Enacted Knowledge (Ingold)

In the AI Story

Michael Polanyi's concept of tacit knowledge — 'we can know more than we can tell' — provided a foundation, but Ingold goes further. For Polanyi, tacit knowledge is knowledge that happens to be difficult to articulate. For Ingold, enacted knowledge is knowledge whose existence depends on bodily engagement with material. The senior software engineer's ability to 'feel' a codebase is enacted knowledge — built through thousands of hours of debugging, watching systems fail, tracing execution paths. Each encounter deposited a trace, invisible to external observation, that educated her perception. This knowledge informs her architectural decisions in ways she cannot fully explain. It operates below articulation, guiding judgment through felt rightness rather than rule-application.

The implications for AI-mediated work are structural. When the engineer reviews Claude-generated code rather than writing it by hand, she exercises representational knowledge — evaluating specifications, assessing logic, comparing against documented patterns. This is genuine expertise. But she does not exercise enacted knowledge. The encounter with the database's surprising behavior under load, the friction of watching a migration fail and diagnosing why, the trace that would have been deposited by that struggle — these do not occur. Over time, the reservoir of enacted knowledge that her judgment draws upon stops replenishing. She retains the knowledge she built through prior hands-on work, but she does not develop new enacted knowledge. Whether her judgment degrades, plateaus, or transforms into a different but equally valuable form of expertise is an open empirical question.

Ingold's research across cultures demonstrates that enacted knowledge is not a luxury of artisanal work but a universal feature of skilled practice. The Cree hunter's knowledge of tracking caribou cannot be transmitted by instruction — it must be grown through the body's engagement with snow, tracks, weather, and landscape. The Inuit builder's knowledge of constructing igloos depends on hands that have learned to read snow's density and structural properties through years of handling. Representational knowledge — instructions, manuals, specifications — can guide the novice, but it cannot substitute for the enacted knowledge that develops only through sustained material engagement. When AI removes that engagement from professional domains, the question is not whether representational knowledge is valuable (it is), but whether it can sustain expert judgment without the enacted substrate that material engagement provides.

Origin

The concept has roots in phenomenology (Heidegger, Merleau-Ponty), ecological psychology (Gibson), and the sociology of practice (Bourdieu's habitus, Lave's situated cognition). But Ingold synthesizes these traditions into a specifically anthropological framework through decades of fieldwork. His 2000 The Perception of the Environment introduced the distinction between knowing about and knowing through. Making (2013) developed it into a systematic critique of how modernity separates conception from execution. By the 2020s, as AI made the separation frictionless, Ingold's framework became newly urgent — not as philosophy but as diagnosis of what is structurally being eliminated from knowledge work.

Key Ideas

Knowledge is not transferred, it is grown. Enacted knowledge develops through sustained bodily engagement; it cannot be extracted from one practitioner and installed in another through instruction or observation.

Articulation produces representation, not knowledge. When enacted knowledge is described in language, the description is useful but incomplete — the knowledge itself remains in the body's educated engagement with material.

AI processes representations, not enactments. Large language models can manipulate descriptions of making with extraordinary sophistication, but they cannot produce the enacted knowledge that comes from hands on material.

The review is not the encounter. Evaluating AI-generated output exercises representational judgment; it does not produce the enacted knowledge that writing the code, debugging the failure, or handling the material would have deposited.

Debates & Critiques

Critics argue that knowledge worth preserving can always be articulated and transmitted — that truly important engineering principles are teachable, and what cannot be taught is craft superstition. Ingold's evidence from multiple cultures suggests otherwise — that the inarticulate dimension is real, productive, and detectable in the quality of the work. The debates also concern whether AI collaboration might produce a new form of enacted knowledge — a different relationship between body and tool that generates its own traces. Segal's account suggests it might; Ingold's framework predicts it will differ in kind from the knowledge that material resistance produces.