My research focuses on the interplay between performance and social considerations in human-AI joint decision-making. Performance considerations can include optimizing for decision quality, individual scores, or travel time to destinations. Social considerations may involve maintaining and improving user perceptions, teamwork, and the overall interaction experience through compromise and cooperation. In interactive, interdependent, and contextual joint decision-making processes, each action taken by an AI agent can impact both performance and social outcomes. In this dissertation, I present three projects that explore how optimizing and compromising actions can influence the joint decision-making process, and offer insights on how an agent might balance the two.