A Framework for Human-Autonomy Team Research

On a team, autonomy must be able to work alongside human counterparts and carry out the fundamentals of teamwork and taskwork. In this paper we refer to these machine teammates as autonomy. These HumanAutonomy Teams (HATs) need to be assembled to have the appropriate roles and responsibilities and to interact in an interdependent manner. One challenge in assembling an effective Human-Autonomy Team involves doing research on human-autonomy teaming that can provide input to autonomy development BEFORE the autonomy is developed. We propose here a five-step process to doing HAT research and provide four examples of the application of this process. The five steps involve 1) knowledge elicitation to determine the essential aspects of HAT in a given domain, 2) development of a synthetic task environment with Wizard of Oz capability, 3) development of measurement strategies, 4) human subject experimentation, and 5) translation to the developers of artificial intelligence and robots that will serve as teammates.