Sougwen Chung matters to movement-and-computation research because her work treats the machine as a situated collaborator rather than a neutral output device. Her public record documents a practice in which robots draw alongside her, trained on traces of her own mark-making. The work offers a concrete way to discuss agency, feedback and embodied data without pretending that a model has a human interior.

Context

Chung is an artist and researcher whose practice sits between drawing, robotics and machine learning. Her 2019 TED talk, “Why I teach robots to paint with me,” describes the motivation for building systems that learn from her gestures rather than simply executing prewritten commands. Her website documents the DOUG and OMNI series and related exhibitions. These sources establish the public facts used here; no private interview or invented quotation is implied.

The public record

The DOUG systems are robotic drawing companions trained on Chung’s own marks and gestures. In the documented setup, the machine’s output is not a finished replacement for the artist. It becomes another mark-making event that can be observed, answered and folded into a continuing practice. OMNI extends the inquiry through robotic and generative systems that explore how a body’s archive can become a source of machine behaviour.

This distinction is important. “Trained on herself” does not mean the machine has acquired Chung’s consciousness. It means that a selected history of traces, movements and decisions shapes a statistical response. The artwork makes that mediation visible instead of hiding it behind a seamless image.

Critical engagement

Chung’s work is a useful counterpoint to current EEG-conditioned or pose-conditioned systems. A sensor can supply a control signal; a generator can supply variation; neither alone establishes agency. In DOUG, agency is negotiated through repeated feedback between a person, a machine and a material surface. The drawing records the relation, not only the model’s prediction.

The approach also raises a data question. A personal archive can make a system more specific, but specificity is not the same as understanding. What the machine learns depends on which gestures were captured, how they were segmented and which outputs were accepted. A somatic system would need to preserve those choices and allow the practitioner to reject an interpretation without treating rejection as noise.

Field significance

Chung’s contribution is methodological as much as aesthetic. She shows that a human–machine practice can be organised around co-presence, trace and response rather than prompt and answer. For movement research, that suggests evaluating a system over an encounter: latency, surprise, adjustment and the participant’s ability to redirect the process. It also supports a modest claim about authorship. The machine may generate marks, but the practice includes the conditions, selections and feedback through which those marks become meaningful.

The archive’s profile of choreographic robotics and discussion of body-specific calibration extend this field context. Chung’s public work does not solve those problems; it gives them a materially legible form.

References

Chung, S. (2019). Why I teach robots to paint with me [Video]. TED. https://www.ted.com/talks/sougwen_chung_why_i_teach_robots_to_paint_with_me

Chung, S. (n.d.). Sougwen Chung. https://sougwen.com/

MIT Media Lab. (n.d.). Sougwen Chung. https://www.media.mit.edu/people/sougwen/overview/