|
A Cognition Briefing Contributed by: David Vernon, Khalifa University of Science, Technology, and Research, UAE The term cognitive architecture originated with the seminal cognitivist work of Newell et al. Consequently, the term has a very specific meaning in the cognitivist paradigm, even though the term is used freely by proponents of alternative emergent and hybrid approaches. In the cognitivist paradigm, cognitive architectures represent attempts to create unified theories of cognition, i.e. theories that cover a broad range of cognitive issues, such as attention, memory, problem solving, decision making, learning, from several aspects including psychology, neuroscience, and computer science. Newell's Soar architecture, Anderson's ACT-R architecture, and Minsky's Society of Mind are all candidate unified theories of cognition. For emergent approaches to cognition, which a focus on development from a primitive state to a fully cognitive state over the life-time of the system, the architecture of the system is equivalent to its phylogenetic configuration: the initial state from which it subsequently develops. In the cognitivist paradigm, the focus in a cognitive architecture is on the aspects of cognition that are constant over time and that are relatively independent of the task. Since cognitive architectures represent the fixed part of cognition, they cannot accomplish anything in their own right and need to be provided with or acquire knowledge to peform any given task. This combination of a given cognitive architecture and a particular knowledge set is generally referred to as a cognitive model. In most cognitivist systems the knowledge incorporated into the model is normally determined by the human designer, although there is in increasing use of machine learning to augment and adapt this knowledge. The specification of a cognitive architecture consists of its representational assumptions, the characteristics of its memories, and the processes that operate on those memories. The cognitive architecture defines the manner in which a cognitive agent manages the primitive resources at its disposal. For cognitivist approaches, these resources are the computational system in which the physical symbol system is realized. The architecture specifies the formalisms for knowledge representations and the memory used to store them, the processes that act upon that knowledge, and the learning mechanisms that acquire it. Typically, it also provides a way of programming the system so that intelligent systems can be instantiated in some application domain. For emergent approaches, the need to identify an architecture arises from the intrinsic complexity of a cognitive system and the need to provide some form of structure within which to embed the mechanisms for perception, action, adaptation, anticipation, and motivation that enable the ontogenetic development over the system's life-time. It is this complexity that distinguishes an emergent developmental cognitive architecture from a simple connectionist robot control system that typically learns associations for specific tasks. In a sense, the cognitive architecture of an emergent system corresponds to the innate capabilities that are endowed by the system's phylogeny and which don't have to be learned but of course which may be developed further. There resources facilitate the system's ontogensis. They represent the initial point of departure for the cognitive system and they provide the basis and mechanism for its subsequent autonomous development, a development that may impact directly on the architecture itself. As we have stated already, the autonomy involved in this development is important because it places strong constraints on the manner in which the system's knowledge is acquired and by which its semantics are grounded (typically by autonomy-preserving anticipatory and adaptive skill construction) and by which an inter-agent epistemology is achieved (the subjective outcome of a history of shared consensual experiences among phylogenetically-compatible agents). It is important to emphasize that the presence of innate capabilities in emergent systems does not in any way imply that the architecture is functionally modular: that the cognitive system is comprised of distinct modules each one carrying out a specialized cognitive task. If a modularity is present, it may be because it develops this modularity through experience as part of its ontogenesis or epigenesis rather than being prefigured by the phylogeny of the system (e.g. see Karmiloff-Smith's theory of representational redescription.) Even more important, it does not necessarily imply that the innate capabilities are hard-wired cognitive skills as suggested by nativist psychology (e.g. see Note 1). At the same time, neither does it necessarily imply that the cognitive system is a blank slate, devoid of any innate cognitive structures as posited in Piaget's constructivist view of cognitive development (see Note 2). at the very least there must exist a mechanism, structure, and organization which allows the cognitive system to be autonomous, to act effectively to some limited extent, and to develop that autonomy. Finally, since the emergent paradigm sits in opposition to the two pillars of cognitivism – the dualism that posits the logical separation of mind and body, and the functionalism that posits that cognitive mechanisms are independent of the physical platform – it is likely that the architecture will reflect or recognize in some way the morphology of the physical body of which it is embedded and of which it is an intrinsic part. This article is abstracted from A Survey of Artificial Cognitive Systems: Implications for the Autonomous Development of Mental Capabilities in Computational Agents, D. Vernon, G. Metta, G. Sandini (2007), IEEE Transactions on Evolutionary Computation, Vol. 11, No. 2, pp. 151-180, 2007. For an alterative viewpoint, see Cognitive architectures: Research Issues and Challenges by P. Langley, J. E. Laird, and S. Rogers (2006).
Note 1
Note 2
References
|