Seungbeen Lee

[E-mail] [Github] [Google Scholar] [CV]

Hi, I’m Seungbeen, an MS graduate at Yonsei University advised by Youngjae Yu. I studied Psychology and Economics during my undergraduates at Yonsei. I still love discussing about Personality Psychology, Social Psychology, and Game Theory. I’ve always been fascinated by modeling human decision-making, and have insights that humans are significantly influenced by the presence and decision-making of others.

What Can Language Do for Robotics?

Robots are taking on tasks with increasingly high-dimensional action spaces and increasingly numerous constraints. Greater complexity demands higher-level abstractions. Language is one of humanity’s most powerful inventions for compressing experience into concepts and composing those concepts into new meanings. I am interested in two roles for language in robotics: (1) language-grounded motion primitives that translate high-level intentions into reusable behavior, and (2) personality as a compact model of traits and preferences that makes an agent’s behavior coherent and predictable across contexts.

What Can Robots Do for Language and Social Interaction?

Embodiment fundamentally changes what it means to participate as a social agent. I am interested in anthropomorphism not as a matter of humanlike appearance, but as an inference elicited by the richness and consistency of movement and decision-making. In the Heider–Simmel experiment, viewers readily attributed intentions, emotions, and relationships to simple geometric shapes based on motion alone. Similarly, a robot’s gaze, posture, trajectory, distance, and timing can give language situated social meaning. I want to understand which minimal embodied cues create social presence—and how robots can use them legibly without misleading people about their capabilities.

Superhuman Social Intelligence

Must artificial social intelligence be bounded by human ability? Today, we often treat human annotation as the gold standard for the complex, high-dimensional problem of social interaction. Yet people are bounded by their own biases, cognitive limitations, and perceptual resolution. Future robots may integrate fine-grained behavioral and physiological signals, potentially functioning as near-perfect lie detectors even in everyday conversation. I am especially interested in what could become fundamentally different in machine-to-machine sociality. Humans pay a cognitive cost for maintaining every social relationship—a constraint reflected in Dunbar’s number—and our communication depends heavily on language because we cannot directly access another person’s mind. Robots, however, are not shaped by the same evolutionary pressures and may communicate at negligible marginal cost, without language or theory of mind, by exchanging internal representations directly.