ReDBot: Exploring Conversational Recommendation for Decision-Making Support in Group Chats

Group Decision-Making (GDM) commonly takes place online, e.g., in text-based group chats, for daily tasks like choosing a movie or a restaurant. However, reaching a consensus among members in GDM tasks online is non-trivial due to the high workload of collecting necessary information and low awareness of group preferences. In this paper, we explore the design and impact of conversational recommendation for GDM support. Inspired by theories of GDM, we propose a ReDBot that asks questions to identify the group preference and recommends alternatives that match the group preference. We power ReDBot with recent large language models to handle the conversational flow. Our preliminary user study with four three-member groups suggests that ReDBot could reduce members' workload in collecting information, improve awareness of group preferences, and boost consensus-reaching in GDM group chats. We conclude with design considerations for GDM support.