Primitive Classification and Categorical Violence
Durkheim and Mauss's 1903 essay 'De quelques formes primitives de classification' demonstrated that classification systems of different societies are not arbitrary impositions on a neutral world but reflections of social organization projected onto nature. The Zuñi divided the world into seven regions corresponding to seven clans. Australian societies organized nature into the same moieties that organized kinship. The categories were not discovered — they were projected from social structure onto the world, and the projection created the appearance of a natural order that was in fact a social construction. This framework becomes devastating when applied to the AI transition, which is imposing a new taxonomy on the landscape of professional capability: a taxonomy that divides skills into the amplifiable (what AI can process) and the unrecognizable (what AI cannot parse). This classification is not neutral. Like all classification systems, it carries an implicit theory of value.
In the AI Story
Every new classification system is experienced by contemporaries not as a system — not as a human construction — but as a description of reality. The medieval Great Chain of Being, the racial classifications of the nineteenth century — each was naturalized, experienced not as imposition but as revelation, and the naturalization was what made the classification's violence invisible.
The classification AI imposes is undergoing this naturalization. The distinction between what AI can amplify and what it cannot recognize is being treated as discovery rather than classification. Skills the tool can process are regarded as real skills; skills it cannot process are treated as residual or illusory. Segal captures the economic expression in The Orange Pill: 'depth itself was losing its market value... because the market was discovering that, for most purposes, breadth was good enough.' This describes the economic dimension accurately, but Mauss's framework reveals a deeper operation: not merely repricing but reclassification.
The violence operates through omission rather than force. No one decides that tacit knowledge is worthless. No policy is enacted. The classification simply does not include a category for it, and in a world where what is categorized is what is real, the absence of a category is the most effective form of erasure. The skill that no metric measures and no performance review evaluates does not cease to be real philosophically — but it ceases to be socially real, and in a social world, the unrecognizable does not survive.
Origin
'De quelques formes primitives de classification' appeared in L'Année Sociologique in 1903, co-authored by Mauss and Durkheim though attribution of specific contributions remains debated. The framework was extended by Mary Douglas, Pierre Bourdieu, and contemporary work on classification and power.
Key Ideas
Classifications are social. Category systems project social organization onto nature, creating the appearance of natural order.
Naturalization conceals construction. Every classification system is experienced by its contemporaries as description rather than imposition.
The amplifiable and the unrecognizable. AI imposes a taxonomy that renders some capabilities visible and others invisible.
Violence through omission. Reclassification operates not through force but through the absence of categories for what is being erased.
Self-reinforcing naturalization. Once the classification is experienced as reality, practitioners whose knowledge falls outside it come to doubt that their knowledge was ever real.
Appears in the Orange Pill Cycle
Further reading
- Émile Durkheim and Marcel Mauss, 'Primitive Classification' (1903, trans. Rodney Needham 1963)
- Mary Douglas, How Institutions Think (1986)
- Geoffrey Bowker and Susan Leigh Star, Sorting Things Out (1999)
- Pierre Bourdieu, Distinction (1979)