Robin Sloan

Investigating narrative with code

Wow. This is so awesome.

Matt Katz had a hypoth­e­sis…

Maybe microblog­ging pushed [Robin] to write bet­ter, snap­pier sentences?

…so he made a plain-​​text ver­sion of Last Beau­ti­ful, parsed it with Python, and posted the result as a Google Spread­sheet. All in like 20 minutes.

The result? Hypoth­e­sis con­firmed, because fully 75% of my sen­tences are under 140 characters.

This is my new favorite thing.

Here’s the orig­i­nal. I’m work­ing on v1.1 right now.

The pale blue fuzz of readership

Sooo here’s what read­ing looks like:

That’s a graph of read­ers’ paths through The Truth About the East Wind. The x-​​axis is elapsed read­ing time, in min­utes. The y-​​axis is progress through the story; the higher you get on the graph, the closer you are to the end of the page.

So if you’re some­one who scrolled through the story… you’re in there! One of those ghostly blue ten­drils is you. The page is rigged up with a very sim­ple (and totally anony­mous) scroll-​​tracker that dis­patches data points to Sim­pleDB at reg­u­lar inter­vals. It’s a book that phones home.

If you’ve ever talked to me about the Kin­dle, you know this is some­thing I’m totally obsessed with; call it instru­mented read­ing. This post at Snark­mar­ket sketches it out in a sci-​​fi way (and, P.S., has one of my favorite titles of any Snark­mar­ket post ever). So, after talk­ing about it for a looong time, I decided to actu­ally col­lect the data. And you know what?

I have no idea what to do with it.

The aggre­gate behav­ior isn’t very sur­pris­ing. “Yup. Peo­ple scroll down the page.” If any­thing, the sur­prise is sim­ply that a lot of peo­ple spent 10 or more min­utes with this story—which is pretty awesome.

It’s the indi­vid­ual graphs that are interesting:

I feel like that graph tells a lit­tle story. What hap­pened around min­utes 10 and 12? Did this reader go back to savor an image—or to double-​​check a con­fus­ing name?

Seri­ously, these graphs are almost like lit­tle nar­ra­tives themselves:

And this one? No idea:

So sure, these are kinda fun to look at, but they don’t really deliver any­thing action­able. And I don’t think the aggre­gate graph up above does, either. I mean, is there any­thing I can change about the story, or about its pre­sen­ta­tion, based on what I see there? Not really. Not yet.

But this is just a first step. Like the story itself, it’s a pro­to­type—a proof-​​of-​​concept. I’ve got my hands on a cool tool here… and I think I’m prob­a­bly mea­sur­ing the wrong thing.

So what should I mea­sure instead?

Nerd notes: The data gets piped to Sim­pleDB via a lit­tle wrap­per built with Sina­tra, a Ruby frame­work that is the best thing I’ve yet dis­cov­ered for mak­ing super-​​simple tools like this. It’s just fan­tas­tic. The graphs were plot­ted with gchartrb and the Google Chart API. Does every­body already know about this? It’s like magic. What a wacky, won­der­ful ser­vice from Google.