I’ve really enjoyed reading the “syuzhet” debate, but with all due respect to both Jockers and Swafford, the post which grabbed me the most was Scott Enderle’s–not so much for his specific observations, but simply because he pulled back a bit from the back-and-forth and asked a slightly more fundamental question beyond the technical limitations Swafford pointed out–in this case, is the Fourier curve a good model to apply to “sentiment” at all?
It seemed to me that the more Swafford and others pointed out larger conceptual problems, the more Jockers dug in and defended the fundamental workings of his tool, but he never took a step back and defended the ultimate logic of applying it to the problem in the first place. In other words–arguments over whether nor not the 85% accuracy rate that the Stanford tool could produce was “good enough” were interesting, but absolutely nothing in this entire exchange–with all due apologies to the late Mr. Vonnegut–ever convinced me the whole enterprise has any inherent value. I’m just not convinced that even a more accurate plot curve or a more detailed and nuanced sentiment curve would TELL us that much.
I didn’t say that the would tell us nothing, mind you–there is some value in being able to establish certain universal or common themes or plot structures. I can think of plenty of possible insights one might gain by being able to rather quickly graph the general emotional arc of thousands of novels from a particular genre, time or place; or to verify certain universals in plot construction. Jockers, I’m certain, has since come up with a number of such findings.
However…he seems to have begun his project with the notion that there are “six or seven” basic plots, and while he later acknowledges that there arguments that that estimate might be a little low, it’s low by a matter of degree, not kind. In other words–Yes, according to Syuzhet, there are indeed only a handful of plots in all of modern fiction.
Well…OK then! So what? I don’t mean to be glib, but while Jockers had a ready answer for every one of Swafford’s objections, I think he completely overlooks the implicit objection that underlies her entire argument–that his version of “distant reading” is ultimately ONLY interesting “from a distance”, so to speak. The closer you get, the less informative it seems. A program which has to “smooth out” a model of a real thing to such a degree that others find the exceptions and misreadings more indicative than the overall interpretation might simply be TOO distant to be of much use.
I don’t know if I even believe that, to be honest. In a different mood, I might take Jockers’ side in this argument. Ultimately, I find the entire exchange to be mostly a reminder that one of the LEAST true cliches I’ve ever heard is “necessity is the mother of invention.” That, I believe, is rarely true. Edison thought the phonograph would be a teaching tool. He didn’t think the motion picture had commercial applications. Alexander Graham Bell thought he was merely improving the telegraph. I doubt Berners Lee had internet trolls in mind a few decades ago.
It’s OK to find unexpected uses for new tools; it can be exciting to play with a new technology just to see what it can do. But if we’re going to be Digital historians, then we need to go looking for tools that will help us do useful things, rather than look for things just to put our tools to use.