1. constitutive autonomy behavioral autonomy (artificial?) These considerations make it evident that there is a pressing need of finding a principled way of integrating these two approaches into one coherent framework of autonomous systems research. What kind of research methodology is up to this task? On the side of behavioral autonomy one of the most popular approaches is evolutionary robotics (Harvey et al. 2005). However, it models only behavior and the evolved agents are constitutively autonomous by definition only. More thought needs to be given as to how natural cognition is constrained by the constitutive processes which give rise to living systems. Is it the case that adding further biological mechanisms into the behavioral approach makes it closer to being autonomous in the constitutive sense (e.g. Di Paolo 2003)? On the side of constitutive autonomy we find cellular automata (e.g. Varela, Maturana & Uribe 1974) as well as simulated (e.g. Mavelli & Ruiz-Mirazo 2007) and actual chemistry (e.g. Bitbol & Luisi 2004). The problem here is how to get from systems that self- constitute to systems that self-constitute and do something interesting at the same time. Autonomy: a review and a reappraisal Tom Froese, Nathaniel Virgo & Eduardo Izquierdo Centre for Computational Neuroscience & Robotics, University of Sussex, Brighton, UK By distinguishing between behavioral and constitutive autonomy, we can see that this question actually demands two distinct responses. Despite the lack of a commonly accepted definition it seems reasonable to say that today’s systems are indeed more behaviorally autonomous (than at the start of ECAL, for example). However, the vast majority of this kind of research does not address constitutive autonomy. The question of how viability constraints emerge from the internal operations of a system while coupled to its environment is particularly relevant. However, more work is starting to be done in this area (e.g. Ikegami & Suzuki 2007, Di Paolo 2003). Taking both approaches to autonomy into account will be important for future research in artificial life. Here, we review the diversity of approaches and categorize them by introducing a conceptual distinction between behavioral and constitutive autonomy (Froese, Virgo & Izquierdo 2007). On the one hand, researchers associated within an engineering context mostly focus on the behavioral capacity of a system. These researchers tend to marginalize the differences between the autonomy of artificial agents and that of living organisms. On the other hand, researchers coming from a biological background often use the word to denote (metabolic) self-constitution. These researchers tend to treat the differences in autonomy between living and artificial agents as absolute. We argue that making a clear distinction between the two approaches can resolve a lot of this apparent opposition. In the field of artificial life there is no agreement on what defines ‘autonomy’. This makes it difficult to measure progress made towards understanding as well as engineering autonomous systems. Autonomy: a review Are today’s artificial agents more autonomous? E-Mail: {t.froese, n.d.virgo, e.j.izquierdo}@sussex.ac.uk http://lifeandmind.wordpress.com Mars rover Seagull Simulated agent (Beer 1995) Simulated metabolism (Bourgine & Stewart 2004) Behavioral autonomy: the concept ‘autonomy’ is used to refer to the robustness and flexibility of a system’s behavior. Constitutive autonomy: the concept is used to refer to the self-constitution of the system through its own operations. Beer, R.D. (1995), “A dynamical systems perspective on agent-environment interaction”, Artificial Intelligence, 72(1-2), pp. 173-215 Bitbol, M. & Luisi, P.L. (2004), “Autopoiesis with or without cognition: defining life at its edge”, J. R. Soc. Interface, 1(1), pp. 99-107 Boden, M.A. (1996), “Autonomy and Artificiality”, in: M.A. Boden (ed.), The Philosophy of Artificial Life, Oxford Uni. Press, pp. 95-108 Bourgine, P., & Stewart, J. (2004), “Autopoiesis and Cognition”, Artificial Life, 10(3), pp. 327-345 Brooks, R.A. (1991), “Intelligence without representation”, Artificial Intelligence, 47(1-3), pp. 139-160 Di Paolo, E.A. (2003), “Organismically-inspired robotics: homeostatic adaptation and teleology beyond the closed sensorimotor loop”, in: K. Murase & T. Asakura (eds.), Dynamical Systems Approach to Embodiment and Sociality, Adelaide, Australia: Advanced Knowledge Int., pp. 19-42 Franklin, S. (1995), Artificial Minds, Cambridge, MA: The MIT Press Froese, T., Virgo, N. & Izquierdo, E. (2007), “Autonomy: a review and a reappraisal”, in: F. Almeida e Costa et al. (eds.), Proc. of the 9 th Euro. Conf. on Artificial Life, Berlin, Germany: Springer-Verlag Harvey, I., Di Paolo, E.A., Wood, R., Quinn, M. & Tuci, E. A. (2005), ‘Evolutionary Robotics: A new scientific tool for studying cognition’, Artificial Life, 11(1-2), pp. 79-98 Ikegami, T. & Suzuki, K. (2007), “From Homeostatic to Homeodynamic Self”, BioSystems, in press Mavelli, F. & Ruiz-Mirazo, K. (2007), “Stochastic simulations of minimal self-reproducing cellular systems”, Phil. Trans. R. Soc. B, in press Varela, F.J., Maturana, H.R. & Uribe, R. (1974), “Autopoiesis: The organization of living systems, its characterization and a model”, BioSystems, 5, pp. 187-196 References Motivation New robotics A behavioral definition of autonomy can generally accommodate both artificial and biological agents. At the same time, however, it has difficulties in specifying exactly what makes such systems autonomous. Consequently, the requirements are often trivially met in many cases (e.g. Franklin 1995, p. 233). As an ambiguous and very inclusive approach, it threatens to make the concept of autonomy meaningless. The constitutive approach can provide a precise definition of autonomy in operational terms (e.g. Varela, Maturana & Uribe 1974). This has the consequence that its applicability is mainly restricted to actual organisms, since its aim is to distinguish life from non-life. Most current artificial life agents do not posess constitutive autonomy. Constitutive autonomyBehavioral autonomy Autonomy: a reappraisal Finally, it is important to note that the widespread disregard of the dimension of constitutive autonomy is a serious shortcoming not only for scientific research, but also in terms of our own understanding of what it means to be human. As Boden (1996) points out: “what science tells us about human autonomy is practically important, because it affects the way in which ordinary people see themselves – which includes the way in which they believe it is possible to behave”. The field of artificial life is therefore also faced by an ethical imperative to invest more effort into improving our understanding of constitutive autonomy. Only then can we ground our understanding of human freedom not only in terms of the behavior involved in mere external constraint satisfaction, but also in terms of the creativity involved in dynamic and open-ended self-realization. Concluding remarks While advances have been made in designing and understanding systems which are purely self-constituting (e.g. Bourgine & Stewart 2004) or purely behavioral (e.g. Beer 1995), little effort has been made to tackle these two complementary aspects of life in an integrated fashion. The major challenge for future artificial life research will be to address this shortcoming. Only when we are able to investigate both constitutive and behavioral autonomy via synthetic means can the field of artificial life claim to provide one coherent framework of autonomous systems research. The future of artificial life? GOFAI robotics Life