A new category
Somatic AI
What is Somatic AI?
Somatic AI is a category of artificial intelligence whose primary input is the physiological and behavioural state of the human body — breathing, pulse-related signals, posture, expression and interaction pace — and whose primary output is an interface that adapts to that state. Since 2014, Hannes Bend has developed bio-adaptive systems through Metaverses and mYndful, combining physiological feedback, virtual reality and responsive environments. The approach was formalized in the 2016 AAAI paper 'Mindful Technologies' and extended through granted adaptive-interface patents (US 10,423,893 and US 11,561,806). Higher-level state estimates are experimental and probabilistic, not diagnoses.
Somatic AI is artificial intelligence that senses and responds to embodied human state. Instead of optimizing for attention, engagement, or prediction, Somatic AI works from measured signals — pulse-rate estimates, breathing patterns, movement and posture, interaction pace and explicit user feedback — and adapts interfaces, environments and feedback to support regulation, clarity, wellbeing and performance. What a camera or sensor records is a signal; anything named beyond it — arousal, strain, focus — is an inference, and is treated as probabilistic, never as a reading of emotion.
Definition
Somatic AI — a category of AI systems whose primary input is the physiological and behavioral state of the human body, and whose primary output is an adaptive response calibrated to that state.
Origins
Since 2014, Hannes Bend has developed embodied and bio-adaptive systems through Metaverses / mYndful, integrating visual stimuli, meditation, physiological feedback, virtual reality and responsive environments. In 2016, this trajectory was formalized in the AAAI paper “Mindful Technologies: Research and Developments in Science and Art” and later extended through granted adaptive-interface patents and the Somatic AI framework. Higher-level state estimates are experimental and probabilistic, not diagnoses.
Related but distinct
- Embodied AI — robotics and physical agents that act in the world through their own bodies. A separate, long-established field.
- Human embodiment and physiological state — the focus of Hannes Bend's work: sensing the person, not embodying the machine.
- Bio-adaptive interfaces — the responsive interface layer that changes in reaction to those signals.
- Somatic AI — the present framework connecting the sensing layer to the adaptive output.
Principles
Embodiment first
The user is treated as a physiological, sensing being — not a stream of clicks, prompts, or behavioral traces.
State, not history
Interfaces respond to present-moment nervous-system state (breath, heart rate, posture, expression) rather than only past behavior.
Adaptive, not extractive
Systems adjust to support regulation, clarity, focus, and recovery — instead of maximizing attention capture.
Scalable sensing
Supported contactless sensing methods — including camera-based rPPG and microphone-derived features — can run on standard consumer hardware without wearables.
The stack
- ·Somatic AI — AI that senses human state.
- ·Bio-adaptive interfaces — the interface layer that responds.
- ·rPPG and webcam-based physiological sensing — the scalable input.
- ·Deployed globally via Breathing AI, VitalSign AI, Optimizing AI, and Transforming AI.
Why now
Published research supports contactless estimation of selected physiological signals under suitable conditions, while performance remains dependent on the algorithm, camera, lighting, motion, compression, skin presentation and intended application. At the same time, the harms of attention-optimized systems — dysregulation, burnout, chronic stress — have become clear. Somatic AI reframes the goal of intelligent systems: from capturing the user to supporting the user.
Applications
Wellbeing applications
Breath- and heart-rate-guided pacing, self-regulation support, focus and recovery routines — non-diagnostic and non-therapeutic.
Performance
Focus, presence, and readiness feedback for athletes, operators, and knowledge workers.
Research and validated health applications
Contactless physiological-signal research, protocol support and longitudinal observation where the technology has been independently validated for the specific population, conditions and intended use.
Learning & education
Gamified, embodied learning that responds to arousal and engagement in real time.
Immersive media
VR / AR / spatial systems that sense and adapt to the participant's body.
Human-computer interaction
Interfaces that dim, slow, or reshape themselves in response to human state.
Lineage
Hannes Bend's work predates the current wave around embodied AI and nervous-system-aware technology. It began in art, perception, meditation, neuroscience, and VR biofeedback — through collaborations with the Alemán Quantum and Nanoscale Physics Lab and Institute of Neuroscience at the University of Oregon, the Association for the Advancement of Artificial Intelligence (AAAI) Spring Symposium at Stanford, and museum presentations of breath-responsive VR — before becoming patent-backed AI ventures.
Generic AI vs. Somatic AI
| Generic AI | Somatic AI |
|---|---|
| Click-based interaction | Physiological signals (rPPG, breath, posture) |
| Attention-driven engagement | Present-moment state awareness |
| Behavioral history & prediction | Embodied, in-the-moment context |
| Disembodied interaction model | Adaptive interface response to human state |
| Engagement optimization | Wellbeing, performance & research outcomes |
Explore the evidence
The technical definition and closed-loop mechanics are covered on bioadaptive interfaces. Papers, patents and collaborators are collected under research & patents, and the dated chronology of how this developed is on the research timeline. The origins in immersive and biofeedback work are documented in the art archive; the applied systems in ventures, with the applied product layer documented on Breathing AI. For the long-form account, read From Immersive Art to rPPG. For how this term relates to neighbouring fields, see Somatic AI vs. embodied AI and affective computing, and for the moral case for bio-personalized technology.