Seungbeen Lee

[Google Scholar] [CV]

I am an external collaborator at Carnegie Mellon University, where I work with Yonatan Bisk. I received my M.S. from Yonsei University, advised by Youngjae Yu, and earned bachelor’s degrees in Psychology and Economics at the same university. I am interested in building long-term, healthy relationships between people and human-like social robots to help people feel happier and less bored. I also remain fascinated by how humans think and interact, especially in social psychology, evolutionary psychology, and game theory.

Robot Motion Semantics

When interacting with robots, people infer meaning from a robot’s capabilities, perception, abstractions, generalizations, and intentions. This is natural and inevitable, given how keenly humans read others’ nonverbal cues. I am therefore interested in building foundational theories and automated tools for generating robot motions that people can intuitively interpret—not only motions that accomplish a task, but also ones that communicate the robot’s goals, methods, and internal state. My first testbed is the generation of iconic gestures across diverse robot morphologies using large language models. More broadly, I am interested in open-vocabulary intent communication in task settings.

Personality Building

Intelligence and competence allow us to predict an agent’s behavior and form expectations even in unfamiliar contexts. Personality goes one step further: when multiple behaviors can successfully accomplish a task, personality provides a latent structure that shapes which behavior the agent chooses. Robotics research currently introduces personality through human-like traits, such as preferred interpersonal distance or speaking patterns, but usually only in a limited set of contexts. I believe personality can provide a foundation for building deep trust in robots, and I want to develop automated methods and evaluation tools for designing coherent robot personalities across contexts.