Patterning Behaviors for Human-Robot Teams

Abstract: Effective human-AI teaming is hindered by a fundamental conflict between machine-centric optimization and human-centric cognition. While algorithms can generate high performing behaviors, these actions can be unpredictable to humans, compromising team fluency and safety. This 'predictability gap' represents a critical barrier to the widespread adoption of intelligent systems in collaborative settings. My work seeks to narrow this gap by integrating human cognitive strengths into robot behavior. Results indicate that this approach improves human perceptions of the robots they work with, as well as improving team performance on a variety of collaborative tasks.

Bio: Clare Lohrmann is a doctoral candidate at the University of Colorado Boulder, mentored jointly by Drs. Bradley Hayes and Alessandro Roncone. Her research is in the area of human-robot collaboration, where she develops algorithms that alter robot behavior to make robots more predictable to the people who interact and work with them. Her work draws from a variety of fields both inside computing and beyond, such as cognitive science, trajectory optimization, reinforcement learning, and behavioral economics. Outside of the lab, she has a small hobby farm and raises heritage and endangered poultry. In the fall, she will begin a tenure-track role at Western Washington University in Bellingham, WA.