The Beauty in Structure

Only a math major could write a headline like that.

Or maybe an Asperger’s-prone anal-retentive math major.

But there is value in structure. If I have a database full of the best information in the world – lists of bribes taken by top government leaders, poisonous ingredients in common foodstuffs, etc – but there’s no list of data fields, if it’s not organized in any way, if I can’t write queries to return specific data, then it’s just numbers and words in a pile. The value is in the structure, and understanding the structure.

It’s the same with stories. There’s huge value in stories, and in fact we do extract it regularly when we read them (assuming they’re well-written, which isn’t always the case. But I digress.) But that value is

1. often limited in time, meaning that if I read the same story a year later, much of that value has dissipated; and

2. is isolated from the rest of the stories on the same topic (short of linking – which is a form of value-enhancement (which Jeff Jarvis is absolutely right about), but a limited form of value-enhancement.

It’s as if I had a sheet with one bribe paid to a government officials, that was written a year ago. That’s useful, sure. But if everyday someone in my office was creating another sheet, with another bribe, to another official, and we didn’t find a way to marry all that into a real-time database that we could poll – well, we’d be pretty stupid.

And so it is with newsrooms, that treat each story like it was a single sheet. We put them in an archive, it’s true. And we even tag them. But then we ask people to read them as if they were single sheets. You can see as many of them as you want, but you have to note the date yourself, the person involved, and go over all the duplications.

It doesn’t make sense.

To be sure, it’s much harder to come up with a data structure for stories. Possibly impossible.

But we can do it for limited areas of coverage/applications, such as Politifact. We can do it broadly for all stories (changing date fields, extracting entities and individuals, writing summaries) and maybe learn how to do it at a more detailed level for some kinds of stories – sports, natural disasters, markets – where there’s a more set form and structure.

The trick is a mental one: Thinking of stories as sets of data, rather than as glittering pearls of prose. They are that, too – but they need to be both.