Leveraging Variation in Human Feedback Expression to Enable Natural Human Teaching for Robots

With robots increasingly integrated into everyday life, they must engage with a diverse range of users, each exhibiting different interaction behavior. A key area for this engagement is human-in-the-loop robotic systems, where individuals utilize their knowledge in real-world tasks to teach robots. Human teaching is multifaceted, encompassing not only the content of feedback but also the various ways in which feedback is expressed. Oversimplifying this complexity can result in robots misunderstanding human intentions, ultimately leading to reduced task performance and user experience quality. To address the challenge, my work focuses on identifying, modeling and interpreting the variation in feedback expression to support natural human teaching for robots.