A cognitive model of social relations for artificial companions

Abstract

Artificial companions are made to establish and maintain long-term relationships with users. In order to model the companion's social relations and to capture its dynamics, we propose a neural network model based on a formal representation of social relations. Based on psychological theories, we characterize social relations over two dimensions, namely liking and dominance. These two dimensions are formally described as a combination of beliefs and goals of the agent's mental state. Such a model allows us to automatically compute the social relation of a virtual agent towards its interlocutor depending on its beliefs and goals. © 2014 Springer International Publishing Switzerland.