Exploring the Socially Adept Agent

Imagine an agent who is charged with seeking information from around the globe. It is the task of this agent to visit different sites, to search for information, and to report back to its 'superior' (which may or may not be human). In the course of this search, it will interact with others, human or otherwise. It will make decisions about the situations it finds itself in, based on, for example, the behaviour of others, its own 'knowledge', and the state of the environment as it sees it. In these decisions, the actions of others will play a large part. Artificial agents do not exist in the world in solitude. As a result, they will be required to interact with others, and reason about the behaviour of others. These interactions will become increasingly important and complex as the agents become more sophisticated, and more so when we expect them to interact with human agents. This paper introduces and discusses the concept of 'social adeptness' in artificial agents. Social adeptness, the ability to exist and interact in a complex heterogeneous society of agents, involves agents reasoning about (amongst others) communication (verbal and non verbal), the emotions, the trustworthiness of others, the morality and ethics of interactions and behaviours of themselves and others. This paper draws from work in philosophy, psychology and sociology to present a unified view of what is necessary for social adeptness. In addition, it presents and discusses what has been done so far to allow artificial agents to carry out sensible and meaningful interactions in society. This includes the development of Artificial Morality, research into emotional displays in agents, and the development of a formalism for the concept of trust in others. The aim is to develop a formal view of social adeptness.