Evaluating Human Perception of Autonomous System Teammate-likeness

The paper presents preliminary results on an evaluation of the perception of teammate-likeness in Human-Autonomy Teaming (HAT). Increases in an autonomous system’s core teaming skills could increase human perception of the teammate-likeness of an autonomous system. In many HAT systems, the autonomous systems have lacked humanness (or human-like characteristics), based on the technology developed to the date. This may affect how much humans perceive the autonomous system as a teammate, rather than just a tool. Increased perception of teammate-likeness would enable human and autonomous systems to work together more closely and may enable the autonomous system behavior to more closely match the human’s expectations of a teammate’s behavior, benefiting coordination and cooperation. Previous work with commercial flight pilots showed the need for different feedback modality requirements of an autonomous teammate. Pilots felt autonomous teammates should provide visible cues of actions in the work environment, in addition to verbal communication to provide a sense of presence in the cockpit. These results motivated the present study to investigate the impact of feedback modality on teammate-likeness of autonomous system. The evaluation study assessed the impact of centralized feedback vs. distributed feedback of a Cognitive Assistant working with a human to manage the outbreak of a fire aboard a simulated Mars station. The Cognitive Assistant and human teamed to manage both sequential and concurrent fires, thereby manipulating task demands. The study showed that human perception of autonomous system teammate-likeness increasing with distributed feedback model due to increased awareness on CA’s actions.