Thecla Schiphorst: The Somatic Turn in How We Build Machines
The pioneer who brought somatic practice into human-computer interaction — from choreographing with Merce Cunningham's software to wearable systems that sense the felt body — on why technology must begin from embodied knowing
Long before "embodied AI" became a conference track, Thecla Schiphorst was arguing that computers would never understand movement until their designers learned to understand it from the inside.
Her path is unusual even among interdisciplinary researchers. Trained in dance and somatics, she was a member of the original design team behind Life Forms — the pioneering computer choreography software — and collaborated with Merce Cunningham from 1990 to 2005, supporting the creation of new dance made with the computer. She went on to become a professor at Simon Fraser University's School of Interactive Arts and Technology, where she has spent decades building what she calls a "somatic turn" in human-computer interaction: the argument that movement experience and somatic awareness are not decorative additions to technology design but necessary foundations of it.
Her research group carries a name that is itself a small manifesto — whisper[s]: wearable, handheld, intimate, sensory, personal, expressive, responsive systems.
You were working at the intersection of somatics and computing decades before it became fashionable. How did that begin?
It began with a dissatisfaction. I was trained in dance and in somatic practices, and I was also working with computers early — with Life Forms, with Cunningham. And what I kept noticing was that the technology was being designed by people who thought about the body as a thing to be represented: a skeleton to be animated, positions to be tracked, poses to be recorded. The body as an object.
But that is not what a body is, to anyone who has done somatic work. The body is not an object you observe from outside; it is the ground of experience, the place from which you know anything at all. And I became convinced that if we were going to build technologies that engaged with human movement — and I could see that we were going to build a great many of them — the people building them needed to understand the body the way a somatic practitioner understands it. From the inside. As lived, felt, first-person experience.
That conviction has basically driven everything I've done since. It's not enough to represent the body accurately. You have to design from embodied knowing, not just about the body.
What does it actually mean to "design from embodied knowing"? It sounds lovely but abstract.
It's concrete, actually. Let me give you the contrast.
The conventional way to design a movement-based technology is to decide what the system should detect — a gesture, a pose, an action — and then build sensors and algorithms to detect it. The designer's own body never enters the process except as a test case. The knowledge that guides the design is technical knowledge about signals and classification.
Designing from embodied knowing inverts this. It starts with the designer cultivating their own somatic awareness — actually doing the movement practice, actually attending to the felt qualities of movement, developing what I've called somatic "connoisseurship." Only from that cultivated first-person sensitivity do you then ask what the technology should sense and support. The design is guided by knowledge that lives in the designer's own moving body, not just in their analytical understanding.
This matters because movement qualities that you have not learned to feel, you cannot design for. If you have never cultivated the felt distinction between a movement that is genuinely released and one that merely looks relaxed, you will build a system that cannot tell them apart — because you couldn't either. The sensitivity has to exist in the designer before it can be built into the technology.
How does that connect to the current wave of AI movement systems?
The current systems are extraordinary at what they do, and what they do is mostly still representation — very sophisticated representation, learned from enormous data, but representation nonetheless. They model how movement appears. What I've spent my career arguing is that appearance is not the whole of movement, and that the missing part — the felt, qualitative, first-person dimension — doesn't get added by scaling up the same approach. You don't reach the inside of movement by getting better and better at the outside.
I'm encouraged, though, by the turn toward wearable and physiological sensing — reading the body's signals directly rather than only observing it from outside. That's closer to what I've always thought was necessary. A system that senses the body's own signals is at least reaching toward the interior, rather than perfecting the exterior view.
But — and this is the crucial point — the sensing hardware is not enough by itself. A system that reads muscle signals but is designed by people who have not cultivated their own somatic awareness will still miss what matters, because it won't know what in those signals to attend to. The interior sensing has to be married to interior knowing. The technology and the trained human perception have to develop together.
Your group's work is described as "intimate" and "personal." Those aren't words one usually associates with computing systems.
No, and deliberately so. Most computing is designed for a generic user — an abstraction, an average. Somatic experience is the opposite of generic. It is always this body, in this moment, with this history. The felt sense of movement is irreducibly first-person and particular.
So when we design wearable, intimate systems, we're insisting that the technology meet the person at the level of their actual, particular, felt experience — not at the level of an averaged model of a user. This is technically harder and philosophically necessary. A system that engages your movement should engage your movement, with its particular qualities and history, not a statistical composite of human movement in general.
I think this is one of the deepest challenges for AI movement systems. They are built on averages — enormous averages, learned from millions of examples. And somatic experience is precisely what does not average. The particular is not noise to be smoothed away; it is the phenomenon itself.
What would you say to a young researcher who wants to work at this intersection — somatics and AI — today?
I would say: do the movement practice. Genuinely. Not as background reading but as sustained, embodied training. You cannot design from embodied knowing that you do not have. The temptation, especially now, is to treat somatics as a set of concepts you can read about and then apply. But somatic knowledge is not conceptual knowledge. It lives in a trained body, and training a body takes years and cannot be shortcut.
And then I would say: hold both. Hold the rigour of the technology and the depth of the somatic practice, and refuse to let either one dominate. The failure mode on one side is technically sophisticated systems that are somatically naive. The failure mode on the other side is beautiful somatic ideas that never become anything real. The whole opportunity is in the marriage, and the marriage is hard because it asks you to be genuinely serious about two very different kinds of knowing at once.
That's the work. It has been my work for thirty years, and it has never been more relevant than it is right now.
Thecla Schiphorst is a professor at the School of Interactive Arts and Technology at Simon Fraser University. A member of the original Life Forms computer-choreography design team, she collaborated with Merce Cunningham from 1990 to 2005. Her research on embodied interaction, wearable systems, and the somatic foundations of human-computer interaction has been foundational to the "somatic turn" in HCI. She leads the whisper[s] research group (wearable, handheld, intimate, sensory, personal, expressive, responsive systems).
References
Schiphorst, T. (2009). soft(n): Toward a somaesthetics of touch. CHI '09 Extended Abstracts. https://doi.org/10.1145/1520340.1520345
Schiphorst, T. (2011). Self-evidence: Applying somatic connoisseurship to experience design. CHI '11 Extended Abstracts. https://doi.org/10.1145/1979742.1979640
Höök, K., et al. (2018). The somatic turn in human-computer interaction. interactions, 25(5). https://interactions.acm.org/archive/view/september-october-2018/the-somatic-turn-in-human-computer-interaction