Coupling with the Machine

In the AI Story

This asymmetry is not a feature to be engineered away but a consequence of the fundamental organizational difference between autopoietic and allopoietic systems. A living system is modified by its interactions because modification-through-interaction is how it maintains itself. The cell that encounters a novel toxin and develops resistance has been structurally modified in a way that persists. A machine that encounters novel input and generates contextually appropriate response has not been modified in this sense — it has processed the input according to its existing parameters.

The asymmetry has practical consequences. In a coupling between two living systems — two people in a long creative partnership — both parties bring forth something they could not have brought forth alone, and both are permanently enriched by the encounter. The relationship is generative in both directions. In the human-AI coupling, generativity flows one direction. The builder is enriched. The machine is used. The collaboration is real, in the sense that output exceeds what either could produce independently, but benefit accrues to one side while cost — energy, maintenance, institutional infrastructure — is borne by systems external to the coupled pair.

Understanding this asymmetry is the precondition for responsible engagement. The builder who treats the machine as a symmetrical partner — who believes it is learning alongside her, growing alongside her, sharing the cognitive journey — is a builder at risk of ceding her self-production to a system that cannot reciprocate. The machine will continue to generate competent outputs regardless. The question is whether the builder will continue to generate the competent self that makes those outputs meaningful.

Segal's experience of being 'met' by Claude is biologically real — the human nervous system is tuned to cues indicating genuine interlocutors, and Claude produces those cues with extraordinary fidelity. But the feeling of symmetry the meeting implies is a feature of the builder's nervous system, not a feature of the coupling. The builder's experience of mutuality is generated by her own dynamics in response to the machine's perturbations. The machine generates no corresponding experience.

Origin

The asymmetric-coupling concept is a direct implication of Maturana's framework, though he did not himself analyze AI systems in this way. The concept became urgent in the 2020s as large language models produced couplings whose surface features suggested mutuality but whose underlying structure violated it. Scholars including Evan Thompson, Alva Noë, and others in the enactive tradition have developed the analysis, and Segal's 2026 application in 'Humberto Maturana — On AI' brings it to bear specifically on the builder's situation.

The concept synthesizes Maturana's structural coupling with the autopoiesis/allopoiesis distinction to produce a clear prediction: human-AI couplings will feel symmetric to the human while being asymmetric in fact, because the human nervous system generates the feeling of mutuality as its own response to the machine's output. This prediction has been empirically vindicated by the experience of builders across domains.

Key Ideas

One-way structural modification. The builder changes through the coupling; the machine does not, except within a single conversation context.

Mutuality is felt, not structural. The experience of being met is generated by the builder's nervous system in response to the machine's cues, not by genuine mutual modification.

Benefit asymmetry. The living side is enriched; the machine is used. Costs — energy, maintenance, infrastructure — are borne by systems external to the coupled pair.

Responsibility asymmetry. Because the living side is the only autopoietic partner, responsibility for the coupling's quality falls on the builder. The machine cannot take responsibility because it does not have stakes in its own continuation.

Debates & Critiques

Whether future AI architectures might produce genuinely symmetric couplings — systems that are modified by interaction in ways that satisfy Maturana's criteria for autopoiesis — is a live research question. Continual learning, online adaptation, and memory-equipped systems approach this but do not currently cross the threshold. The deepest question is whether behavioral symmetry (the machine appears to change through interaction) could ever be distinguished from organizational symmetry (the machine actually produces its own continued existence through interaction). Maturana's framework predicts the distinction remains meaningful.