A computational model of the allocentric and egocentric spatial memory by means of virtual agents, or how simple virtual agents can help to build complex computational models Action editor: Gregg Oden Cyril Brom a,⇑ , Jan Vyhna ´nek a , Jir ˇı ´ Lukavsky ´ b , David Waller c , Rudolf Kadlec a a Faculty of Mathematics and Physics, Charles University in Prague, Czech Republic b Institute of Psychology, Academy of Sciences of the Czech Republic, Czech Republic c Department of Psychology, Miami University, United States Received 30 September 2010; received in revised form 2 August 2011; accepted 17 August 2011 Available online 5 October 2011 Abstract The ability to acquire, remember and use information about locations of objects in one’s proximal surrounding is a fundamental aspect of human spatial cognition. In this paper, we present a computational model of this ability. The model provides a possible expla- nation of contradictory results from experimental psychology related to this ability, namely explanation of why some experiments have reproduced the so-called “disorientation effect” while others have failed to do so. Additionally, in contrast to other computational mod- els of various aspects of spatial cognition, our model is integrated within an intelligent virtual agent. Thus, on a more general level, this paper also demonstrates that it is possible to use intelligent virtual agents as a powerful research tool in computational cognitive sciences. Ó 2011 Elsevier B.V. All rights reserved. Keywords: Spatial cognition; Paradigm of pointing; Disorientation effect; Intelligent virtual agent 1. Introduction Computational approaches to cognitive science have become increasingly important in the past decade (Burgess, 2007; Sun, 2008). This is likely because, compared to ver- bally-based theories, in silico simulations enable a researcher to specify hypothetical mechanisms in precise detail, system- atically explore the model and manipulate its parameters, and generate new predictions (Burgess, 2007; Marsella & Gratch, 2009; Sun, 2008; Tyrrell, 1993). At the same time, simulations can be both more complex and well-specified than analytical or numerical models (Kokko, 2007, chap. 8). Intelligent virtual agents (IVAs) are pieces of software that are both autonomous and graphically embodied in a 2D or 3D virtual environment, capitalizing on the general agent metaphor used in software engineering and artificial intelligence (Wooldridge, 2002). IVAs are currently used in a large variety of applications, for instance, in educational simulations, virtual storytelling, cultural heritage, and computer games. In general, research on IVAs attempts to make these entities more believable, that is, to increase the agents’ abil- ity to appear and behave in a lifelike manner. Believable IVAs enable users to suspend their disbelief by providing a convincing portrayal of the personality the user expects (Loyall, 1997). It is worth noting that for designers of many IVAs, the goal is to imitate a character’s behavior, but not necessarily to develop a cognitively plausible model pro- ducing the behavior. Nonetheless, these agents can be used as tools for investigating plausible computational models 1389-0417/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.cogsys.2011.09.001 ⇑ Corresponding author. Address: Faculty of Mathematics and Physics, Charles University in Prague, Room 312, Malostranske Namesti 25, Prague 11800, Czech Republic. Tel.: +420 221 914 216; fax: +420 221 914 281. E-mail address: brom@ksvi.mff.cuni.cz (C. Brom). www.elsevier.com/locate/cogsys Available online at www.sciencedirect.com Cognitive Systems Research 17–18 (2012) 1–24