Tuesday, August 29, 2017

Reply to a traditional critic about computational criticism: Or, It’s time to escape the prison-house of critical language [#DH]

A couple of weeks ago I posted “In search of a small world net: Computing an emblem in Heart of Darkness” [1], which was about a computational procedure for establishing that a phrase we can call the Emblem is ‘central’ to Conrad’s text. Here is the phrase, in one of two slightly different versions: My Intended, my ivory, my station, my river, my—. A colleague asked me, in the person of more traditional critic: If this were true, why would we need computation to prove it? That is, since I’ve already made an argument in more or less traditional terms, what’s the point of the computational exercise?

To a critic for whom such questions are real, I have no response.

But I am not such a critic.

To be sure, my degree is in English literature, from the State University of New York at Buffalo in ancient days of 1978. But my dissertation was, in a large degree, an exercise in knowledge representation of the sort then current in the cognitive sciences. By the time I had completed my degree I ‘spoke’ both literary criticism and cognitive science with native competence. As a literary critic I can read Heart of Darkness, spot Kurtz’s Emblem, note its ‘centrality’, and make a fairly standard kind of case for it. But for my cognitive science side, what does that mean, ‘central’ to the text? The standard critical argument doesn’t satisfy my curiosity. Thus I’ve suggested the small-world-net exercise to satisfy the cognitive scientist in me.

But that’s not a reason for a literary critic to pay attention to that kind of argument. From my point of view, such a critic suffers from a limited imagination, an imagination confined to the prison-house of discursive criticism [2].

The standard argumentation I’ve offered about the centrality of the Emblem ultimately rest on intuitions about texts and language. What does it mean to talk of the semantic center of Heart of Darkness, or any other text for that matter? What do we mean by center, anyhow? Certainly not the physical center of the book itself. We mean the center of, well, you know, the ‘meaning space’ of the text. And what is that?

The fact is for years critics have been using loose spatial imagery in talking about texts. Texts have insides and outsides, so that some things are in a text while others are not. The text is capable of hiding meaning. It also has a surface, and that surface can be read. Such a so-called surface reading is of course different from a close reading, which is certainly different from distant reading. Do texts also have semantic edges as well? What about a top and a bottom?

We’re so used to this imagery that we don’t notice that it’s pretty much built on nothing. These vague spatial terms constitute our rock-bottom understanding of texts, not as physical objects, but as mental objects. We’ve agreed on these terms and have arrived at ways of using them but, really, what are we saying? There’s very little there.

It may see strange to think of such language as the bars and walls of a conceptual prison, but that’s what it is. It limits how we allow ourselves to think about our craft. We defend this language, not because it’s perspicuous, but because it’s comfortable – and we’ve grown lazy and complacent.

I've you've of a mind, better pour yourself a drink. This is going take awhile.

Beyond the transcendental homunculus: What’s next for literary criticism?

What kind of a future does this standard spatialized literary criticism have? Forget about current institutional pressures for the moment. In purely intellectual terms, what’s the future of literary criticism? Where is there to go?

For years I’ve been reading that there’s been no substantial new development since Barbara Butler and queer theory. I’m not entirely sure that’s true. Does animal studies count? What about eco criticism? Is Tim Morton a major figure? He’s given the Wellek Lectures (2014); he’s touring the world giving speeches these days; Bjork has endorsed his work and I believe he’s completing a general audience book for Penguin. Sounds pretty major to me. And I think he’s brilliant. But – the future of literary criticism?

What tools does discursive criticism have for thinking about the mind? I submit that the most important tools are the texts themselves – a theme to which I’ll return – along with the critics’ ability to spot interesting patterns in the texts. But when it comes to explaining those patterns, what tools?

As far as I can tell, those tools various kinds of mental faculties operated by invisible homunculi. That’s what psychoanalytic psychology (in its various forms) is like–and, by the way, I actually like psychoanalytic ideas and use them in my work [3]. As far as I can tell that’s the most explicit ‘model’ on offer. Ideology? What’s that but a hegemonic mental faculty operated by an authoritarian homunculus, and a transcendental homunculus at that, though we try hard to pretend that these guys (they’re always guys) are immanent. I’m exaggerating of course, but in comparison with the work that’s been done in the cognitive a neurosciences over the past half-century, the psychologies of literary criticism are pretty ghostly.

