What an Ontology Explorer Taught Me
I set out to build a tool that explains ontologies. I ended up building one that's careful about how it explains them, and that was the harder, better project.
The Method Was the Point
I set out to build a tool that explains ontologies. What I actually built was a tool that’s careful about how it explains them, and that turned out to be the more interesting project. Ethnographers have a word for this: reflexivity, the discipline of disclosing your own method so a reader can judge the account. It became the spine of the whole thing.
A summary you can’t interrogate is just a rumor with good formatting.
Field Notes That Show Their Work
Every plain-language summary in Emic is generated by a model, and every one says so. A provenance line under each note records which model wrote it, which generation of the prompt produced it, and when. That isn’t legal boilerplate. It’s the fieldworker telling you where the account came from, so you can weight it accordingly. A note written by one model in February is a different kind of evidence than one rewritten last week, and you can see which you’re looking at.
The summaries also carry a standing disclaimer: they’re aids for understanding, not authoritative definitions. The authoritative text is always the verbatim spec, kept one click away. The sources and licenses behind every ontology are credited on a dedicated page.
Reflexivity isn’t a disclaimer you bury. It’s metadata you attach to every claim.
Declared vs. Inherited
The same instinct shows up in how Emic handles inheritance. A concept in CIM rarely stands alone; it inherits most of what it is from ancestors. Storage Volume declares 5 properties of its own and inherits 52 more from 6 ancestors above it. Emic shows that split explicitly, attributing each property to the class that actually declares it, instead of presenting one flat list of 57 that pretends the concept invented them all.
What a concept declares and what it inherits are different facts. Flattening them is a small lie that adds up.
A First Week in the Field
The last piece is tours: short, narrated walks through a handful of related concepts, in sequence. “From raw capacity to a mounted disk” in CIM; “How D3FEND frames a defense” in the cyber-defense ontology. They exist because dropping someone into a 1,648-node graph is not an introduction. A tour is the local who walks you around in your first week, before you know the map.
What’s Still Open
Plenty. The live-generation tier exists, but the public site doesn’t use it, so every note is only as current as its last batch. Six ontologies are mapped, which is a start, not a survey. And the hardest gap is the one between a concept’s formal name and the word a practitioner would actually say, the jargon that never makes it into the schema. Going emic gets you the insider’s categories. It doesn’t yet get you the insider’s slang.
The ontology gives you what a domain wrote down. The last mile is everything the domain knows but never bothered to formalize.
The Takeaway
The Takeaway: Building an explorer for ontologies taught me that the explaining is the easy part and the honesty about the explaining is the hard, valuable one: show the model, show the date, separate declared from inherited, keep the source one click away. Going emic means reading a domain in its own categories. Doing it responsibly means always showing your work.