The centrifugal development of artificial agents: a research agenda

This paper presents a research agenda underlying a PhD proposal on embodied software agents. The research aims to demonstrate that it is possible, for embodied software agents, to develop first-person meanings for environmental perturbations. According to the proposed approach, first-person meanings will be created through the agent centrifugal development. We argue that centrifugally developed agents, which create first-person meanings for environmental perturbations, will perform better in identified classes of tasks. They will also contribute to advance the state of the art regarding embodiment, particularly with respect to the grounding problem. For traditional AI, the mind is a symbol processing system that can exist without the body or act in an independent way. Although traditional approaches have successfully solved several problems, they are trapped in the grounding problem. Embodied cognition approach departs from the dominant paradigm by focusing on cognition as an embodied situated activity. Organismic embodiment supporters argue that the systems development method (centrifugally versus centripetally) is decisive to the system autonomy and, thereby, to their cognitive abilities. This research will use models of cellular division and evolution from cellular biology and will propose processes, through which meaning grounding can be achieved as an emergent property of centrifugally developed agents.