No, I believe that what we (in my person as a literary critic) have to offer is the texts themselves, and the patterns we spot and describe in them. Consider these remarks by Tony Jackson [4, p. 202]:
That is, a literary interpretation, if we are allowed to distinguish it as a distinct kind of interpretation, joins in with the literariness of the text. Literary interpretation is a peculiar and, I would say, unique conjunction of argument and literature, analytic approach and art form being analyzed.
That’s obvious enough, isn’t it? An interpretation is coupled to the text, and draws its substance from the text. Certainly it draws its claim to authority from the canonical text, not from the critic, who is merely an academic.

Taken by alone, the conceptual apparatus we use in interpretation is thin. Perhaps that’s one of the problems of critique, the ratio of interpreting theory to object text has tipped too far toward theory, resulting in an intellectually thin and unsatisfying discourse. Is it any coincidence that this era of High Theory has also given rise to the star critic who is granted the authority of that same Theory?

Can academic literary criticism thrive and prosper under a regime where theory works hard to pry itself free of literary texts, rendering them mere examples for displays of critical brilliance?

Are there other ways, newer psychologies, even computing?

The cognitive sciences, on the other hand, have a much richer range of ways of thinking about what the mind does. It’s just that they don’t quite get to literary texts. However, hat interests me, as I’ve said many times and ways, is computation [5]. I certainly don’t think the mind IS a digital computer, but as a way of thinking about processes and mechanisms, it beats homunculi, especially transcendental hegemonic homunculi. Of course, cognitive criticism (aka distant reading) doesn’t involve thinking about the mind as some kind of computational process. Rather, it uses computation as a tool for analyzing texts, both individually (e.g. character networks) and in large bodies.

Yet, there are connections, always connections. Let’s follow a few links. While many find the phrase “digital humanities” to be all but meaningless, let’s exploit that vagueness for a second. Digital humanities shades over into media studies and platform studies and there we find games. Several weeks ago I saw this tweet by James Ryan:

It caught my attention because it promised the "best overview of computational narrative modeling that I've read”. And that led to a recent computer science dissertation by David Elson from Columbia University (Modeling Narrative Discourse [6]).

I downloaded the dissertation and skimmed the section Ryan had pointed out and, yes, it is very good. Where does it start? Vladimer Propp’s Morphology of the Folktale. Then it works its way to the material I was reading when I was in graduate school writing abstracts for the American Journal of Computational Linguistics (now simply Computational Linguistics). I forget whether or not it mentions the work of David Ferrucci (leader of IBM’s Watson team [7]), but no mind. Both Elson and Ryan are in gaming (Ryan is doing a PhD at Santa Cruz). So I go to Ryan’s website and find that he working on an interesting project, History of Story Generation [8]. Bingo! Surely that’s digital humanities. If you follow Ryan’s twitter feed [9], it contains a wealth of documents in that history. Ryan is a fan of Sheldon Klein, who worked with Lévi-Strauss on the computational modeling of myth [10]. And Kline was a teacher of Roger Schank [11], who is perhaps the best known of the symbolic AI researchers on stories.

Why are game designers and programmers interested in the history of computational attempts to model narrative discourse? It’s not purely out of interest in history. They need those models in their own work. When you play a computer game you’re a player in a narrative. And the people who write those programs need models of narrative. What better source than real narratives, and descriptive and analytic work about those narratives.

And it’s somewhere in that mix that my little proposal about a small-world network in Heart of Darkness has a home. For it is grounded in a mathematical technique for modeling semantic space yada using high dimensional spaces. High-dimensional space? you ask. For the purposes of this piece it’s more than the three we can visualize. Depending on the particular application, we might be talking about hundreds or thousands or even more dimensions, many more. This YouTube video, “Thinking visually about higher dimensions” will get you started:

Thinking in high dimensions may be, certainly is, abstract. But it is more concrete than the vague spatial metaphors literary critics currently use, inside the text, surface reading, and so forth. The mathematics of computational criticism is thus one way out of the prison house of purely discursive criticism [2], but it isn’t the only way, as we’ll see in a bit. There is description, too.

In my graduate school days at SUNY Buffalo I learned computational semantics from David Hays. I developed a semantic model for a Shakespeare sonnet [12]. The idea was (and remains) that one produces a text by ‘walking’ a path through a semantic network. I never developed the model to the point where it could be implemented on a computer, but it was nonetheless still a real exercise. But the process of learning how to do that changed how I think about language and the mind. Even if we don’t know how to create a model adequate to literary texts – the stories generated by those AI programs were all toy stories – figuring out what you have to do to get anything at all was a bracing experience.

