Cognitio 2011

Abstract: The goal of the cognitive architectures is to suggest a unified theory of cognition. Researcher to propose a cognitive architecture came up with different idea. Some, such as Anderson and Newell (Anderson, 1983, Newell, 1990, Anderson and Lebiere, 2003) who have postulated the symbol manipulation to study cognition. Others proposed that cognition is the construction of skills through a self-organization process(Brooks, 1991). However, recent researches suggest that hybrid systems deemed more appropriate when it comes to proposing a unified theory of cognition. Hybrid system postulates a combination of both cognitive and emergent system to propose a unified theory of cognition. However, there is no consensus regarding the manner in which one could combine both cognitivist and emergent systems to create a hybrid system. In this paper, we propose our experiments for the creation of hybrid systems. We have developed a cognitive agent named CELTS inspired by the latest research in neurosciences. The agent is both capable of symbol manipulation and emergent behavior. The agent’s architecture is based on thousands of simple processors that collaborate to generate emergent decision by a deliberation process inspired by the Baars theory of consciousness (Baars, 1988). The agent also integrates several types of learning that we have integrated in its architecture such as episodic learning, procedural learning, causal learning and emotional learning.

The CELTS agent was applied in a virtual learning environment for teaching the manipulation of the Canadarm2 robotic arm on the space station. Several papers have been published about the architecture of CELTS (Faghihi et al., 2011, Faghihi et al., 2010a, Faghihi et al., 2010b). In this presentation, we will give an overview of this research and a glimpse at our newest results.