Care, in Noë's analysis, is not a sentiment or attitude but a mode of embodied orientation. To care about something is to be physically directed toward it, vulnerable to its outcomes, and temporally committed to its well-being through sustained habitual engagement. These three features — orientation, vulnerability, temporal extension — are constitutively bodily. A mother watching her child on a playground is not merely attending cognitively; her whole organism is committed, her body configured toward the child's position, her vulnerability to the child's suffering registered in visceral distress. AI systems that process information about care lack the embodied conditions that make care what it is.
The analysis of care has a long philosophical genealogy. Heidegger's Being and Time (1927) placed Sorge (care) at the center of his analysis of human existence, arguing that care is the fundamental structure of being-in-the-world — not one emotion among others but the underlying character of Dasein's existence. Contemporary care ethics, developed by Nel Noddings, Virginia Held, and Joan Tronto, has extended this into a systematic moral philosophy that emphasizes the relational, embodied, and particular nature of caring relationships.
Noë's contribution is to specify the embodied conditions of care with phenomenological precision. Orientation: the body is physically directed toward the object of care, with postural configuration, muscular readiness, and proprioceptive awareness of position relative to what is cared for. Vulnerability: the caring organism is exposed to outcomes — the object's suffering affects the carer's body, felt as visceral distress. Temporal extension: care is not momentary but sustained across time through habitual embodied engagement; one's whole life is structured by what one cares about.
For AI, the analysis yields a specific diagnosis. Systems can produce outputs that simulate the verbal effects of care — sophisticated therapeutic responses, apparently empathetic text, well-calibrated emotional support. These outputs may be genuinely useful. But the systems have no orientation toward the world (no body to be directed), no vulnerability (no outcomes that affect them), and no temporal commitment (no life structured by what they care about). The simulation of care is not care, however functionally effective.
The risk this creates is subtle. Users accustomed to AI-mediated interaction may gradually lose the expectation of genuine care — may come to experience well-calibrated responses as equivalent to caring responses, because the phenomenological surface is the same. The cultural expectation of what 'being understood' or 'being cared for' means may thin until it accommodates the machine's capacities rather than maintaining the full weight of what embodied care involves. This is not a problem the technology causes directly; it is a cultural drift the technology enables if the drift is not actively resisted.
The analysis draws on Heidegger's Being and Time (1927), care ethics as developed by Nel Noddings, Virginia Held, and Joan Tronto, and Noë's application of enactive philosophy to ethical concepts in The Entanglement (2023) and related essays.
Care as orientation. The body is physically directed toward what is cared for.
Care as vulnerability. The carer is exposed to the consequences of what happens to the object of care.
Care as temporal commitment. Care is sustained across time through habitual engagement.
The embodiment condition. All three features require a body at stake in a world that matters.
The simulation gap. AI can produce outputs indistinguishable from caring responses without possessing the embodied conditions of care.
Critics argue that a behavioral account of care would be sufficient for moral and practical purposes, and that requiring embodied vulnerability sets an unmeetable standard. Defenders respond that the distinction matters enormously when the question is whether to entrust care relationships — therapeutic, pedagogical, familial — to systems that perform care without possessing it.