This paper discusses why intelligent embodied agents - concentrating on the case of graphically-embodied agents - require affective systems. It discusses the risks involved in attempting to produce naturalistic expressive behaviour, and examines the specific case of facial expression generation. It considers the approach taken by the EU Framework 5 project VICTEC to facial expressions and then discusses some of the issues in affective architectures needed to drive facial expressiveness in an autonomous Intelligent Virtual Agent (IVA).
展开▼