Artificial Life and Higher Level Cognition

Abstract Artificial Life is the study of biological phenomena through their reproduction in artificial systems. Cognition is a form of biological adaptation and it therefore falls within the province of Artificial Life. However, neural network models and Artificial Life simulations of entire ecosystems tend to address simple behaviors and elementary forms of cognition, while it is often thought that higher level (human) cognition is best accounted for using symbol manipulation models that ignore biology. This type of division of work is rejected and it is argued that Artificial Life should be able to simulate higher level cognition by showing how higher level cognition emerges evolutionarily, developmentally, and culturally/historically from lower level cognition. Three directions of research addressing the emergence of more complex forms of cognition are described using “ecological” neural networks that live in a (simulated) physical environment: (a) networks that learn to predict the consequences of their actions in the environment; (b) networks that are not passive receivers of input but control in various ways the input to which they respond; (c) populations of networks that evolve language to help classifying environmental inputs in useful categories. Other directions that should be explored are the evolution of modular network architectures that can support more sophisticated cognitive abilities and the role of man-made social, cultural, and technological environments in shaping the form of cognition typical of modern man.