Investigating human’s impression of a shared experience-based dialogue robot

Abstract

In this study, we investigated the impressions conveyed by a robot which interacts with a user, and then carries out a conversation about the content of their interaction. In the proposed system, the robot does not converse with the user based on a predefined dialogue scenario, nor does it unfold the conversation interactively based on user’s utterances. It rather generates dialogue contents by reflecting on the events previously experienced with the user. In the first phase, the user and the robot virtually share a place and time in a scenario where users are successively presented with a variety of T-shirts designs that they rate according to their preference or sensitivity. Through this interaction, the robot captures the user’s preference using Kansei Retrieval Agents. In the second phase, the robot carries out a conversation with the user based on the preference information gathered during the first phase. During the experiment, we asked the participants to interact with two robots, one embedding the proposed reflective conversation, and a comparison robot which conducted dialogue based on a predefined scenario. Obtained results seem to indicate that the robot equipped with the proposed system was more able to convey a friendly impression to the subjects.