I’ve been working hard over the last two month, writing a response to two reviews of recent work in computational criticism:
Nan Z. Da, The Computational Case against Computational Literary Studies, Critical Inquiry 45, Spring 2019, 601-639.
Franco Moretti and Oleg Sobchuk, Hidden in Plain Sight: Data Visualization in the Humanities, New Left Review 118, July August 2019, 86-119.
As I’ve working it’s just been growing and growing and has turned out to be more involved than I had guestimated or intended when I started.
Which is not all that unusual.
First I say a bit about what’s been going on with that essay and then to the point, which is about crafting intellectual frameworks in which we can study cultural evolution. THAT’s what’s been taking me so long with this essay. I’d misjudged that maturity of existing frameworks.
How I got here
Back in early October I’d posted draft material for the main argument, which would be in the second part of the article [1]. That centered on that 3300 node graph in Macroanalysis, by Matt Jockers, that’s interested me since I first encountered it in 2014. Each node in the graph represents roughly 600 pieces of information about one 19th century Anglophone novel. This is the graph:
I figured that would be the most difficult part of the piece to write, which is why I drafted it first. I figured the first part of the article would be much easier since the material I wanted to cover wasn’t as sophisticated and difficult.
It’s six weeks later and I’m still working on draft material of the easier part of the article (I’ll be ready to post that in a couple of days, or maybe the end of the week). A week ago I published material for an appendix on cultural evolution [2]. When I started this project I hadn’t even intended that there would be such an appendix because I hadn’t intended to get that deep into cultural evolution.
What happened? Had I misunderestimated the nature of the task? Yes and no.
On the one hand material I’m currently drafting is not as conceptually sophisticated and dense as that I’d devoted to the graph. However, I’ve already written quite a bit about that graph and I had some draft material sitting around on my hard-drive. So putting that all together wasn’t difficult. But, for various reasons, I’d underestimated the amount of time I’d have to spend on tree structures and network structures for biological and cultural phylogenies, a subject in which I’d published two decades ago [3].
I’ve now got a reasonable sense of what I’ve got to do to bring closure to the current draft material and I’ve got concrete plans for next steps. All I’m doing here is taking a breather, clearing the decks for the final push as it were.
Frameworks for studying the evolution of literary culture
In her review Da devoted a long paragraph to that graph, dismissing it as a lot of effort for a trivial result. She’s wrong, of course, but I can see why she would think that. She didn’t have an intellectual framework in which to understand the significance of that graph. Nor did Jockers himself; he knew it was an interesting conceptual object, but couldn’t explain quite why. My project in this essay has been, then, to set up such a framework.
What does that have to do with Moretti and Sobchuk, who didn’t even discuss that graph? They called for more theory and clearly believed evolutionary theory was called for. But they had little to say about what it might look like. That’s what I’m supplying in the draft I’m currently working on and it appears that it will be useful in explicating that graph in ways I hadn’t anticipated. More framing.
I figure we need (at least) three interlinked conceptual frameworks:
- Statics: This framework characterizes the entities in our model, if you will, and specifies the structural relations between them.
- Dynamics: This framework characterizes the way these entities interact over time.
- Experimental Method: This framework links those two to empirical evidence.
Computational criticism has concentrated on the methodological issues comprising that last framework and hasn’t given any attention to the first two. Why not? Because, Moretti aside, almost know one is interested in explaining literary history using an evolutionary framework.
I, on the other hand, have spent a great deal of time on the first two. That’s why I latched onto Jockers’ graph back in 2014 when I first saw it. I could see that it somehow related to the conceptual framework I had been working on. That framework sees cultural evolution as a unidirectional phenomenon over the long haul. Just what that direction is, that’s not so clear. But it turns out that we don’t need to work that out in order to see Jockers’ graph as evidence for unidirectional evolution. That’s what I’ve been working in the various things I’ve written about it, including the draft material for this essay [1]. If I read her right, Da simply takes a unidirectional history without really having thought about it – I’ve argued this in the draft – while Jockers is just puzzled.
But again, what does this have to do with Moretti and Sobchuk? Their issue is treelike histories vs. nets; they seem to want their large scale historical hypotheses to hang on those structures (p. 112):
Tree-like, linear, reticulate . . . why should we even care about the shape of cultural history? We should, because that shape is implicitly a hypothesis about the forces that operate within history; the tentative, intuitive beginning of a theoretical framework.
What’s important about tree-like phylogenies is that you can use information about the current state of the system to infer the order of its past history (biology: vertical transmission of genetic information). You can’t do that with net-like phylogenies (biology: horizontal transmission of genetic information).
Jockers’ graph is clearly a net, not a tree. But it isn’t a cultural phylogeny either. It’s something else. And, as I’m typing this, I’m thinking that it’s something that you don’t even see in biological evolution. Why not? Because cultural evolution takes place in human minds, albeit human minds operating in a complex dance of cooperative and competitive interaction.
THAT’s why the current draft has been such a slog. Teasing that out has been difficult, and I’m not quite there. What surprised Jockers about the graph is that its nodes – one for each text – are in roughly chronological order, but there is no temporal information in the data on which the graph is based. Where did that order come from if there is no temporal information in the underlying data? It came from form. Each text has a form. Jockers used standard methods for characterizing that form. The graph thus represents the evolving stochastic form of the 19th century Anglophone novel. There is historical information in the relationships between the stochastic forms of those novels, just as there is historical information in the relationships between biological species where almost all inheritance is passed along vertically, giving us a three-like phylogeny. Horizontal transmission, which is characteristic of reticular phylogenies, obscures the historical signal.
Well, Matthew Jockers created a graph that has a historical signal despite the fact that it is network in form. That strikes me as being very important, as well as interesting. But it doesn’t explain itself. My objective in this essay is to lay the groundwork for a framework in which he can understand and work with that graph.
More later.
References
[1] William Benzon, On the direction of literary history: How should we interpret that 3300 node graph in Macroanalysis, Version 2, October 11, 2019, https://www.academia.edu/40550795/On_the_direction_of_literary_history_How_should_we_interpret_that_3300_node_graph_in_Macroanalysis_Version_2
[2] William Benzon, A quick guide to cultural evolution for humanists, November 14, 2019, https://www.academia.edu/40930224/A_quick_guide_to_cultural_evolution_for_humanists
[3] William Benzon, Culture as an Evolutionary Arena, Journal of Social and Evolutionary Systems 19 (4): 321-362, 1996, https://www.academia.edu/235113/Culture_as_an_Evolutionary_Arena
No comments:
Post a Comment