Evaluating images of virtual agents

Abstract

This study examined the perceived attributes of virtual agents, based on their appearance. The aim was to determine the nature of the psychological processes that influence judgements. It was expected that many of the usual characterisitics of stereotypical judgements would influence attributions made with regard to computer agents. Fourteen agents from a population of 150 were used to obtain similarity judgements, preference data, dimensional analysis, and personality judgements. Multidimensional Scaling analysis suggested that computer agents were categorized according to 2 main dimensions (gender and anthropomorphism). The clustering of agents were across dimensions predicted attributions made, with more positive attrbutions made to female agents, less positive to cartoon agents, and the most negative responses to male agents.