Human Satisfaction in Ad Hoc Human-Agent Teams.

With recent progress in connectivity and the transition into knowledge economies, mixed human and agent teams will become increasingly commonplace in both our personal and professional spheres. Hence, further examination of factors that affect human satisfaction in these types of teams can inform the design and use of effective human-agent teams. In particular, we are interested in virtual and ad-hoc team scenarios where a human is paired with an agent, and both need to assess and adapt to the capabilities of the partner to maximize team performance. We designed, implemented, and experimented with an environment in which virtual human-agent teams repeatedly collaborate to complete heterogeneous task sets. We investigate the role humans and autonomous agents play, as task allocators, on human satisfaction of virtual and adhoc human-agent repeated teamwork. We evaluated the satisfaction of human participants, recruited from the Amazon Mechanical Turk platform, using survey questions focusing on different aspects of team collaboration, including the process, allocation strategy, team performance, and agent partner. We find that participants are more satisfied (a) with the process when they allocate and (b) with the team outcome when the agent is allocating. We analyze and identify the factors that contributed to this result. This work contributes to our understanding of how to allocate tasks in virtual and ad hoc human-agent to improve team effectiveness.