Monday, October 7, 2019

On the direction of literary history: How should we interpret that 3300 node graph in Macroanalysis? [#DH]

That's the name of a draft I've put on for comment. Here's the link:

Abstract, table of contents, and introduction below.

* * * * *

Abstract: In Macroanalysis (2013) Matthew Jockers created a graph depicting similarity relationships between 3300 19th century Anglophone novels, each characterized by 600 features. The graph is derived from a database that contains no date information. When projected onto two-dimensions and visualized, however, the graph has a gradient that is aligned with time. I 1) interpret the graph as a trace of the activity of complex dynamical system (19th century Anglophone novels), 2) conclude that the system has an inherent temporal direction, and 3) contrast it with systems that evolve through random or through cyclic trajectories. I suggest that as this system evolves the range of design possibilities for novels becomes larger, allowing them to encompass a greater range of human experience. I conclude by asserting that evolving literary culture is itself a force in history.


What’s Up? 2
3300 Anglophone novels in a graph 2
Da’s critique 6
A trace of a complex dynamical system 9
Trajectories in time: random, cyclical, and linear 10
Change and growth 14
Literary culture is a force in history 17

What’s Up?

I’m currently working on a response to two articles:
Franco Moretti and Oleg Sobchuk, Hidden in Plain Sight: Data Visualization in the Humanities, New Left Review 118, July August 2019, 86-119.

Nan Z. Da, The Computational Case against Computational Literary Studies, Critical Inquiry 45, Spring 2019, 601-639.
The first part of my article will be address an issue raised by Moretti and Sobchuk at the end of their article, namely the general relationship between literary form and history (pp. 108 ff.) in terms taken from evolutionary biology, where tree-like phylogenies are contrasted with reticular phylogenies. I’ll be address that in this piece, which is draft material for the second part of my article.

I plant to center my argument on the 3300 node graph that Matthew Jockers published in Chapter 9 of Macroanalysis (2013, p. 165). Da dismissed it in one rather long paragraph (pp. 610-611) while Moretti and Sobchuk didn’t mention it all, though their article centered on visualization and that graph is one of the most striking visualizations in contemporary computational criticism. Since it clearly demonstrates a reticular pattern of relationships among 19th century Anglophone novelists, it IS germane to their interest in reticular phylogeny, but for some reason they didn’t mention it. Whatever.

It’s not the reticular form of that graph that interests me here. What interests me here is that, while there is no temporal information (e.g. publication dates) in the underlying database, the graph displays an unmistakable temporal gradient, a fact that surprised Jockers, presumably because he wasn’t looking for that at all. He was simply examining similarity between texts. That the graph has a temporal gradient suggests to me that the underlying social process – the creation of 3300 Anglophone novels in the 19th century – can be thought of as a complex dynamical system having an temporal inherent direction that is of a different kind from mere succession. Da seems unaware of such a possibility and dismisses Jockers’ unintended result as trivial. I know from private communication that Moretti dismisses it as well.

The purpose of this draft essay is to indicate why I think they are wrong.

No comments:

Post a Comment