Understanding Autonomous Interaction

Autonomy is a necessary part of the design of agents flexible enough to function effectively and efficiently in a sophisticated world. Much work, however, has taken a very restricted view of what is entailed by autonomous interaction; in particular, the effects of an interaction have, to some extent, been guaranteed. In this paper, we argue that no facet of interaction can ever be guaranteed, and that if agents are to be autonomous, they must be able to cope with this inherent uncertainty. We propose a model of autonomous interaction in response, which addresses these concerns, and which can be viewed as a process of motivated discovery. This approach has two important aspects: first, modelling the motivations of the agent allows a more adequate model of autonomy to be achieved, and also provides a control strategy for the process of interaction; second, the discovery paradigm provides a suitable framework for effective action and reasoning in an uncertain environment.