Analysis of Human and Agent Characteristics on Human-Agent Team Performance and Trust

The human-agent team represents a new construct in how the United States Department of Defense is orchestrating mission planning and mission accomplishment. In order for mission planning and accomplishment to be successful, several requirements must be met a firm understanding of human trust in automated agents, how human and automated agent characteristics influence human-agent team performance, and how humans behave. This thesis applies a combination of modeling techniques and human experimentation to understand the concepts aforementioned. The modeling techniques used include static modeling in SysML activity diagrams and dynamic modeling of both human and agent behavior in IMPRINT. Additionally, this research consisted of human experimentation in a dynamic, event-driven, teaming environment known as Space Navigator. Both the modeling and the experimenting depict that the agents reliability has a significant effect upon the human-agent team performance. Additionally, this research found that the age, gender, and education level of the human user has a relationship with the perceived trust the user has in the agent. Finally, it was found that patterns of compliant human behavior, archetypes, can be created to classify human users.