ACT-R Tutorial

ACT-R Tutorial Niels A. Taatgen (taatgen@cmu.edu) Hedderik van Rijn (hedderik@ai.rug.nl) Psychology, Carnegie Mellon University 5000 Forbes Av., Pittsburgh, PA 15213 USA Artificial Intelligence, University of Groningen Grote Kruisstraat 2/1, 9712 TS Groningen, Netherlands ACT-R (Anderson, Bothell, Byrne, Douglass, Lebiere & Qin, 2004) is a cognitive theory and simulation system for developing cognitive models. It assumes cognition emerges through the interaction of a procedural memory of productions with a declarative memory of chunks and independent modules for external perception and actions. Since its release in 1993, ACT-R has supported the development of over 100 cognitive models, published in the literature by many different researchers. These models cover topics as diverse as driving behavior, implicit memory, learning backgammon, metaphor processing, and emotion. This tutorial will discuss the latest version of ACT-R, ACT-R 6.0, which is more interruptible, achieves greater across-task parameter consistency, has better mechanisms of production learning, and is more in correspondence with our knowledge of brain functioning. The tutorial has no prerequisite knowledge, and is intended to on the one hand give an overview of the theory, and on the other hand offer some direct demonstration of ACT-R models. Although a half day is not sufficient to cover all material, it can whet the appetite for and serve as a kick start to the full ACT-R tutorial that is available online at http://act-r.psy.cmu.edu/. This website also provides for the necessary software, and overview of researchers using ACT-R, and it has a list of ACT-R publications (many of them downloadable). Although these individual research paradigms have produced interesting models by themselves, the full potential of the architecture can only be seen when they work together in models of complex cognition, which is the focus of a large proportion of current ACT-R research. In addition to the modeling paradigms we will discuss imaging research that shows how components of the ACT-R architecture can be mapped onto brain regions. During the tutorial, following Taatgen, Lebiere and Anderson (2006) four popular research paradigms within ACT-R will be used as a vehicle both to explain the architecture and to explain how ACT-R accounts for these phenomena. Figure 1: Overview of the ACT-R architecture Instance learning Learning by retrieving old experiences from memory, similar to Logan’s instance theory. References Anderson, J. R., Bothell, D., Byrne, M.D., Douglass, S., Lebiere, C., Qin, Y. (2004) An integrated theory of Mind. Psychological Review, 111, 1036-1060. Available online: http://act-r.psy.cmu.edu/papers/403/IntegratedTheory.pdf Utility learning Learning which of several available strategies is optimal by keeping track of costs and probability of success. Taatgen, N.A., Lebiere, C. & Anderson, J.R. (2006). Modeling paradigms in ACT-R. In R. Sun (ed.), Cognition and Multi-Agent Interaction: From Cognitive Modeling to Social Simulation (pp. 29-52). Cambridge University Press. Available online: http://www.ai.rug.nl/~niels/publications/taatgenLebiereAnd erson.pdf Perceptual/Motor constrained processing Models in which the main factor in explaining human performance lies in the limitations of their perceptual and motor systems. Rule learning Models in which new production rules are learned on the basis of combination of old rules and substitution of declarative knowledge.