Artificial companions

Abstract

AbstractAbstractThe paper concerns the closely related topics of the possibility of machines having identifiable personalities, and the possible future legal responsibilities of machines. As will become clear, I wish to explore these topics in terms which are basically those of computational linguistics; and by 'machines' here, I intend software entities, rather than robots, and in particular the sort of software agents now being encountered on the web, ranging at present from technical advisers to mere chatbots. I call them Companions. The body of the paper explores the following related aspects of a Companion in more detail: what kind and level of personality should be in a machine agent so as to be acceptable to a human user, more particularly to one who may fear technology and have no experience of it; and what levels of responsibility and legal attribution for responsibility can we expect or desire from entities like web agents in the near future?