Assigning a personality to a spoken dialogue agent through self-disclosure of behavior

In this study, we propose a method to assign a personality to a spoken dialogue agent and evaluate effectiveness of the method. Recently, some research studies on human-agent interaction (HAI) showed that it is possible to assign a per- sonality by controlling non-verbal information of an agent. However, we consider that verbal information sent from an agent would result in the agent developing a personality. In this paper, we focus on self-disclosures about behavior as an important factor of assigning a personality, and analyze the effects of self-disclosures on assigning a personality to the agent by evaluation of impressions. We discovered that self-disclosures performed by the agent are as effective for personality recognition as those performed by humans. Fur- ther, self-disclosures allow a user to recognize an agent's per- sonality, and humanlike self-disclosure does not impede per- sonality recognition even if utterer of these self-disclosures is an agent. Based on these facts, we verified effectiveness of our proposal method to assigning a personality by self-disclosures about behavior.