I leaped out of bed because my head was buzzing and wouldn’t stop, so here goes:⁰¹

By definition the semantic web needs ontologies, right?⁰² and those are defined by shared vocabularies, but shared vocabularies aren’t static. They don’t emerge instantly and once they do, they change constantly… Language, not just vocabulary—syntax too, is one of the great intentional and accidental ways in which whole swaths of people are disenfranchised. A controlled vocabulary can potentially silence and isolate thinkers who relate ideas differently or are not fluent in the language defined by the dominant culture. In the case of objects, the concession to the dominant culture matters less than it does around ideas. Does it matter for raw data? Probably. If only for findability.

In the real world, and in the world I want to inhabit, there’s no such thing as a standard connection between data, documents, people, and ideas. (I mean for bigger, nonspecialized systems. Obviously the shared vocabularies that emerge around individual object and document types within specialized systems are very useful.) But when I consider giant quivering blobs of raw data, I want to be able to grab the blobs I’m looking for, smash them together, and let the computer do its computerly work—look for the mathematical patterns invisible to me because of my looping, sliding, sideway crabwalking mind; and then I will use my messy human mind to determine the meaning in those patterns.

For example, I want to be able to grab historical economic data from the 1930′s on a county in rural Mississippi, present day epigenomic data from people born in the same county after 1970 (Fine, this is happening in the future), data on the dental health and daily personal habits of those people, and data on the population of fish in Mississippi rivers between the 1930s and now. I don’t need the computer to make sense of all that, just to give me mathematical patterns. I think everyone may be focusing on answering questions when we ought really to be supporting the asking itself.

Obviously the tricky part is making the tools that allow people to work with data that way, to make those mad demands, to perform those kind of intuitive, inductive searches… Inductive search. I’m not sure how else to describe that sort of inquiry. (I wonder if it already means something else. Too sleepy to research it.) There are several discrete parts to that scenario: finding the blobs of data, which is where dense tagging makes sense, handing it over to the computer⁰³, studying the patterns, and starting again with a slightly more nuanced question based on the results of the last inquiry.

So that’s an exciting example for raw data, but the same thing simply doesn’t work for ideas or objects (or does it?). Kate Ray, (who made a nifty short documentary giving a very general idea of the Semantic web) says some relationships, such as manufacturing, are certainly static and ought to be defined, and I can’t help thinking the reason the definition of the relationship between those two objects or airplane parts is presently consistent and tidy is because it is not sitting out in the open, growing a crazy fuzz of lesser lines connecting each physical piece to some abstract idea in a particularly personal and idiosyncratic way. (I’m deliberately misreading a bit here, but this would be the time and place.) If for some reason I stumble upon this imaginary object, this recognizable chunk of an airplane, right this very second, I will relate it, in this order to: the long delirious burning blue, the firebombing of Dresden, a Martin Margiela flight suit on a guy I met at a work party some ten years ago, the painful vintage red heels I wore to the same party, ebay, rain, walking tours in England, the rest cure, incarceration, lead levels, and handcuffs, labor pain, qi, inner ankes… I’ll stop, but the point is, it may be possible to ask a machine to make sense of that stuff of mine (maybe it’s even me.) But, that odd, kinked chain will be totally different ten minutes from now and unrecognizable by tomorrow morning. (I just reread that list thirty seconds later and it’s already wildly different.) What does that mean?

In fact, the time axis is the fascinating and puzzling thing to me about graphs of personal data such as Facebooks’s social graph or hunch’s taste graph. Facebook obviously fails to make sense of the time axis, given how poorly it represents the real weight a user places on her personal relationships. I wonder how or whether hunch takes time into account. When I answer their questions I am often acutely aware that my answer was or will be different one day or one year or ten years out in either direction. It’s the same thing that has always made me squirm about Cosmo quizzes and the Meyers Briggs test. I want to make a three dimensional graph of lumpy, crazy, mismatched, ongoing, relentless data that just won’t stay still, any more than the brains and bodies of the humans creating the data do. And then I want to demand things of it.