the performance of specific tasks. Studies of control have often focused on how its influence is exerted on the rest of the processing system. The present dissertation describes work aimed at supplementing this 'top-down' view of control with an account of the 'bottom-up' influences that determine the behavior of control itself. The dissertation brings together reports of two sets of computational modeling studies, each implementing an hypothesis concerning the regulation of control. The first set of studies draws on behavioral and neuroimaging data to suggest that control may respond to the occurrence of conflicts in information processing. The second set of studies builds on earlier work with recurrent neural networks to suggest how routine sequential behavior might emerge from a system that acts based in part on an assessment of its own internal states. Together, the two sets of studies provide a view of how cognitive control---at a number of functional levels and across behavioral domains---might not only regulate information processing, but also respond to it, displaying organized, strategic behavior without the need for a homunculus.