Descriptor: Appraising Trust in Interactions Between Human Supervisors and Autonomous Robotic Systems (ATHAI)

With the ongoing development of autonomous systems, significant attention has focused on deploying autonomous robotic systems to complete a task under human supervision. An important pillar of this interaction is a notion of human trust in the autonomous system, however the modeling of trust in such scenarios has been hampered by the lack of a standard open-access dataset with longitudinal measurements of trust. By recording measurements of trust at a single point in time (often at the end of interactions), these datasets do not capture the temporal dynamics of trust over the course of interactions. We propose to address these challenges with a novel multi-modal dataset that aggregates autonomous system status and task performance, self-reported human trust, and human interventions in the autonomous system during interactions. We collect the data using a software-based interface that simulates a swarm of autonomous robotic agents performing a foraging task in a rectangular area. By recording metrics pertaining to the autonomous robotic swarm’s task performance and human trust, the dataset provides a foundation for the modeling of human trust dynamics and their influence on human interventions. In addition, the dataset’s inclusion of timestamped human responses to a secondary task permits analysis of human attention allocation between tasks over time. It is anticipated that the dataset will provide a benchmark for human trust models, facilitating the development of more trust-aware autonomous robotic systems. IEEE SOCIETY/COUNCIL Robotics and Automation Society (RAS).