Much of the current conversation around AI focuses on bias, ethics, and governance — but rarely does it dig into the history of categories themselves and their lasting impact.
In this paper in the Cambridge Forum on AI: Culture and Society, Juan Cortés and I argue that Linnaean classification — rigid hierarchies and fixed labels — has quietly become the template for how we structure data, metadata, and annotation pipelines, systematically sidelining Indigenous knowledge systems that have understood nature as a living web of relationships for millennia. The categories shaping how we train models were designed in the 18th century to sort and control the natural world.
Drawing on Buffon’s dynamic vision of nature, Kakataibo taxonomies from the Peruvian Amazon, K’iche’ Maya maize cosmology, and other ethnobiological systems from Latin America, we argue why a new classification built on multiplicity, relationality, and cultural context, rather than fixed schemas, what we call rhizomatic hylomorphism, generates living maps where meaning emerges from relationships rather than predetermined labels.
Read the full piece (open access)
Cortes, Juan, and José-Carlos Mariátegui. 2026. «Limitations of the Linnaean categorization model in the age of AI.» Cambridge Forum on AI: Culture and Society 2 (AI & Archives): e5. https://doi.org/10.1017/cfc.2025.10010.
Cambridge Forum on AI: Culture and Society, Themed Issue: AI & Archives
Guest edited by Katie Mackinnon, Louis Ravn, Nanna Thylstrup Joo Eun Seo and Caroline Bassett.