Can a Virtual Listener Replace a Human Listener in Active Listening Conversation?

The elderly who live alone are increasing rapidly in these years. For their mental health, it is reported useful to maintain their social life with others. Active listening is a communication technique that the listener listens to the speaker carefully and attentively by confirming or asking for more details about what they heard. Our project is aiming to develop a computer graphics animated agent who can engage active listening dialog with the elderly users. However, there is a research issue exists in this task: whether the CG animated virtual listener can be really perceived as a human active listeners even if they perform the same behaviors. This paper presents the experiment results in investigating this issue. We conducted an experiment which collects active listening teleconferences in two conditions, human-human and human-agent. In the human-agent condition, the human listener's video is substituted with a CG character performing the same behaviors. We then compared the impression in these two conditions base on the evaluation done by the participants themselves. The results showed that generally there were no significant differences between these two conditions. This means that a virtual listener can engage active listening at the level similar to human listeners if their behaviors are appropriately modeled.