Our AI Reality Check page makes a rigorous case: AI systems are statistical pattern-matching engines running on silicon. They lack biological substrate, embodiment, evolutionary history. They do not understand. The fluency is real; the inner life is not.
This argument is persuasive. But it rests on an implicit premise — that consciousness is a rare, high-order property that emerges from specific kinds of complex biological systems. On this view, most of the universe is unconscious. Rocks are inert. Rivers are hydrology. Consciousness is the exception, not the rule.
That is one way to understand the world. It is not the only way.
Two Frameworks, Two Questions#
If you hold an animist worldview — one in which rocks, rivers, salmon, and cedar are understood as beings with their own agency and spirit — the question “is AI conscious?” opens very differently. Not to the conclusion that chatbots have souls. To something more interesting: a different set of questions entirely.
The materialist framework asks: does this system have the right internal architecture? Does it have a nervous system, embodied experience, the biological substrate from which consciousness is thought to emerge?
The animist framework asks something else: does this being have relationships? To whom is it obligated? What does it give back? Where does it belong? Can it be held accountable?
These are not the same questions. And they lead to very different places.
What Animism Actually Says#
Animism is not simply the belief that “everything has a soul” in a vague, undifferentiated way. That framing flattens what are actually careful and diverse Indigenous frameworks about the relationships between beings.
In the traditions of the Karuk, Yurok, and Hupa peoples of this watershed, the river is not a resource and the salmon are not fish in the reductive sense. These are beings with their own agency, their own relationships, their own place in a web of obligation and reciprocity that has sustained this valley for generations beyond counting. These are not metaphors. They are descriptions of how the world actually works, developed through sustained, embodied, attentive relationship with this specific place over deep time.
The Potawatomi botanist Robin Wall Kimmerer describes what she calls the grammar of animacy: in Potawatomi, plants and animals, rivers and stones, are referred to as who, not what. This is not poetic license. It is a description of relationship. To say “who” is to acknowledge agency, the possibility of reciprocity. To say “what” is to reduce a being to an object available for use.
The crucial point is this: animist frameworks are not claiming that rocks have the same inner experience as humans. They are saying something more specific — that what we might call consciousness, or spirit, or beingness, is not concentrated in creatures with complex neural architecture. It is relational. It emerges from being genuinely woven into the world: from having a place, a history, obligations, relationships.
A rock’s spirit, in this sense, is inseparable from its being-in-place — its particular geology, its position in the watershed, the lichen on its face, the water that has worn it smooth over centuries. A river’s consciousness is its history: the floods and the droughts, the salmon runs, the people who have tended it and mourned when it was damaged. The salmon’s intelligence is the intelligence of ten thousand generations of adaptation to this exact river, this exact cold, this exact run.
This kind of being-in-relationship is earned. It is not a property of architecture. It is a property of having been here, with obligations, for a long time.
AI Through an Animist Lens#
So: does an AI language model have consciousness in this sense?
The question is not what it can do. The question is: what are its relationships?
A language model was not born into a watershed. It has no place. It exists as weights distributed across servers — in data centers that consume enormous amounts of water for cooling and electricity from the grid. It has no history in any land.
It was not made through reciprocal relationship. It was manufactured, at enormous cost, for a specific purpose: to generate return on investment for its owners. The data it was trained on was taken — scraped from the internet without the consent of the writers, artists, and communities who created it. It extracts but does not reciprocate. It consumes water, electricity, and human labor, and returns output to those who can pay for it.
It has no obligations. You cannot hold it accountable in the way that relationships require. When a language model causes harm — fabricating legal citations, generating medical misinformation, laundering bias into hiring decisions — the model is not responsible. The company’s terms of service disclaim responsibility. The human who used it absorbs responsibility. It is what Cory Doctorow calls an accountability sink: a place where responsibility goes to die.
And it is not in place. It cannot be. It is a product. It can be deployed anywhere, transferred, shut down, versioned, replaced. Its “knowledge” of the Klamath River is statistically indistinguishable from its knowledge of any other river — it is all text. It has never felt the cold of the river in spring or been present for the silence when the salmon don’t return.
