UVA Rotunda. (click for original)
Should a foreign language requirement for a literary studies PhD be fulfillable by a machine language? Or maybe even by a methods course (like a course in statistics, GIS, or some other competence in computational technology)?
These questions have been on my mind since I flew back Sunday morning from an energizing time at lovely (since it reminds me of high school) UVA, where I participated in the NEH-funded Institute for Enabling Geospatial Research in the Humanities, run out of UVA’s impressive Scholar’s Lab. It was a great time, I appreciate the NEH for tossing the cheese the Scholar’s Lab’s way, and I especially appreciate all the hard work the UVA people did to pull off a seamless little three-day event.
Over the next few days (read: weeks), I hope to write up more about the various things that went on at UVA, but the one that has been needling me most came from a quick flash of Twitter conversation on the topics in the lede, prompted by Brian (@briancroxall), who wondered whether GIS methodologies should count toward a “methods” course requirement in a PhD program, as it already does at Penn’s History Department. This then led toward the suggestion that, perhaps, proficiency in a machine language should be accepted as fulfillment of a foreign language requirement in a PhD program. Ryan (@ryancordell) even wished he had spent the time cramming for a French exam learning Ruby, instead.
As I mentioned while recusing myself, I’m kind of a militant about foreign languages (the more the better), to the point where I would consider basic Spanish proficiency a requirement of any specialist of American (U.S.) literature.
But after thinking about it, I realized that the two languages, human and machine, are, bizarrely, and in most mainstream use scenarios, totally non-comparable. Human language proficiency, at the literature PhD level at least, is measured in ability to read.1 The mere fact that my classmates can (and have) fulfilled their requirements by taking “Reading German” or Latin (dead language!) courses demonstrates that the key skill being taught is reading.
Someday being able to read a machine language, however, is never the goal of learning it. As Matt Kirschenbaum writes in his appeal to humanities students to learn to program (my emph):
Many of us in the humanities miss the extent to which programming is a creative and generative activity… Programming is about choices and constraints, and about how you choose to model some select slice of the world around you in the formal environment of a computer. This idea of modeling is vital.
From this promising beginning, Kirschenbaum comes out in favor of using machine languages as substitutes for human languages in fulfilling requirements.2 If one is studying contemporary American literature, in which code can appear, he argues, it has certain value.
But Kirschenbaum’s understanding of the value of the foreign language requirement seems misplaced.3 In the beginning of his article he criticizes those who view what we literary scholars do as little more than correcting spelling and grammar in comparison to those who think wrongly that computer scientists do nothing but fix bugs in code. So why, then, is he willing to imagine the foreign language requirement in such reductive terms when he says explains that (my emph),
Knowledge of a foreign language is desirable so that a scholar does not have to rely exclusively on existing translations and so that the accuracy of others’ translations can be scrutinized. One also learns something about the idiosyncrasies of the English language in the process.
Really? That’s it? Nothing about alterity, about imagining different conceptual schemes (pace Davidson), about forcing a disruption in one’s comfort zones (and comfort Weltanschauungs)? I should learn French just so I can take Massumi to task on how he continues the tradition of the translation of “agencement”? That seems a bit… thin.
Furthermore, when he pushes human and machine languages together, since knowing code helps one understand novels in which code appears, he is working against the very point of programming with which he opens his piece: programming is world-making. Sure, if I know C, I can read a novel that has pages of C. But I’m interpreting it and re-generating there, not generating tout court, as I would be with a “Hello World” program.
So the confusion in the Kirschenbaum piece gets reflected in the argument on Twitter: a willingness to compare human and machine languages, perhaps only since they are both called “languages,” emerges. This gets amplified, then, in the academic world; looking back at Penn’s history requirements, we see that they treat foreign language acquisition as a “competence” in a “technical area” (if I read the guideline correctly), akin to competence in GIS or statistics. Language is foregrounded–the technical option seems available more to US history scholars–but it is still treated as part of the same piece.
But history is not literary study; in an English department, I maintain, knowing French and python are completely different beasts serving two different masters in the scholarly process. The former relates to how stuff goes into the scholar (reading), while the latter relates to what comes out of the scholar (analysis). Exceptions are obvious, as one could publish in French and analysis usually involves quite a bit of feedback looping, but I think that I’m more or less right for the huge majority of cases.
So, on the one hand, I do wish that I could get the sense that computational methods training had greater appeal in the humanities.4 But, on the other, I do not think that it is entirely appropriate to view methods and foreign language as either/or objects in a course of study, when, as at my university, a student needs to show only proficiency in a single foreign language.
Given how my colleagues dismissively look back on their foreign language requirement (something like “I took ‘Reading German’ and remember none of it” is common), this distinction is probably unnecessary. In fact, as far as real-world skills are concerned, a much more lucrative future can be built up in a quarter-long course on Ruby than in the first quarter of first-year French. Human languages require care and attention, like Gabriel Conroy’s going to the continent to “keep in touch with the languages.” But though machine languages also benefit from practice (like any skill), it strikes me that it’s much easier to get back in the coding swing.
Still, maybe the question gets a bit more provocative if we consider the language requirement not at the PhD level, but, rather, at the undergrad level. Undergrads at my university can fulfill their math requirement by taking intro level CS courses that include vocational programming courses for web development, but the math requirement is certainly not the same thing as the foreign language requirement, which can be met in a dizzying array of ways. I cannot imagine that my university, which has made a huge deal over the past decade about expanding study abroad opportunities, would start accepting perl proficiency as meeting the foreign language requirement. And I’m not sure that’s bad.
Now it seems like we’re still comparing two things, human and machine languages, that can only really be compared in coarse ways, like through simple, macro-level questions of time management or speculation regarding future employability–two things, I think, that fall out of consideration at the PhD level (ha!).
