Transparency in Human-agent Teaming and its Effect on Automation-induced Complacency

Abstract

The battlefield of the future has been envisioned as one soldier operator managing a team of robotic assets to conduct multiple concurrent tasks, and the DoD has been actively investigating the potential of such human-agent teams. Contemporary research shows that one operator managing multiple robotic assets suffers from a variety of performance decrements. Using an intelligent agent as the mediator of the robotic team helps alleviate some problems, while introducing several unique to the supervisory relationship. One such problem is the human-out-of-the-loop condition, which often results in an increase in operator complacent behavior. This proposed study explores how operator knowledge of the work environment and access to the agent's reasoning affects complacent behavior. Additionally, the interaction of operator knowledge and agent reasoning will be explored to see how the presence (or lack thereof) of each affects operator performance, workload, and situation awareness.