Robots As Socially Intelligent Agents

We briefly consider thae nature of a social agent, snmmarise work carried out at Salford, and consider the communication requirements for different levels of socila organisation. We examine the ngineering implications for robot experienmts of these communication requirements. What are social agents? We consider social agents in general to be those involved in interaction between same species individuals for extended periods in common locations. In nature social interaction is mediated by life events and tasks: mating and reproduction together with subsequent care of young, food gathering, protection against predators including sometimes construction of protective artefacts. But social interaction may not just be a question of external tasks: the social domain, once existing, may develop its own internal structure and relationships often resulting in a social hierarchy expressed and mediated through particular social behaviours and rituals. Social intelligence is then the ability of the individual to thrive in a social environment, both from the point of view of life tasks (e.g. successful reproduction) and within the social domain itself (e.g. high or rising social status). When we consider artificial agents, whether entirely software based or also embodied in hardware as with robots, we may continue to approach social interaction from a agent-centred perspective as described above. However there is now an alternative point of view: the integration of such agents into a specifically human social milieu. While agent-centred perspectives are usually concerned with theoretical investigation, the human-centred perspective may also have a pragmatic focus in terms of meeting specific human purposes. Moreover the ’interaction between same-species individuals’ basic to the agent-centred perspective may be completely absent, with the productive analogy being less natural animal or insect societies and more pets or other domesticated animals. Social Agents at Salford Work has been carded out for a number of years at Salford in the field of cooperative task execution with multiple robots (Aylett et al 97, Barnes et al 97) in which robots carry out cooperative object relocation tasks within a multi-agent architecture incorporating behavioural architectures on the robots and a fixed predictive planner agent supervising the task at an abstract level. This type of multi-robot system is aimed at industrial environments, but what is often forgotten here is that these environments are already complex social organisations in which the cooperating robots must also cooperate with human workers. Thus this work correctly seen falls into a humancentred approach rather than a purely agent-centred one. Ruth Aylett Secondly, in the last year the author has worked with colleagues at VUB in Brussels on the origins of language and the grounding of adaptive generation of vocabulary in real world robots (Steels & Vogt 97). This is very much an agent-centred approach in which communicative competence between same-species individuals is the focus. In both cases, communication is a major issue, whether between agents alone or between agents and human.~. If social expertise consists in ’doing the fight thing’, then an important part of this is communicative activity. Underlying this work are two premises, or perhaps more correctly, hypotheses. Firstly that social organisation is supported, developed and expressed by a communicative spectrum all the way from stigmergy (Holland & Becker 96) to human language. At each point on this spectrum there are different requirements in terms of the functionality required of the individual as well as different features in terms of the social expertise supported and expressed. We will briefly discuss these relationships. The second hypothesis is that the behavioural substrate we think of as lying at the more primitive end of this spectrum is essential to the conscious activity we locate at the advanced end, not only historically but also functionally. This hypothesis has a number of practical consequences. One is that embodiment in the real world is vital to experimental activity in spite of the difficulties involved in this and that simulations may on occasion be misleading. Another is that making a service robot ’sociable’ requires more than merely giving it a formalised body of social knowledge at a symbolic level. It sometimes appears that workers in the field identify social expertise with cooperative task execution. A functional analysis of what a social organisation is ’good for’ may support this approach, and at an intuitive level it is clear that ’many hands make light work’ has been a powerful force in human social organisation. A clear functional role can also be identified for communication systems, from the alarm call of a blackbird to the planning of a family day out. Yet it is important to remember as argued above that social expertise may also be reflexive, that is aimed at the social domain itself. Grooming in primate societies, it is argued (de Waal 82), has less to do with personal hygiene than with status, social hierarchy and social differentiation. Indeed it has been argued (Dunbar 93, Aiello & Dunbar 93) that the origins of language and of consciousness lie in the social domain itself rather than in cooperative task execution, that language evolved as a more efficient form of grooming and that consciousness was based on the evolutionary advantage in a social group of being able to ’put yourself in someone else’s shoes’. From: AAAI Technical Report FS-97-02. Compilation copyright © 1997, AAAI (www.aaai.org). All rights reserved.