The Effect of Dominance Manipulation on the Perception and Believability of an Emotional Expression

Abstract

Models of affect are used in virtual characters to predict the emotions that can be shown by the character and thus to increase the believability of the character. In some specific situations it may not be clear which appraisals are the most important and thus which emotion should be generated. For example, both anger and sadness can be shown if another person does something blameworthy that is negative for one’s own goals. Based on experimental and theoretical findings in emotion psychology, we propose a model using social dominance as a way to choose between anger and sadness. We hypothesize that anger should be generated (and expressed) in the dominant virtual character and sadness in the non-dominant character. We test this hypothesis with a virtual reality scenario in which a user and an agent negotiate about job options. The negotiation always fails as a result of actions of the user. We have a $2 \times 2$ experimental setup with agent role (dominant/submissive) and expressed emotion at the end of the scenario (angry/sad) as factors. No significant effect on the believability measure between the different conditions was found so the hypotheses cannot be confirmed. A significant influence of agent role was found on the perception of the emotional expressions, showing that social context influences perception of expressed emotions.