User Evaluation of a Conversational Recommender System

Conversational recommender systems (CRSs) approach user preference acquisition from a conversational point of view, where preferences are captured and put to use in the course of on-going natural language dialogue. The approach is motivated by its aim to make interaction efficient and natural, to acquire preferences from the user in a context when she is motivated to give them, as well as to facilitate exploration of the domain and the development of the user’s preferences. A CRS’s dialogue strategy to achieve these aspects of the interaction is crucial for its performance and usability. This paper reports on a user satisfaction evaluation of ACORN, which is a CRS in the movie domain. The results of the study indicate a high user satisfaction with the interaction from nine usability aspects, and that ACORN’s dialogue strategy is suitable for efficient interaction and user preference modeling, and facilitates domain exploration.