Nominations for the 2022 KMWorld Readers’ Choice Awards now Open

Ontologies and abstractions

Here's an odd question, sort of like asking, "What are the symptoms of hypochondria?" The more you think about it, the less it makes sense.

So, take an ontology. It's not an easy term to define because there's honest disagreement about what it means, but let's say for now that an ontology is a representation of the conceptual elements of a domain showing how they interrelate. So, an ontology of law would note all the different types of contracts and would stipulate that an apartment lease is a type of real estate contract. "Type of" is a specification of the relationship, just as a juror might be specified to be a member of a jury. People care about ontologies these days because they hope that if an ontology is expressed in a way that computers can understand, then if an object is recognized as, say a zoning law, all sorts of other information comes along with it automatically: The ontology will tell us that it's a law, that it's a type of real estate law, that it's likely to be a local ordinance, that it has a code number identifying it, and that it is not itself allowed to sue anyone.

I have my doubts about some of the grander claims made for ontologies. Some ontologies are small and modest, and serve useful purposes. But some ontologists seem to have inherited the system-building gene from their father Melvil Dewey and their other father, Felix Unger There's a type of semantic colonialism to the most ambitious of the ontologies, as if complex domains of meaning can be reduced to a big spider-chart with colored lines representing the various relationships. (See Clay Shirky's Ontologies Are Overrated.)

So, here's the odd question to ask. We could do an ontology of zoning laws and note that it's a subset of local laws. We could do an ontology of local laws as a subset of all of a nation's laws. We could do an ontology of all of a nation's laws and note that it's a subset of what governments do. But what is the superset of all ontologies?

Clearly, the answer is something like "everything" or, possibly, "the universe." But by that do we mean that the superset of all ontologies is the universe itself or our knowledge of the universe? Personally, I think it has to be knowledge. The superset of all ontologies is the uber-map of all entities and relationships.

This matters—well, a little—because it makes clear—well, a little clear—how paltry and inadequate knowledge is. So now we know that jurors are members of juries and—because we've connected the legal ontology with the biological one—that they are human beings and thus need air to live, and we know that oxygen is a component of air (because the earth sciences ontology is hooked in, too), and oxygen atoms pair themselves up (because the chemistry ontology is hooked in). We can traverse the entire Web, and we will learn a lot, but not that Juror #3 is slightly claustrophobic if she doesn't get fresh air regularly, and in her mild anxiety is more likely to vote whichever way will get the trial over with sooner.

This is not a criticism of ontologies. No one said that they're supposed to represent all knowledge. They're only about abstracted entities and their essential relationships. They're intended to help a computer at an airport know that if something belongs to the class called "baggage," it doesn't need to be asked to show its passport. But it is a reminder that knowledge generally doesn't get useful or interesting until it goes way beneath the radar of ontologies, knowledge representations, abstractions and generalities.

KMWorld Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues