What is Natural?: Challenges and Opportunities for Conversational Recommender Systems

Abstract

This poster describes a fundamental user research to uncover the challenges and opportunities for a conversational recommender system. The study was conducted in three different environments: home, work and street. During the sessions, the participants (i) recounted their previous experiences of interacting with conversational agents in a critical incident interview, (ii) performed a task to explore restaurant recommendations with Google Assistant, and (iii) discussed their views on conversational recommendations. We organize the challenges uncovered into three levels: environment, conversation, and system. Further, we illustrate the identified opportunities of whispering interaction, collaborative recommendations, driving scenario, and local-like conversations for delivering recommendations.