In a human-agent team, the performance of each human/agent is only partially known initially, and revealed over time as in the classic multi-armed bandit problem. An ideal team of the future will have this kind of a human-agent setting. Understanding the dynamics of such teams is helpful in the design of efficient teams. With this goal, we conducted experiments to understand the decision-making process in human-agent teams. In the talk, I will primarily address three questions---how are the opinions of individuals integrated into the group decision, how does the team rely on agents, and how is an agent’s response integrated into the team’s decision. I will present some models used to represent the decision-making process including Bayes rule, centrality, and prospect theory. Finally, I will discuss the results and the inferences we draw from our analysis about team behaviour.