Métis

In the AI Story

Scott adopted métis from classical Greek because the term named something his research kept encountering: the gap between what institutions knew about the systems they governed and what the inhabitants of those systems knew. The gap was not a failure of data collection. It was structural. The knowledge institutions needed most — the fine-grained, contextual, adaptive knowledge that determines whether policies work or fail — was precisely the knowledge institutional structures could not accommodate. Too local to aggregate. Too embodied to document. Too dialogical to capture in a policy framework.

The concept connects Scott's framework to older traditions in philosophy of knowledge. Michael Polanyi's tacit knowledge — 'we know more than we can tell' — describes a similar phenomenon from a different angle. Hubert Dreyfus's work on embodied expertise, drawing on Heidegger and Merleau-Ponty, describes the same terrain through phenomenology. Harry Collins's taxonomy of tacit knowledge provides the most systematic contemporary treatment. Scott's distinctive contribution was to embed the concept in political analysis — to argue that the exclusion of métis from institutional governance is not merely an epistemological oversight but a structural feature of how centralized institutions engage with the systems they administer.

The application to AI is particularly consequential because AI systems occupy an unusual epistemic position in relation to métis. Language models trained on billions of words have, in some sense, absorbed the traces of millions of practitioners' métis — patterns of reasoning, domain-specific vocabularies, implicit models of how systems behave. The model cannot feel a codebase the way a senior engineer feels it. But it can produce outputs that resemble what the senior engineer would produce, because it has been trained on the written residue that practitioners with métis have left in their wake. Henry Farrell, writing after Scott's death, called this an uncanny middle ground: 'not techne, even if they are not métis either.'

The laparoscopic surgery case from Harry Collins's work illustrates the stakes. When abdominal surgery transitioned from open technique to camera-mediated practice between 1987 and 1997, surgeons trained exclusively in the new technique did not develop the tactile intuition that open surgeons possessed. The métis transferred only imperfectly across the technological transition. The new practice produced better outcomes on many dimensions — shorter recovery times, lower infection rates — but it also produced a new generation of surgeons whose métis was different, and whose different métis had implications that only became visible as the cohort aged into senior positions.

Origin

The use of métis as an analytical category was pioneered by classicists Marcel Detienne and Jean-Pierre Vernant in their 1974 book Les ruses de l'intelligence: la mètis des Grecs. Their work examined how Greek thought distinguished the cunning, adaptive intelligence of Odysseus and Hephaestus from the formal, systematic knowledge of philosophy and mathematics. Scott encountered the concept through his fieldwork and found it a more precise name for what he was documenting than the alternatives available in political science or anthropology.

Key Ideas

Local, embodied, dialogical. The three features that distinguish métis from formal knowledge — features that make it valuable and, simultaneously, make it invisible to centralized administration.

Not primitive but adapted. Métis is not what comes before science but what comes alongside it — a form of knowing adapted to conditions that formal knowledge has abstracted away.

The extraction problem. Attempts to codify métis into explicit rules produce something that looks like the knowledge from the outside but lacks its essential quality: the ability to respond to the unexpected.

Institutional blindness. Institutions do not overlook métis because they are careless. They cannot see it because the institutional form requires legible inputs, and métis resists legibility by its nature.

Debates & Critiques

The concept has been criticized for romanticizing folk knowledge and underestimating the genuine superiority of scientific methods in many domains. Scott's response was that he was not arguing for métis against science but for both against the assumption that either alone is sufficient. Recent AI scholars have debated whether large language models can absorb métis through statistical pattern-matching on text. Collins's position — that they produce mimeomorphic reproductions that fail at polimorphic boundaries — represents one answer; the opposing view argues that the distinction is one of degree rather than kind.