Philosophy
What if the biggest flaw in AI is not just its carbon footprint, but the worldview beneath it? Indigenous ways of knowing treat land, rivers, plants, and memory as relationships, while AI turns them into data to be extracted, stored, and sold.
Other Ways of Knowing the World
Indigenous onto-epistemologies; the ways of being and knowing that emerge from specific, long-standing relationships with particular landscapes offer something that neither deep ecology nor mainstream AI ethics has yet managed to articulate: a conception of knowledge that is relational, reciprocal, and accountable.
Robin Wall Kimmerer, the Potawatomi botanist and author of Braiding Sweetgrass, speaks of a grammar of animacy, a way of using language that recognises plants, rivers, and animals as beings with agency, not objects for use. To say the river in English is already to reduce it to a thing. To say “ki”, as she proposes from Potawatomi, is to acknowledge it as a living subject with its own interior life and its own claims on our behaviour. This is not merely poetic. It is epistemological. It changes what questions you ask, what relationships you enter, and what obligations you recognise. Aldo Leopold, writing in A Sand County Almanack in 1949, articulated what he called the land ethic, the proposition that the boundaries of the moral community must be extended beyond humans to include soils, waters, plants, and animals. A thing is right, he argued, when it tends to preserve the integrity, stability, and beauty of the biotic community; it is wrong when it tends otherwise. This was a remarkable philosophical move for its time, anticipating deep ecology by decades. But Leopold’s land ethic, read alongside indigenous onto-epistemologies, reveals an important asymmetry: where Leopold proposes an ethics extended from the human outward to include nature, many indigenous traditions begin from a position of fundamental relatedness; one that does not require a philosopher’s decree to extend the moral circle because the circle was never closed to begin with.
The Pachamama worldview of Andean communities, the concept of Buen Vivir, living well in relationship rather than living better through accumulation, offers a political-philosophical alternative to the growth logics that underpin AI’s current development trajectory. Where the AI economy asks how we can extract more value from more data more efficiently, Buen Vivir asks: what kind of relationships with the living world make life worth living? These are not quaint alternatives to be displayed in a museum of cultures. They are living epistemologies that have sustained human communities in relationship with specific ecosystems for millennia, and they carry within them a critique of extractive rationality that no algorithm has yet managed to produce.
AI, by contrast, is built on a grammar of objectification. Everything that enters the system is data that is discrete, extractable, and recombinant. The river becomes a dataset of flow rates and salinity levels. The forest becomes a carbon sequestration calculation. The community elder’s knowledge of medicinal plants becomes a training input.The act of translation into data is not neutral; it is a philosophical act with political consequences. When traditional ecological knowledge is digitised, indexed, and fed into a machine learning system, something is preserved, certain information and something is lost: the relational context in which that knowledge lives, the obligations it carries, the community structures that make it transmissible across generations. The data point floats free of its roots. It becomes available for uses its original holders would never have sanctioned. An ecologically just AI would need to operate from a fundamentally different premise: that knowledge is not property to be extracted but a relationship to be entered carefully, with consent and accountability. That the communities whose knowledge and landscapes are implicated in AI’s environmental promises must be not just consulted but positioned as decision-makers. The criteria for evaluating an AI system’s success must include not just efficiency and accuracy, but justice – who benefits, who bears the cost, and who gets to ask the question in the first place.