“Tell Tell” - Effects of Personality of Kansei Agents

Abstract

The purpose of this study is to propose an emotion generation model that consists of pseudo-personality and pseudo-moods. Agents implementing this model have various personalities for each user. In order to verify its usefulness, we implemented the model on an agent called “Tell Tell” and performed an experimental performance evaluation using a web application. In addition, we tried to performed an experiment to identify the nurtured agents. From the experimental results of performance evaluations, it was found that the agent of the proposed method had a positive difference in impression between "enthusiasm-lose interest" and "human-like- mechanical". However, in the items of getting tired and fun, there was a significant decrease in subjects with many interactions. From the experimental results of the nurtured agents, it showed that the proposed agent had a higher discrimination rate than the agent withpersonality only. Therefore, we analyzed the interaction of each subject, the agent of the user who was able to discriminate had little change in personality, and the user who couldn't discriminate had changed personality many times. In the future, it is necessary to improve the impression of getting tired and increase the discrimination rate by implementing an algorithm that changes the facial expression depending on the intensity even for the same personality.