What Ontology Actually Is | Casey Reacts

I read Jawahar Talluri's article, "What an Ontology Actually Is," and I think it's a solid introduction to why ontologies matter—but I also think it oversimplifies what an ontology actually is.

In this response, I discuss where I agree with the article, where I think its definition falls short, and why I prefer to think of an ontology in three parts:

1. A language — a vocabulary for talking about a domain.

2. A model of reality — a representation of the things that exist and how they relate.

3. A mathematical structure — a graph that formally connects those things.

I also discuss why I think instance data belongs within the ontology, why I'm not convinced that constraints are a defining component of ontology itself, and why the article's framing seems primarily designed around using ontologies as infrastructure for generative AI and enterprise knowledge graphs.

The article makes a good case for ontologies as infrastructure. But I think there's a much broader story to tell about what ontologies are, what they can represent, and why they matter beyond GenAI.