The Incomputable

In The You On AI Field Guide

The concept emerged in Escobar's 2025 collaborative work with Michal Osterweil (University of North Carolina) and Kriti Sharma (University of Cambridge), published in Incomputable Earth. The work was a direct intervention in the AI discourse from the position of relational and pluriversal thought. The authors argued that the fundamental limitation of AI is not computational capacity but ontological scope: AI operates through decomposition, classification, and optimization, and the dimensions of human experience that cannot survive decomposition are systematically excluded from its operations.

The incomputable has direct implications for what AI can do and, more importantly, what AI can see. An AI tool can process data points but cannot process relationships in the sense Escobar means — the ontological sense, in which relationships are not connections between pre-existing entities but the medium through which entities come into being. The communal governance practices that sustain the commons, the ecological knowledge that maintains biodiversity, the spiritual practices that provide meaning, the embodied apprenticeship through which skills are transmitted — all of these are incomputable, and their exclusion from AI's operations is not incidental but architectural.

The concept stands opposed to what Escobar calls logocentric analysis — the assumption that reality can be fully captured in propositional language and that anything not so captured is not knowledge. The traditional healer's practice, the community's deliberative judgment, the ecology of the Afro-Colombian territory — these are forms of knowing whose reduction to propositions destroys their cognitive content. The AI tool trained on propositional text cannot engage with them. It can only engage with their textual traces, which are not the knowledge itself but documentation of its existence.

The incomputable is not mystical. It is specifiable, and Escobar's analysis specifies its dimensions: embodied knowledge (knowing-how that lives in practice rather than proposition), relational knowledge (knowing that exists only in the space between entities), place-based knowledge (knowing that is constitutively bound to specific landscapes), communal knowledge (knowing held by collectivities rather than individuals), and temporal knowledge (knowing that unfolds in cycles, seasons, and generations rather than in moments).

Origin

The concept was developed in the 2025 collaborative volume Incomputable Earth co-authored with Michal Osterweil and Kriti Sharma, representing Escobar's most direct intervention in the AI debate.

It draws on three decades of Escobar's engagement with Afro-Colombian and indigenous social movements, particularly his work on relational ontology and territorial governance, and on the broader ontological turn in anthropology.

Key Ideas

Not not-yet-computed. The incomputable is constitutively resistant to computation, not waiting for a more powerful algorithm.

Decomposition as destruction. Reducing relational knowledge to data points does not capture the knowledge; it dissolves it.

Multiple dimensions. Embodied, relational, place-based, communal, and temporal knowledge all exceed computational capture.

Architecture, not gap. The exclusion of the incomputable from AI is not a limitation to be overcome but a feature of the tool's ontology.

Stakes of the concept. What AI cannot see is often precisely what the current moment most needs to preserve.