Communicating Robot Learning across Human-Robot Interaction

Today’s robots are increasingly able to learn. From robot arms that learn from experience to autonomous cars that learn from demonstration, these systems fundamentally change their behaviors over time. While this learning improves the robot’s performance, it also presents a black box to nearby humans: what has the robot learned correctly, what is the robot still confused about, and how will the robot behave in the future? For seamless human-robot interaction we must communicate the robot’s learning to human partners.

This workshop explores communication by bringing together diverse experts from three perspectives. From a robot learning perspective, we must understand how to create algorithms that are inherently explainable and intuitive to human collaborators. From a communication interfaces perspective, we must develop haptics, augmented reality, nonverbal, and soft robotics mechanisms that convey information from the robot to the human. Finally, from a human modeling perspective, we must understand how humans perceive and interpret this feedback to construct mental models of learning robots. The main objectives of this workshop are i) to understand how each of these communities is separately trying to communicate robot learning, and then ii) to discuss how these separate advances should be combined in the future.