Relational Ontology (Indigenous)

In the AI Story

The ontological difference is not abstract philosophy but practical epistemology with immediate consequences for how knowledge is produced, transmitted, and used. A Zapotec farmer's knowledge about soil management is not a collection of facts about soil chemistry, crop requirements, and climate patterns that can be separated and recombined. It is an integrated understanding of the relationships between soil and water and seed and season and community labor and ancestral practice—a relational whole in which each element is understood through its connections to all others. Ask the farmer 'what makes soil fertile?' and the answer is not a list of nutrient levels but a story about how the community has cared for this land across generations, how the forest provides what the field needs, how the timing of planting aligns with rain and ceremony and collective labor.

Western agronomy can measure the farmer's soil and generate recommendations based on chemical analysis. The recommendations may be scientifically valid and practically useless—or worse, harmful—because they ignore the relational context the farmer's knowledge preserves. The soil's fertility is not only its chemical composition but its relationship to the watershed, the forest, the practices of sustainable harvest, the community's capacity to provide labor at key moments. Optimizing any single element in isolation can degrade the system as a whole. The farmer's relational knowledge protects against this degradation by maintaining awareness of the whole. Western agronomy's analytical power comes from its capacity to isolate variables. Its blindness comes from the same source.

AI training processes assume atomistic ontology. Knowledge is extracted from context, broken into units (sentences, paragraphs, documents), processed through statistical patterns, and recombined in response to prompts. The process works spectacularly well for knowledge organized atomistically—scientific papers, technical documentation, factual databases. It works poorly for relational knowledge because the extraction step destroys the structure that constitutes the knowledge's meaning. Including Zapotec agricultural knowledge in an AI training set requires decomposing the relational narratives into data points—soil composition, crop types, planting schedules. The data points are accurate. The knowledge is gone. What the AI learns is the surface information. What it misses is the relational structure—the integrated understanding of land-water-seed-community-ancestor that the narrative preserved.

Srinivasan's framework does not propose that AI must be rejected by indigenous communities but that the current model of inclusion—extracting indigenous knowledge into Western organizational formats—is a form of epistemological violence. Genuine inclusion would require AI architectures capable of processing relational knowledge without decomposing it, or alternatively, recognition that some knowledge should not be processed by AI at all. The knowledge's preservation is more valuable than its amplification, and the institutional support should flow toward preservation—protecting the conditions under which relational knowledge can be transmitted through the practices, relationships, and contexts that constitute it—rather than toward extraction into formats that destroy it.

Origin

Relational ontology has been articulated by indigenous scholars including Vine Deloria Jr., Gregory Cajete, Leanne Betasamosake Simpson, and Shawn Wilson, whose Research Is Ceremony (2008) provided a methodological articulation. The philosophical tradition extends across indigenous cultures globally—from First Nations to Aboriginal Australian to Andean to African cosmologies—with variations in expression but consistency in the core claim that existence is relational. Srinivasan encountered relational ontology through his Zuni fieldwork and developed its implications for technology design through subsequent research. His integration of indigenous philosophy with science and technology studies provided the analytical framework for understanding why AI systems trained on Western knowledge struggle with indigenous knowledge—not because of technical limitations but because of ontological incompatibility.

Key Ideas

Existence through relationship. Entities are not self-contained substances that subsequently enter into relationships—they are constituted through relationships, and their meaning is inseparable from their connections.

Holistic integration vs. disciplinary separation. Knowledge about astronomy-ecology-agriculture-social practice forms a unified whole rather than separable disciplines—the integration is the knowledge's distinctive contribution, not a primitive failure to specialize.

Narrative as knowledge structure. Stories are not illustrations of knowledge but its organizational form—the narrative structure holding together the relational connections in a way that propositional statements cannot replicate.

Knowledge in practice, not extraction. Understanding is embedded in the doing, in the relationships maintained through practice—extracting it into text or data requires converting practice into proposition, relationship into fact, and the conversion is lossy.

AI's atomistic incompatibility. Systems that process knowledge by decomposing it into units cannot faithfully handle relational knowledge—the decomposition is the destruction, making genuine inclusion structurally impossible without redesigning the knowledge architecture itself.