David C. Jeong, Ph.D.
EBODIED DYNAICS LAB EBODIED DYNAICS LAB

NTU Embodied Dynamics Lab

We study how machines see and model the moving human body, and where that breaks down. Our ground is embodied human–AI interaction, including social presence and avatar identity in social VR. The lab is based at the Wee Kim Wee School of Communication and Information, Nanyang Technological University.

How people come to know a made thing as someone.

How do people interpret intelligent machines?

People read a machine before it does anything. Its look, its movement, and how close it stands all shape whether they treat it as a tool or as a social partner. We study how those early readings carry into the interaction that follows.

scroll ↓

It begins as a word: kinematic.

A body is a set of relationships.

It takes up space.

And it moves. It greets you.

Fig. 4One body as a word, as 2D joints, as a 3D shape, and as motion over time.

How does a body change a relationship over time?

A body changes what an AI system can communicate. Gaze, posture, and distance can make a machine feel socially present, and can pull a person’s own identity into the exchange. We work across robots, virtual agents, and avatars, and we track how cooperation and identity shift when the same pairing meets again.

How can movement be measured?

Motion capture, markerless pose estimation1, egocentric video, and computer vision can show how two bodies coordinate second by second, so a relationship reads as a trajectory rather than a single score.

From pixels to people in motion.

Explore the step from a flat image to a body moving through space and time.

Fig. 5A photograph, its 2D joints, a 3D body, and that body over time. The photograph is AI-generated and the model is hand-animated; nothing here was estimated from the photograph. Adapted from Cesium Man (Cesium), CC BY 4.0, with changed proportions, materials and pose traces.

Methods and research environment

We run controlled communication and HCI experiments, and we work with computer vision, XR, motion and pose data, and simulation. Rudolf Laban’s term “kinesphere,”2 the space a body can reach, is how we think about a body acting in a room or in a headset.

The Kinesphere A finely ticked reach boundary (the volume of space a body can act within) with the coupling lens at its centre.
Fig. 6The kinesphere, the space a body reaches without stepping.

Selected evidence

Publications the current work builds on.

Embodied movement and affect

MoEmo Vision Transformer infers emotional state from body movement together with the surrounding scene, for human–robot interaction.

Egocentric vision and body reconstruction

Fish2Mesh Transformer3 reconstructs 3D human meshes from egocentric fisheye video, so behavior can be measured from the wearer’s own point of view.

Identity and social presence in XR

The Alt-Self looks at inclusive avatar representation in social VR. Social Presence in Human–Machine Communication sets out the theory of how a machine comes to feel socially present.

Social kinematics and experimental methods

Earlier work on virtual agents, proxemics, and representative experimental design is the behavioral and methodological base for this work.

Frequently asked

What is embodied human–AI interaction?

Embodied human–AI interaction asks how the body and the space around it shape communication with intelligent systems: physical robots, virtual agents, and avatars in VR.

What does the NTU Embodied Dynamics Lab study?

The lab works on three questions: how people read intelligent machines, how having a body changes a relationship with one over time, and how that movement can be measured. It is based at NTU Singapore’s Wee Kim Wee School of Communication and Information.

Join or collaborate

The lab is in its founding stage, and I’m recruiting a lab manager, research assistants, and PhD students to build it with. I’m looking for people interested in embodied human–AI interaction, social presence, identity, cooperation, XR, computer vision, or behavioral measurement. Good fits may come from communication, HCI, computer vision, cognitive science, or design.

Doctoral applications go through NTU’s Interdisciplinary Graduate Programme in Global Asia, where I supervise a project on whose movement can be measured, whose must be authored by hand, and what is lost in the gap. IGP students take two of their six courses in a second discipline, which fits work that sits between communication and computer vision. There is also an option to spend a year at an overseas university. Applications for the August 2027 intake close on .

Email about joining the lab with a short note about the role you’re interested in and the question you want to study.

For conversations about AI in higher education or curriculum, see Technology & Education.

Back to home