Maia Stiber (John Hopkins)- Human-Aware Robots: Social Signals in Human-Robot Interaction

Abstract: Effective human-robot interaction requires robots that can interpret and adapt to human behavior in real-time. My research advances human-aware robotics by using social signals to understand human internal states and models. I examine human reactions to misalignment—moments when users’ mental models about the robot, task, or world diverge from reality. Misalignments can cause interaction breakdowns, but also provide useful information about task context and user perception. I apply this approach in two areas: providing robots with error awareness and with user confusion awareness in physically situated tasks. To achieve this, I build models that use behavioral signals to detect errors and user confusion, develop proactive, multimodal interactive robots to engage users in the error detection process, and demonstrate that behavioral signals—especially when contextualized—can be used for flexible detection of these misalignments. This research is a step towards deploying human-aware robots that can adapt to user behaviors, improving performance, reliability, and acceptance of interactive robots in everyday environments.