Those semantic/cognitive network models are quite different from the graph Matthew Jockers created to investigate influence in his corpus of 19th Century novels [13]. And it was Jockers’ corpus-wide graph that I had in mind in my small-worlds proposal. Now I’ve found a decade-old article where investigators undertook something like a virtual reading of 12 texts [14] to investigate reader attention. While my interest in the Emblem is somewhat different from theirs – they were modeling reader attention – it inspired me to offer a somewhat different proposal for virtual reading [15].

This proposal, too, involves tracing a path through the text, a path that connects disparate regions in a high-dimensional semantic space – the space of the text. The path is somewhat different in kind from that I envisioned in my model for the Shakespeare sonnet. And yet there is a distant kinship here, as the current statistical and neural net models arose from the collapse of those symbolic models. Moreover, abstract though it is, this path is far ‘closer’ to the text than any traditional close analysis and follows the text in greater detail.

It would be better, however, even for polemical purposes, to drop the metaphor of distance. It belongs to a critical era, an épistème if you will, that is receding into the past. We’re talking about a different kind of intellectual engagement with texts.

Description: The key to the treasure...

Consider these remarks by J. Hillis Miller about the influence of Northrup Frye and Kenneth Burke [16]:
I learned a lot from myth criticism [referring to Northrup Frye], especially the way little details in a Shakespeare play can link up to indicate an “underthought” of reference to some myth or other. It was something I had learned in a different way from Burke. Burke came to Harvard when I was a graduate student and gave a lecture about indexing. What he was talking about was how you read. I had never heard anybody talk about this. He said what you do is notice things that recur in the text, though perhaps in some unostentatious way. If something appears four or five times in the same text, you think it’s probably important. That leads you on a kind of hermeneutical circle: you ask questions, you come back to the text and get some answers, and you go around, and pretty soon you may have a reading.
He offered a similar remark about Derrida: “Derrida for me is even more important for his way of reading than for his invention of big concepts like différance.”

As John Barth has said, the key to the treasure is the treasure. Those patterns. And the critic’s first job is to identify and describe those patterns.

In the past few years a number of critics have been attempting elevate description–description, mind you! mere description!¬–to a central place in literary criticism. Writing in an issue of Representations devoted to description, Sharon Marcus, Heather Love, and Stephen Best [17], point out that “many [scientists] consider description an activity sufficiently worthy in its own right” (p. 1) and go on to observe (pp. 1-21):
We believe that description is a core, if unacknowledged, method in all scholarship and teaching. In order to proceed, interpretations, explanations, and prescriptions must give an account of—describe—what they interpret, explain, or evaluate. Description makes objects and phenomena available for analysis and synthesis, and is rarely as simple as its critics imply. An elusive object that travels by many names, and sometimes by no name at all, description’s dictionary definitions include representation, drawing, report, portrayal, and account. Description can take many forms, including lists, case studies, sequences, taxonomies, typologies, genealogies, and prevalence studies, and it involves many actions, including observing, measuring, comparing, particularizing, generalizing, and classifying, using words, images, and numbers.

And that’s where I started, with descriptive work on Heart of Darkness. I used my traditional skills to identify an interesting phrase and then, a bit later, switched to my cognitive/computational side and dreamt of a small-world net explanation for the potency of the Emblem.

I’ve written about description, endlessly it seems, so I don’t way to try to summarize that work here. It’s readily available on the web [18]. Rather, I would only make two points:
  1. Description is not necessarily verbal. It can take the form of diagrams as well. Visual devices are central to my own practice [19], and I was influenced in this by Lévi-Strauss.
  2. Description, as far as I can tell, is fundamentally a craft skill that comes only with practice. It doesn’t require a lot of abstract knowledge of critical theory, cognition, mathematics, or whatever else. You roll up your sleeve and work with the text.
The literary critic who really really likes the texts themselves and is impatient with abstract theory, that critic can cultivate the practice of description. And those descriptions can be used and built-upon by critics of other sensibilities, whether cognitive, evolutionary, or even good old High Theory.

And this takes us back where we began, with the critic who asked, If this were true, why would we need computation to prove it? If in asking that question they’re saying, I don’t give a crop about your stinkin’ computers, then, as I’ve been saying, I have nothing to say to them. What if their concern is a bit different, what if they’re asking, What happens to my skill at finding and describing patterns in texts? That’s a different kind of question.

And to that question I have a different answer: Those skills, finding and analyzing patterns in texts, will be more needed than ever.


