Palantir and ontology

Business
Opinion
KnowledgeGraphs
Palantir didn’t reinvent ontology.

They just quietly redefined what the word means in enterprise software and most people arguing about it haven’t noticed.

In the academic and semantic-web tradition, an ontology is a formal specification of a conceptualization: classes, properties, axioms, reasoning. OWL, RDF, SHACL. A model of what exists. It’s science.

In Palantir Foundry an “Ontology” is something else. It still has objects, properties, and links but it adds two things classical ontologies deliberately leave out; kinetics and operational coupling.

The kinetical part is about actions and functions. Not just “a work order has a status,” but “here is the function that closes it, the permissions to do so, and the downstream system it writes back to…” Verbs, not just nouns.

The operational coupling is related to how the model is wired directly into the apps people use to make decisions. Editing the ontology changes what the business can actually do tomorrow morning.

So when Palantir says “ontology,” they mean: a semantic layer + a permissioned action layer + a live binding to operational systems. A digital twin you can transact against, not a schema you can reason over.

Whether that deserves the word “ontology”? What’s in a name, right? In this era concepts (alternative facts…) are rewritten all the time. What’s harder to dispute: this is the shape (actionable abstraction) most enterprises actually need. A static OWL doesn’t close a work order or reroute a flight. A model that owns the verbs does.

The interesting question from the point of knowledge graphs is whether the same pattern (objects + links + actions + governance, all in one operational layer) can be assembled from open standards?

Ontology Platform