Article Preview
TopIntroduction
A cognitive architecture can be viewed as an embodiment of a scientific hypothesis of (human and nonhuman, both animal and artificial) cognition. Cognitive architectures are designed to be capable of performing certain behaviours and functions based on our understanding of minds (Newell & Simon, 1972; Franklin, 1995; Davis, 2002). Cognitive science has developed in a number of directions including intelligent systems, reasoning, knowledge representation, and robotics. The evaluation of cognitive architectures has always been challenging. Several common concepts and different methodologies have been applied to developing new architectures.
There are many examples of cognitive architectures developed for different purposes through using different concepts available in different disciplines; for example, SOAR (Newell, 1990), ACT-R (Anderson, 1993), CRIBB (Bartsch & Wellman, 1989; Wahl & Spada, 2000), CHREST (Gobet et al., 2001), EM-ONE (Singh, 2005), CogAff (Sloman, 2002) and CAMAL (Davis, 2002, 2008). Different cognitive architectures and paradigms can be said to be modelling different aspects of cognition, with different aims, using different metaphors, and from different contexts. To develop a better and more sophisticated cognitive architecture, researchers need to understand: (1) the sufficient description of theoretical, design and implementation levels of different architectures and; (2) the missing, common and generalised factors of relevant cognitive architectures.
The developing Society of Mind Cognitive Architecture (SMCA) (Venkatamuni, 2008) extends the Davis (2002, 2008) CAMAL cognitive architecture with extra processing layers using society of mind and metacognition concepts. Intelligent behaviour can be viewed as a combination of more simple behaviours. Imagine a simple reactive agent that can only move towards and collect a resource in an environment. Building an optimal or metacognition agent cannot be done with just a community of such simple agents, as they need to interact or take help from other agents. Hence from the perspective of Minsky (1985), developing a cognitive architecture requires the development of many different types of agents, with different behaviours and capabilities.
More importantly, such a community of agents requires agents that perform at more abstract levels. In order to provide a truly adaptable framework a “Society of Mind” needs a top layer catalyst like metacognition. Metacognition is a relatively new buzz word in cognitive theory (Adkins, 2004). Metacognition is defined as thinking about thinking and can be viewed in two ways:
Agent behaviours can be analyzed using many different metrics; for example, affect valencing, (pseudo-)metabolic activity, competition and social interaction with respect to environment and microeconomics. The application of economics to artificial life to analyse adaptive behaviors provides a coherent framework from a cognitive architecture across many levels and types of processing.
TopArtificial Minds
Minsky (1985) defines mind as the functioning of the brain. Franklin (1995) defines mind as a mechanism of the brain. Minsky says “minds are just what brains do”. Franklin (1995) argues that the foundation for exploring the mechanisms of mind can be done through the possibilities offered by artificial minds. This gives rise to artificial minds defined as man made systems that exhibit behavioral characteristics of natural minds. However, artificial minds need not be limited to being analogues of natural ones. The possibilities offered by computational techniques, and synthetic designs, are equally valid in defining what constitutes an artificial mind.