Conversational robots play a pivotal role in conversational recommender systems, offering recommendations that resonate with the user. In this study, we focused on the characteristics of humans who are easily affected by others to obtain agreeable recommendations. We assume a situation where a recommender robot suggests a country while conversing with a human participant, and captures images that simulate scenes from the selected country. Simultaneously, a human or a robot side participant attends the conversation. We investigated the effects of a side participant and the mention of other's preferences on improving the acceptability of the robot's recommendations. We conducted a field experiment in a shopping mall to evaluate the acceptance ratio of recommendations and the impressions of conversations. Over a span of 18 days, 571 groups (787 visitors) participated in the experiment. Our results suggest that mentioning the preferences of both human and robot side participants during the recommendation process can improve the acceptance ratio. Furthermore, the questionnaire analysis suggests that even when recommender's suggestions do not align with the target person's preferences, the robot can still maintain high levels of satisfaction and agreement. This study underscores the potential of conversational robots in enhancing user experience in conversational recommender systems.
Read full abstract