Gordon Beavers

In Search of Simple and Responsible Agents

An artificial agent is a computational entity (embodied or otherwise) that interacts with other agents and/or real-world entities. Being reactive is a standard property of agents. The agents considered here exhibit various degrees of sociability in the form of norms, roles, values, cooperation, motives, responsibilities, autonomies, and rights. Intentional agents have been modeled in multi-modal BDI logics, e.g. [8], with operators for belief, desire, and intention. This paper proposes the integration of social notions into BDI agent architectures to account for social decision-making. Although a large collection of notions is needed to explain the actions of complex social agents, this paper provides a somewhat simplified model of social agents built on a small set of agent properties (norms and values) and intentional notions (obligations). Since this model is a starting point for the investigation of social agents, it is expected that the model will be improved and expanded as the result of further research.

Types and limits of agent autonomy

This paper contends that to fully understand interactions between agents, one must understand the dependence and autonomy relations that hold between the interacting agents. Individual variations, interpersonal dependencies, and environmental factors are determinants of autonomy that will be discussed in this paper. The paper concludes with a discussion of situations when autonomy is harmful.