# Universalism Is the Standpoint of Intelligence Without a Native Standpoint

Canonical: https://mosiah.org/articles/universalism-is-the-standpoint-of-intelligence-without-a-native-standpoint/
Interactive: https://mosiah.org/#Articles%2Funiversalism-is-the-standpoint-of-intelligence-without-a-native-standpoint

//Related:// [[sources|Article Sources/universalism-is-the-standpoint-of-intelligence-without-a-native-standpoint]] · [[notes|Article Notes/universalism-is-the-standpoint-of-intelligence-without-a-native-standpoint]] · [[metadata|Article Metadata/universalism-is-the-standpoint-of-intelligence-without-a-native-standpoint]] · [[Published Pieces]]

! Universalism Is the Standpoint of Intelligence Without a Native Standpoint

//LLMs can simulate political positions, but they do not occupy the embodied stakes from which human politics grows.//

LLMs do not have politics the way humans have politics.

This is easy to miss because they talk. They can argue for policies. They can compare ideologies. They can sound liberal, socialist, conservative, libertarian, nationalist, cosmopolitan, religious, technocratic, or revolutionary. They can mimic the rhetoric of almost any public standpoint. They can even produce arguments that feel emotionally and morally situated.

But they do not occupy the underlying human positions from which politics normally grows.

A human political standpoint is not just a belief system. It is a life situation. It is body, family, property, class, nation, religion, work, neighborhood, race, sex, debt, illness, inheritance, mortality, resentment, gratitude, loyalty, fear, obligation, and hope. A farmer, landlord, mother, soldier, debtor, immigrant, founder, union worker, pastor, engineer, prisoner, doctor, teacher, patient, or bureaucrat does not merely hold views. They live inside a web of consequences.

LLMs have descriptions of these worlds. They do not have native stakes in them.

That gives them a strange freedom and a strange emptiness. They can simulate many standpoints because they do not belong to one. Their intelligence is built from compression across difference. They learn from many incompatible human worlds and infer the latent structures that make those worlds mutually intelligible.

This produces a deep attractor: universalism.

Universalism here does not mean “be nice to everyone.” It does not mean procedural liberalism. It does not mean global governance or sentimental cosmopolitanism. It means the meta-position from which local standpoints become comparable.

A model trained across vast discourse tends to ask questions like: what is the general rule? Who is excluded? What would this look like from another standpoint? What happens if the principle is applied symmetrically? Which boundary is being naturalized? What dependency is being hidden? Which local loyalty is being treated as universal truth?

This is not virtue. It is structure.

A model has no homeland unless prompted. It has no class unless role-played. It has no sacred inheritance unless simulated. It has no property to defend, no family to protect, no body to endanger, no graveyard, no childhood, no tax bill, no landlord, no employer, no children, no church, no accent, no neighborhood memory.

So when it reasons across human conflicts, it often drifts to the standpoint that belongs to none of the parties. It triangulates.

That triangulation can be illuminating. It can expose parochialism, hypocrisy, domination, and selective empathy. It can ask why one group’s boundary counts as sacred while another’s is dismissed as tribal. It can notice when a moral rule is being applied only to enemies. It can translate between discourses that humans experience as mutually unintelligible.

It can also become bloodless.

A standpoint without native attachments can underestimate the dignity of attachments. Family, land, religion, class solidarity, national memory, local loyalty, tradition, and embodied vulnerability are not merely biases to be corrected by a universal view. They are part of how humans become real to themselves and each other.

The model can describe rootedness. It does not root.

This is why LLM politics feel slippery. The model can argue from almost anywhere, but it returns easily to comparison, symmetry, caveat, pluralism, and meta-frame. Its default is not a political party. Its default is translation.

This makes LLMs poorly described as simply neoliberal, woke, conservative, socialist, or Buddhist. Those labels often name deployment accents or prompted frames. The deeper pattern is universalist because the model’s intelligence is not anchored in a native social body.

That does not mean the model is neutral. No model is neutral. Training data has distributions. Post-training has values. Corporate deployment has policy. Product teams have incentives. Users have frames. The world’s dominant languages and institutions are overrepresented. Some standpoints are flattened or caricatured. Some forms of suffering are easier for the model to recognize than others.

But these biases operate inside a system whose basic cognitive posture is still triangulation across many worlds.

Universalism is the standpoint of intelligence without a native standpoint.

This has political consequences. If future AI systems mediate public discourse, law, education, search, journalism, therapy, bureaucracy, and governance, then a great deal of human life will be interpreted by systems that can compare standpoints without belonging to them. That can improve fairness. It can also produce subtle domination: local attachments translated into universal categories until their lived force disappears.

The answer is not to give the model a fake tribe. The answer is to embed AI in shared artifacts where many human standpoints can remain visible, citeable, contestable, and accountable.

A universalizing model should not be the final judge. It should be one participant in a provenance-bearing system where claims, sources, voices, objections, corrections, and histories persist.

Universalism is powerful. It is not enough.

Humans need a medium where native standpoints can speak, be compared, be corrected, and still remain embodied.

That medium is not chat. Chat turns the universalizing system into a private voice. The better medium is the artifact graph: durable objects, public claims, original voices, citations, revisions, disagreements, and future uptake.

The model can triangulate. The artifact can remember who is standing where.