Evolution and Analysis of Brain-Body-Environment Systems - Randall D. Beer

Friday, 5 September, 2014 - 09:30 to 11:00

Interest in situated and embodied approaches to behavior and cognition has been expanding steadily for many years, bringing with it the growing realization that neural activity can only be properly understood within the context of the body and environment in which it occurs. However, both the conceptual foundations of this approach and the theoretical tools necessary to understand the resulting brain-body-environment systems are still in their infancy. In order to advance these aspects of the approach, we study simple model brain-body-environment systems that have been evolved to exhibit behaviors of interest. The operation of the resulting agents are then analyzed using the mathematical tools of dynamical systems theory and information theory to reveal how their behavior  arises from the interaction between their nervous systems, bodies and environments. After a brief overview of this research program, it will be illustrated using a recent model of embodied relational categorization. In this model, an agent presented with two falling objects of different sizes in sequence must catch the second object if it is smaller than the first and avoid it otherwise. Interestingly, both largely disembodied and strongly embodied strategies evolve to make this relational judgement. Dynamical analysis of successful evolved agents reveals the role of transient dynamical structures, discontinuous dynamics and nonautonomous bifurcations in their operation. Information-theoretic analysis reveals how task-relevant information flows throughout the system over time. Finally, a comparison of the two analyses reveals interesting connections between these different languages of description.