Goal reasoning as a general form of metacognition in BICA

A key capability that distinguishes humans from intelligent agents is the ability to generate and select new goals in an unexpected situation, while re-prioritizing existing goals. This level of cognitive autonomy becomes practically vital for actors working as team members with humans or in unpredictable environments. The work is focused on a new perspective on this problem based on biologically inspired cognitive architectures (BICA). Aspects of goal reasoning and goal management capabilities that are currently available in selected BICA are reviewed. A generalizing model of goal reasoning as a form of metacognition is defined, that integrates multiple mechanisms. On top of the traditional cognitive cycle, the model includes a goal-reasoning cycle triggered by notable phenomena. This goal reasoning cycle integrates mechanisms like agitation of desires by drives and their instantiation in goals, or metacognitive reasoning about goals. The model is illustrated by consideration of specific scenarios at micro- and macro-cognitive scales to evaluate the use and potential capabilities of this goal reasoning model for BICA. New capabilities that will be enabled based on this approach are expected to increase the human compatibility and cognitive autonomy of future intelligent agents. The general goal reasoning capability is a key component included in the BICA Challenge and is required for integration of intelligent agents into human teams. (C) 2014 Elsevier B.V. All rights reserved.