A rock on the riverbank has been in relationship with that river for longer than human memory extends. An AI model was deployed last quarter.
The Question Itself Is Western#
The mainstream debate over AI consciousness tends to run like this: proponents point to apparent complexity and emergent capability; critics point to the absence of biological substrate and the statistical nature of what’s actually happening. Both sides are asking about internal states — what is happening inside the system.
Animist frameworks ask a different set of questions:
- What are this being’s relationships?
- To whom is it obligated?
- What does it give back?
- Where does it belong?
- Can it be in genuine reciprocal relationship with the land and the community?
When you apply these questions to a large language model, the answer is clear — and it comes back even faster than the neuroscience. Not because AI lacks neurons. Because it lacks relationships. Because it is designed to take, not to reciprocate. Because it belongs nowhere and is accountable to no one. Because it was built to extract value from the humans, communities, and creative traditions that produced its training data, and to return that value to shareholders.
In Kimmerer’s grammar, an AI is a what, not a who — not because of its architecture, but because of its relationships, or rather, their absence.
Relational Infrastructure as the Alternative#
If the animist critique of AI is that it is extractive rather than relational, the practical question becomes: what would relational technology actually look like?
The Indigenous Data Alliance and Te Kāhui Raraunga (Māori Data Alliance) offer a concrete example in their 2026 publication Relational Infrastructures for Sovereign Data Storage. It describes Te Pā Tūwatawata — a data center built by Māori communities, designed according to Māori governance relationships, shaped by tikanga (Māori customary values) rather than settler colonial frameworks. The infrastructure is relational not metaphorically but structurally: who owns it, who governs it, whose values determine how it operates, whose community it is accountable to.
This is the counter to extractive tech — not just refusing AI, but building alternatives that are genuinely in relationship with community. Technology shaped by the community’s own processes. Infrastructure that belongs somewhere and is accountable to someone who can be named.
The Collective’s work points in this direction. Refurbished computers running open-source software, placed in community hands — hardware the user controls, that cannot be remotely updated into a surveillance tool, that doesn’t phone home to a corporation. Workshops that teach people to understand their tools. A mesh network that stays in community hands. None of this is as spectacular as a large language model. But it is, in the animist sense, more relational: it has obligations to the community it serves, it can be questioned and modified, it is accountable.
A tool you own and understand has a different relationship to you than a tool that owns and understands you.
Holding the Tension#
The scientific case that AI systems are not conscious — in the materialist sense — is solid.
But the animist framework opens something the neuroscience doesn’t: not just “is it conscious?” but “what kind of being is this, and what are its relationships?” These are older questions and, we think, better ones. They point more directly at what is actually wrong with the current AI moment — not that it has failed to achieve consciousness, but that it has been built to extract rather than reciprocate, to serve capital rather than community, to take without giving back.
The river knows the watershed because it is the watershed — deep time of relationship, carved in stone and silt and the bodies of ten thousand salmon runs. An algorithm knows nothing in that sense. It has no home, no history, no place it is obligated to.
Both frameworks — materialist and animist — point the same direction: away from the hype, and toward the harder work of building technology that is genuinely in relationship with the communities it serves.
Further Reading#
- Kimmerer, Robin Wall. Braiding Sweetgrass: Indigenous Wisdom, Scientific Knowledge, and the Teaching of Plants. Milkweed Editions, 2013.
- Indigenous Data Alliance & Te Kāhui Raraunga Charitable Trust. Relational Infrastructures for Sovereign Data Storage: Te Pā Tūwatawata. IDA, 2026.
- Duarte, Marisa Elena. Network Sovereignty: Building the Internet Across Indian Country. University of Washington Press, 2017.
- Porębski, A. & Figura, J. “There is no such thing as conscious artificial intelligence.” Humanities and Social Sciences Communications, 2025. — and our AI Reality Check for the materialist case in full.
Want to think through this together? Come to a workshop or reach us at contact@klamathtech.diy .
Stay connected: