Robot Societies

Abstract for a talk at Robotix97 by Kerstin Dautenhahn


What is a "robot society"? From our own experience we know quite well what a "human society" is, namely not just a collection of individual human beings who gather accidentally at a certain point in time and space. It's rather a structured group. The smallest group which we know is the "family", its members have close contacts and long-term relationships. The same individuals can belong to several other groups, at different levels, up to what we call "society". In a society, in addition to one-to-one contacts we find one-to-many, many-to-many etc. interactions and communication between members. The common point in what we call "group" or "society" are common goals, conventions, interests. Cooperative behavior, division of labour, traditions and culture (e.g. educational systems) could only develop in groups of individuals. They result in new patterns of life and living (thinking and behavior). Therefore, I only like to apply the term "robot society" to groups of robots which share common interests, which cooperate and communicate in order to 1) increase the survial chances of the individual ("selfishness"), 2) increase the "fitness" of the group(s) which they belong to, as an entity. Additonally, in "family-like" small group structures, helping and caressing occurs without any obvious "benefit", it can (at least for humans) has a quality in itself (e.g. friendship).
In my talk I will outline possible applications where groups of robots can cooperatively solve a task better, or survive longer, than one single robot (see the biological model of social insects, bees, termites etc.). Moreover, I will introduce my own research on robot societies, which is much inspired by work in biology and developmental psychology on social behavior, i.e. it is more inspired by mammal-type societies. My work on "artificial social intelligence" focuses individual relationships" between robots in a hilly landscape, including topics like recognition of conspecifics and imitation. The realization of these systems is along the "artificial life" direction, e.g. I am studying how local interactions between simple components can result in complex structures on the next, more global level (e.g. a complex pattern of behavior). The big challenge is to find mechanisms which exploit the characteristics of complexity in artificial matter. I outline how this can be studied with a group of small robots in the "hilly landscape scenario", where small robots can move up and down hills, recharge batteries, communicate and recognize each other.

Lit, e.g.:
K. Dautenhahn (1995) Getting to know each other - artificial social intelligence for autonomous robots. Robotics and Autonomous systems

K. Dautenhahn (1997) I could be you - the phenomenological dimension of social understanding, Cybernetics and Systems