Knowing from the Inside

In the AI Story

The concept organizes Ingold's critique of what he calls the 'anthropology of the armchair': the tradition of trying to understand human practices by reading about them rather than by engaging with them. The armchair anthropologist can produce extensive propositional knowledge about a practice — its history, its structure, its variations — without ever developing the inside knowledge that practitioners carry. She can describe without knowing.

The distinction has specific bite for AI discourse. Large language models have access to vast amounts of propositional knowledge about every human practice. They can describe carpentry, surgery, software engineering, farming, in exquisite detail. What they lack — what they cannot have, on Ingold's view — is inside knowledge, because inside knowledge requires inhabitation, and the models do not inhabit anything. They process representations of practices; they do not live inside practices. The output may be indistinguishable from what an expert would produce, at the level of propositional content. But the knowledge that produced the output is of a fundamentally different kind than the knowledge the expert carries.

This matters practically when the situation at hand exceeds the boundaries of what the model has been trained on. Propositional knowledge is connectable — it can be combined, rearranged, applied across contexts — but only up to a point. Inside knowledge is what allows the expert to recognize that this situation, though superficially similar to familiar ones, has a specific feature that invalidates the standard response. The expert's inside knowledge of the domain is what alerts her to the exception. Without inside knowledge, the exception is not perceived, and the standard response is applied — correctly, by the letter of propositional knowledge, and disastrously, because the situation was not standard.

The implication for the AI-era workforce is that inside knowledge is becoming simultaneously more valuable and harder to develop. More valuable, because AI's propositional competence increases the premium on the judgment that only inside knowledge enables. Harder to develop, because the conditions under which inside knowledge accumulates — sustained, friction-rich, hands-on practice — are precisely what AI reduces or eliminates. This is the structural paradox Ingold's framework identifies: the skill that AI cannot replicate is the skill that AI-assisted workflows make harder to cultivate.

Origin

Ingold developed the concept across his major works from 2000 onward, and in 2013 launched the research project 'Knowing from the Inside' (KFI) at the University of Aberdeen, funded by the European Research Council. The project brought together anthropologists, archaeologists, artists, and architects to explore how knowledge is produced through making rather than through representation.

The concept draws on Gilbert Ryle, Michael Polanyi's tacit knowledge, Merleau-Ponty's phenomenology of perception, and Gibson's ecological psychology. Ingold's distinctive contribution is to insist that inside knowledge is not reducible to tacit know-how but constitutes a distinct epistemological category with its own conditions and its own implications for how inquiry should be conducted.

Key Ideas

A third kind of knowing. Distinct from propositional knowledge and practical know-how, inside knowledge is the embodied familiarity that comes from inhabiting a domain.

Inhabitation produces it. There are no shortcuts; inside knowledge requires the practitioner's own sustained engagement with the domain over time.

AI has propositional mastery but no inside knowledge. Language models process representations of practices without inhabiting them, which limits the reliability of their outputs in non-standard cases.

Judgment depends on inside knowledge. The capacity to recognize when a situation exceeds the standard response requires the cultivated sensitivity that only inhabitation produces.

The paradox of the AI era. Inside knowledge becomes more valuable as AI automates propositional work, but the conditions for developing it are eroded by the same automation.

Debates & Critiques

The concept has been challenged on the grounds that it risks mystifying expertise — making it sound like something that cannot be studied or taught. Ingold's defense is that inside knowledge can be cultivated (the KFI project aims to do exactly this) but cannot be transmitted in the way propositional knowledge can. The pedagogical challenge is to structure environments in which inhabitation can take place, not to devise more efficient methods of information transfer.