Mary Catherine Bateson

In the [YOU] on AI Field Guide

The cycle begins with a question asked at a dinner table: “What am I for?”—a twelve-year-old’s response to learning that AI can now do most of what she might have planned to do. Bateson’s framework is the most complete answer the cycle offers. The question presupposes that identity resides in capability—that the self is what it can do—and the AI transition is experienced as existential precisely because it attacks the content of expertise while leaving intact the practice of engagement. Bateson’s answer is that the self was never the thing you did. It was always the process of doing it—the ongoing, adaptive, improvisational act of composing a self from whatever materials the world provided. The materials have changed. The process has not.

The engineers in Trivandrum who, within days of working with AI tools, began reaching across disciplinary walls they had spent years treating as permanent boundaries are people whose composed careers responded to a change in chord changes by finding new notes. They are not deploying a different competency from the one that built their previous work; they are deploying the same compositional practice in new materials. The walls between domains turned out to be artifacts of the previous arrangement’s translation costs. When the constraint changed, the composition changed with it—and the people who could compose responded. The ones who froze were the ones who had located their identity in the content of their expertise rather than in the quality of their engagement with it.

Bateson’s framework also illuminates the specific cognitive capacity that the AI partnership most needs and that AI most lacks. A large language model is an engine of focused attention: it finds, at superhuman scale, every pattern in every dataset that matches the query. What it cannot do is notice what it is not looking for—the anomalous finch beak at the edge of Darwin’s awareness, the disturbance that signals something important is happening in a direction focused attention is not pointed. Peripheral vision is the human contribution to the partnership that AI cannot replicate, because it depends on what Bateson called the whole person—the full biological, emotional, experiential history that makes each human observer a unique instrument of perception, calibrated by a specific life to register specific disturbances.

Origin

Mary Catherine Bateson was born in 1939 to Gregory Bateson, the anthropologist and cybernetician, and Margaret Mead, the most famous anthropologist of the twentieth century. She grew up in a household where languages changed, continents changed, and intellectual frameworks changed sometimes within a single dinner conversation, absorbing as a formative condition the lesson that would organize her career: that the compositional practice is what persists, not the specific content of any composition. She was trained as a linguist and folklorist, then became an anthropologist, teaching at Northeastern, Amherst, and George Mason universities and the Radcliffe Institute. Her work moves between the autobiography of learning—With a Daughter’s Eye (1984), her memoir of her parents—and the analytical anthropology of how people compose lives in conditions of discontinuity.

Composing a Life (1989) emerged from intensive conversations with four women whose careers had, by conventional metrics, been interrupted: an administrator, a chef, a scientist, an artist. Bateson found that the interruptions were not failures of planning but the occasions of composition—that the women who thrived through discontinuity were the ones whose identity resided in the quality of their engagement rather than the content of any particular expertise. Peripheral Visions (1994) extended the compositional framework to learning itself, arguing that deutero-learning—the concept her father Gregory had developed to name the learning of how to learn—is not a foundation laid in childhood but a continuous practice maintained through adulthood, and that the most consequential learning happens at the margins of focused attention rather than at its center. Her 2018 conversation with Edge.org touched directly on the AI moment, including her recognition that “the key to learning is the discovery of pattern in the unfamiliar, treating it as a resource rather than a threat.”

Key Ideas

Composing rather than planning. The plan presupposes a stable environment and fails when the environment changes. The composition presupposes nothing about stability; it treats the world as a jazz musician treats the chord changes—as the conditions within which improvisation happens rather than the obstacles that prevent the plan from executing. The musicians who responded to the AI transition were the ones who composed: not the ones who abandoned their expertise, but the ones who held their expertise as material rather than identity and found new notes when the chord changes shifted. The ones who froze had built their identity on the plan; when the plan was invalidated, the identity was threatened.

Peripheral vision. Darwin almost missed the finches. The discovery that launched evolutionary biology was registered not at the center of focused attention—which was directed at geology—but at the periphery, where the birds’ anomalous beak variation was noticed without being looked for. Bateson developed this concept to name the cognitive mode that focused attention cannot replicate: the embodied, biographical, idiosyncratic sensitivity to patterns that do not fit the existing framework. AI provides focused attention at superhuman scale. It does not and cannot provide peripheral vision, because peripheral vision is calibrated by a specific life lived in specific environments, accumulating specific experiences that shape what a particular organism registers at the edge of awareness. This is the irreplaceable human contribution to the AI partnership.

The improvisational self. The linear career model locates identity in expertise—the software engineer is her mastery of Python, the lawyer is her knowledge of contract law. When the expertise is threatened, the identity is threatened. Bateson’s framework relocates identity to the quality of engagement—the way one attends to problems, listens to collaborators, finds connections between disparate domains, maintains coherence through disruption. This quality of engagement is not threatened by AI; it is what makes the AI partnership valuable in the first place. The self is not a noun; it is a verb. It is the ongoing act of composition, always responsive to materials that no plan could have predicted.

Continuity through discontinuity. The woman who was a laboratory biologist, interrupted her career for five years, and returned as a public health administrator had not lost five years. She had gained cross-domain intelligence: organizational skills refined by managing a household with young children, a tolerance for interruption and incomplete information, a capacity for simultaneous attention to multiple demands. The interruption forced a transfer of learning that produced meta-understanding—a grasp of the principles that connected domains. The engineer whose pre-AI and post-AI careers seem disconnected is expressing the same compositional practice in different materials. The thread is not the content; it is the quality of attention.

Deutero-learning and learning as a way of life. Gregory Bateson distinguished proto-learning—acquiring specific solutions to specific problems—from deutero-learning—learning how to learn, acquiring the habits of attention and inquiry that shape how all subsequent problems are approached. Mary Catherine extended this distinction to argue that deutero-learning is not a foundation laid in childhood but a continuous practice maintained through adulthood, modified by each discontinuity. The appropriate educational response to the AI transition is not curricular—adding AI to the syllabus—but processual: redesigning learning environments to cultivate the habits of attention and inquiry that produce people capable of composing, rather than people who can execute a plan until the plan is invalidated.