Enhancing conversational access to information through a socially intelligent agent

Abstract

Intelligent access to information could benefit of an effective and natural interaction metaphor. In this perspective, Embodied Conversational Agents (ECAs) can be seen as a promising approach to give to the user the illusion of cooperating with a partner rather than just using a tool. Embedding the HCI technology with human preferences and behavior justifies the attempt of implementing emotional and social intelligence aimed at exceeding the single ability to help the user. In this paper we present an ECA's architecture and methods useful to interpret the user attitude during her dialog with an ECA and behaving 'believably' in its turn. In particular, we present an agent architecture that is general enough to be applied in several application domains and that employs several ECA's bodies according to the context requirements. © 2010 Springer-Verlag Berlin Heidelberg.