Towards An Agentic Conversational Approach to Group Recommender Systems
Group Recommender Systems (GRSs) have yet to achieve widespread practical adoption, despite being an active area of study. This research aims to deepen our understanding of group decision-making by developing a conversational GRS embedded directly into an existing chat application in the form of a bot. We investigate how group dynamics and user roles influence the outcome of group decisions and how a bot can support group discussions through an agentic framework that continuously adapts its behavior based on interaction signals received from the group. To this end, we conduct a series of user studies to empirically evaluate the system across different decision-making domains, group sizes, and user demographics, with the ultimate goal of developing a conversational GRS framework that facilitates the decision-making process and guides groups toward more balanced and mutually satisfactory decisions.