Beyond information transfer: the emergence of embodied communication
50th Anniversary Summit of Artificial Intelligence
9-14 July 2006, Monte Verita, Switzerland
The “Cognition Is Computation” and “Communication Is Information Transfer” metaphors are among the most deeply entrenched constitutive assumptions of traditional AI research. However, more recently there is a growing consensus in the cognitive sciences that it is crucial for our understanding of cognition and communication that we study them as a biological phenomena. Consequently, the functional definition of communication as an exchange of information with selective advantage is inadequate because it only pertains to the descriptive domain of the observer. What is needed instead is a shift of focus to the underlying dynamical mechanisms and morphological structures, which causally enable communication to emerge as a coordination of behavior between embodied, situated, and autonomous systems. In what follows I will first briefly outline the progress that has been made in artificially evolving behavioral coordination in multi-agent systems, and then briefly describe why these developments are particularly relevant for the future of embodied cognitive science as a whole.
Almost since the beginning of research in artificial life there have been attempts at using synthetic models to gain insights into the origins of communication. In practice this usually takes the form of artificially evolving communication between embodied and situated agents (Steels 2003). However, within this broad framework the two main paradigms of cognitive science can nevertheless be distinguished: the computational approach to cognition characterizes communication in functional terms as a transfer of information between sender and receiver, while the embodied approach defines communication in operational terms as a form of behavioral coordination taking place between two or more structure-determined dynamical systems (Maturana & Varela 1987). Recently, the computational approach has come under criticism; for while there are examples of communicative behavior between living organisms which can be said to describe a certain state of affairs, these descriptions require the existence of previous consensual agreement which can only be achieved on the grounds of pre-existing communicative abilities (Di Paolo 1997). This shortcoming of the computational approach is also reflected in models in which a communication channel is explicitly made available to the agents as part of the experimental setup (e.g. MacLennan & Burghardt 1994; Werner & Dyer 1991). In this manner the problems regarding the biological origins of communication are ignored: communicative behavior does not first have to originate from non-communicative behavior but merely has to be fine-tuned by artificial evolution.
In contrast, the use of evolutionary robotics methodology (Harvey et al. 2005) in combination with a dynamical systems approach to cognition (van Gelder 1999) provides a promising framework for investigating the origin of biologically grounded communicative behavior via synthetic means. In fact, the dynamical systems approach can be seen to directly follow from Maturana & Varela’s autopoietic tradition which emphasizes the biological dimension of cognition (Beer 2004). Indeed, more recently some research at the University of Sussex has successfully used this kind of approach to evolve embodied behavioral coordination without dedicated communication channels. For example, to evolve movement coordination through role allocation between two simulated autonomous agents equipped only with wheels and proximity sensors (Quinn 2001), and a similar task has been successfully implemented with robots (Quinn et al. 2002). These developments are important for the new paradigm because they present us with concrete examples where non-trivial, situated, and embodied communication between autonomous agents has evolved from non-communicative behavior. This focus on the dynamics of behavior, as well as the morphology of the organism and its environment, permit an operational analysis of how these three theoretical entities are related, and how they enable non-trivial communicative behavior to emerge. Further work in this area could explore the effects of body morphology on communication.
As a final remark I would like to suggest the possibility that research into the origins of communication with the use of artificial life models could enable embodied cognitive science to finally go beyond the computational paradigm. On the one hand, social cognition in general, and embodied communication in particular, present us with one strong example of how to avoid the trap of a restriction to low-level sensorimotor accounts of cognitive behavior, and move on to the study of more complex cognitive phenomena such as language and social systems, while still retaining the same methodological commitments. This is because an embodied form of communication allows an agent to make use of the same primary cognitive capabilities which have been evolved for intelligent action-in-the-world in the higher-level cognitive domain of social phenomena. Indeed, if we accept the claim that the conscious observer arises through ‘languaging’ (Maturana 1978), then this kind of research could provide us with a path towards the holy grail of AI: machine consciousness.
Beer, R. (2004), ‘Autopoiesis and Cognition in the Game of Life’, Artificial Life, 10, pp. 309-326
Di Paolo, E.A. (1997), ‘An investigation into the evolution of communication’, Adaptive Behavior, 6(2), pp. 285-324
Harvey, I., Di Paolo, E., Wood, R., Quinn, M. and Tuci, E. A. (2005), ‘Evolutionary Robotics: A new scientific tool for studying cognition’, Artificial Life, 11(1-2), pp. 79-98.
MacLennan, B.J., & Burghardt, G.M. (1994), ‘Synthetic ecology and the evolution of cooperative communication’, Adaptive Behavior, 2(2), pp. 151-188
Maturana, H.R. (1978), ‘Biology of Language: The Epistemology of Reality’, Psychology and Biology of Language and Thought: Essays in Honor of Eric Lenneberg, Miller, G.A., & Lenneberg, E. (eds.), New York, NY: Academic Press, pp. 27-63
Maturana, H.R., & Varela, F.J. (1987), The Tree of Knowledge: The Biological Roots of Human Understanding, Boston, MA: Shambhala Publications
Quinn, M. (2001), ‘Evolving communication without dedicated communication channels’, Proc. of the 6th Euro. Conf. on Artificial Life, ECAL’01, Kelemen, J. and Sosik, P. (eds.), Springer, pp. 357-366
Steels, L. (2003), ‘The Evolution of Communication Systems by Adaptive Agents’, Adaptive Agents and Multi-Agent Systems, Alonso, E. et al. (eds.), LNAI 2636, Springer Verlag, pp. 125-140
Werner, G.M. & Dyer, M.G. (1991), ‘Evolution of Communication in Artificial Organisms’, Artificial Life II, Langton, C.G., Taylor, C., Farmer, J.D., & Rasmussen, S. (eds.), Addison-Wesley, pp. 659-687