The paradigmatic illustration is the surgeon who pauses mid-procedure because something in her hands says that this tissue is not what it should be. She has no words for what she has detected. If pressed, she might say 'it didn't feel right' — four words that carry, compressed, thousands of hours of practice. Her body knows something her mind has not yet formulated, and the body's knowledge, in this moment, is more reliable than any imaging system.
The neural substrate involves the integration of cognitive processing with somatic markers — the visceral signals Damasio identified as the body's contribution to decision-making. Cortical, limbic, and interoceptive systems operate as a unit, producing a state that is at once a thought and a feeling, a judgment and a sensation.
This has consequences for what AI can and cannot amplify. The senior engineer from Trivandrum whom Segal describes — whose twenty percent of remaining work turned out to be the part that mattered — possessed embodied knowledge of this kind. His architectural instinct was not a set of rules he could articulate. It was a set of responses he could feel, built through years of practice, consolidated through default mode processing, integrated into an understanding that operated below explicit consciousness but above the threshold of reliability.
AI does not have a body — and this is not an incidental observation. A 2025 paper in AI & Society argued that large language models lack being-in-the-world, which makes it impossible for them to represent the world in a practically sensible way. The representations are accurate; they lack the felt dimension that gives human understanding its functional depth.
The concept emerged from Immordino-Yang's synthesis of Damasio's somatic marker hypothesis with research on expertise and embodied cognition. Its operational force is diagnostic: it names what AI systems structurally cannot have and what frictionless interaction atrophies in humans.
Feeling and thinking are not sequential. In emotional thought they are simultaneous aspects of a single neural event.
The body is part of the cognition. Visceral signals, somatic markers, and interoceptive processing are constitutive, not adjunctive.
Embodied knowledge is deposited, not transferred. Each hour of emotionally engaged practice deposits a layer; the layers accumulate into reliable judgment.
AI can amplify but not produce it. The tool extends reach; it does not replace the somatic history that makes understanding felt.
The feeling of knowing is the signal worth amplifying. It distinguishes competent output from work that carries the quality of having been earned.