[1] In search of a small world net: "Computing an emblem in Heart of Darkness." New Savanna, blog post, August 15, 2017: https://new-savanna.blogspot.com/2017/08/in-search-of-small-world-net-computing.html

[2] See my working paper, Prospects: The Limits of Discursive Thinking and the Future of Literary Criticism, November 2015, 72 pp.: https://www.academia.edu/17168348/Prospects_The_Limits_of_Discursive_Thinking_and_the_Future_of_Literary_Criticism

[3] I’ve written a number of posts on psychoanalysis at New Savanna. You can access them through this link: https://new-savanna.blogspot.com/search/label/psychoanalysis

You might want to start with: "Neural Weather, an Informal Defense of Psychoanalytic Ideas", August 25, 2013: https://new-savanna.blogspot.com/2013/08/neural-weather-informal-defense-of.html

[4] Tony Jackson. “Literary Interpretation” and Cognitive Literary Studies. Poetics Today 24 (2) 2003: 191-205.

[5] For my major methodological statement about computation and the literary mind, see Literary Morphology: Nine Propositions in a Naturalist Theory of Form. PsyArt: An Online Journal for the Psychological Study of the Arts, August 2006, Article 060608. http://www.psyartjournal.com/article/show/l_benzon-literary_morphology_nine_propositions_in

I have a more recent working paper that is worth looking at as well, Sharing Experience: Computation, Form, and Meaning in the Work of Literature (2016) 21 pp.: https://www.academia.edu/28764246/Sharing_Experience_Computation_Form_and_Meaning_in_the_Work_of_Literature

[6] David Elson. Modeling Narrative Discourse. Department of Computer Science, Columbia University, 2012: http://www.cs.columbia.edu/~delson/pubs/Modeling-Narrative-Discourse_Elson_R4.pdf

[7] Selmer Bringsjord and David Ferrucci. Artificial Intelligence and Literary Creativity: Inside the Mind of BRUTUS, a Storytelling Machine. Psychology Press: 1999.

[8] James Ryan. History of Story Generation. Website: https://www.jamesryan.world/projects#/storygen-history/

[9] James Ryan’s Twitter Feed: https://twitter.com/xfoml

[10] Sheldon Klein et al. Modeling Propp and Levi-Strauss in a Meta-Symbolic Simulation System. Computer Science Department, University of Wisconsin, Technical Report No. 226, October 1974: http://pages.cs.wisc.edu/~sklein/Simulation-Meta-Symbolique d'Hypotheses-Propp & Levi-Strauss.pdf

[11] Roger Schank. Wikipedia. Accessed 28 August, 2017: https://en.wikipedia.org/wiki/Roger_Schank

[12] William Benzon. Cognitive Networks and Literary Semantics. MLN 91: 952-982, 1976: https://www.academia.edu/235111/Cognitive_Networks_and_Literary_Semantics

[13] Matthew L. Jockers. Macroanalysis: Digital Methods & Literary History. University of Illinois Press, 2013. x + 192 pp. ISBN 978-0252-07907-8

[14] E. Alvarez-Lacalle, B. Dorow, J.-P. Eckmann, and E. Moses. Hierarchical structures induce long-range dynamical correlations in written texts. PNAS Vol. 103 no. 21. May 23, 2006: 7956–7961. doi: 10.1073/pnas.0510673103

[15] "Virtual reading as a path through a multidimensional-dimensional semantic space." Blog post. New Savanna August 26, 2017: https://new-savanna.blogspot.com/2017/08/virtual-reading-as-path-through.html

[16] I quote Miller in a blog post, "How to Notice Things in Texts, or The Key to the Treasure Really Is the Treasure." New Savanna, September 27, 2013: http://new-savanna.blogspot.com/2013/09/how-to-notice-things-in-texts-or-key-to.html

[17] Sharon Marcus, Heather Love, and Stephen Best. Building a Better Description. Representations 135, Summer 2016, pp. 1-21. Downloadable copy: https://academiccommons.columbia.edu/catalog/ac:201713

[18] I have many posts at New Savanna about description: https://new-savanna.blogspot.com/search/label/description

I have organized many of them into three working papers that you can download at Academia.edu: https://independent.academia.edu/BillBenzon/Description

[19] On the importance of visual devices in describing literary texts see my working paper, Description 3: The Primacy of Visualization, October 2015, 48 pp.: https://www.academia.edu/16835585/Description_3_The_Primacy_of_Visualization

No comments:

Post a Comment