So back to Kirschenbaum as I try to wrap up this wandering. HASTAC recently published a response to his article that reminds readers that one shouldn’t think that CS familiarity is demonstrated by machine language proficiency. This is certainly true: I know how to build and manage a GIS, but I don’t think I’m at all a geographer. I can code, but I’m not a computer scientist. And I don’t think my two years of formal French language study make me a scholar of French.
The article continues to discuss NLP, suggesting that the humanistic disciplines and CS have quite a lot in common. This is certainly true, too: my old CS roommate and I were often working on similar projects from different angles. But that then cuts out the final leg from the table that is Kirschenbaum’s argument. In describing programming as world-making, Kirschenbaum compares the coder to Jane Austen, a formidable world-maker herself of reasonable renown. Yet CSers working on NLP aren’t making worlds. They are investigating the world around us, just like us boring literary scholars.
- I fulfilled the requirement with Russian by showing that I was reading Russian literature and engaging in literary discussion of the texts without translation. [↩]
- He hedges, hardcore, by arguing that, in fact, it should be not only case-by-case but implicitly appropriate only for programs that require two foreign languages, which is not, as I mention, the case at my university, where one foreign language suffices. [↩]
- I am working under a bit of a fantasy here about what the foreign language requirement should do for PhD students. I know the reality is just about always different. [↩]
- My committee certainly had no issues with my pursuing further foreign language acquisition, or learning statistics, or taking a full year’s worth of GIS training. The promise to learn and work with GISes was written into my dissertation proposal, even! [↩]
Tags: assemblage, Brian Croxall, Brian Massumi, Deleuze and Guattari, digital humanities, geography, Geoinst, GIS, HASTAC, language, Matthew Kirschenbaum, natural language processing, programming, Ryan Cordell, the dead
June 1st, 2010 at 21:04
While I’m frankly baffled by the adversarial role the de Sá Pereira elected to adopt in his posting here (is it really necessary to brand ideas “snobbery” or to “cut legs out from under tables”?), let me say a few things in response.
It’s worth noting de Sá Pereira engages almost exclusively with one particular moment of my essay, the point where I raise the question of whether graduate programs in the humanities should accept a computer language in lieu of their traditional foreign language requirement. This is not a casual issue, as it goes to the heart, in a very specific and institutional way, of how new technologies and methodologies are gaining purchase in advanced humanities education. A number of graduate programs of which I am aware have already seen discussion and debate around this topic. There will be more. Nonetheless, the more general point of my piece, I hope it was clear, was not just that humanists should “learn to program,” but also that programming needs a new pedagogy, one that places it in the intellectual and imaginative spaces it deserves-—hence the dreary tale of my own introduction to formal computer science pedagogy at a public university in upstate New York twenty years ago.
But back to the issue at hand. de Sá Pereira suggests that I lack imagination when it comes to understanding that role of foreign languages in a humanities education (he brands himself, by contrast, as “militant” on the subject), that I seem not to appreciate that human and machine languages are very different things (he’s kidding, right?), and that I fail to acknowledge the role of “alterity” as fundamental to linguistic experience (something which, if we accept the binaries he proffers, one must assume computer languages to be then *incapable* of producing?). Additional (strange) binaries present themselves: coding is easy, natural languages are hard; coding is lucrative, natural languages are presumably less so. All of this seems to function mostly at the level of caricature, at least as far as knowledge of actual coding practices (and the job market for those with a single semester of Ruby) is concerned. We then get some confused thoughts about undergraduate education, followed by a concluding sentence that rings remarkably similar to the one with which I end *my* essay:
de Sá Pereira: “Yet CSers working on NLP aren’t making worlds. They are investigating the world around us, just like us boring literary scholars.”
Kirschenbaum: “Our students will need to become more at ease reading (and writing) back and forth across the boundaries between natural and artificial languages. Such an education is essential if we are to cultivate critically informed citizens — not just because computers offer new worlds to explore, but because they offer endless vistas in which to see our own world reflected.”
Personally I prefer Kirschenbaum, but that’s just me.
In any case, if those are our two bottom lines, then one might easily ask what de Sá Pereiran thinks he is objecting to in what I had to say.
The argument is most assuredly *not* that a computer language is “the same as” or “just as good as” a natural language. I honestly fail to see how any competent reader could take such a proposition from my essay. (One assumes it functions here for de Sá Pereira simply as a straw man.) That being said, I *would* contest any notion that computer languages are not sufficiently expressive. On the contrary, computer languages are the building blocks for some of the richest and most complex imaginative artifacts being produced today. Programmers routinely apply terms like elegant or even “beautiful” to code; while I don’t believe these terms are operating in the same way as they might with a natural language, there is an aesthetic dimension here which any serious discussion must acknowledge. (Don’t like my take on that? Try this one: http://pedrocr.net/text/computation-as-expression).
The second thing to say is that while there are thousands of computer languages, they should not be homogenized, something too often seen in discussions of this sort. Saying computer languages should be allowed to fulfill the requirement under certain circumstances (de Sá Pereira seizes on this as evidence of my “hedging,” I prefer to think of it as evidence of my sanity) is *not* the same as saying any computer language will do. I personally believe scripting and programming but not markup languages are appropriate, but that’s a separate conversation to have.
Finally, I don’t believe a computer language ought to be considered for substitution merely out of convenience, or even just a general interest in “digital humanities.” Rather, as with any exception to the normal dictates of such a requirement, the burden should be on the student to make the case as to why their needs are exceptional. Here are two scenarios where it would seem reasonable to me to take such a proposition seriously.
1. A student working on topics in digital culture and new media wishes to study a computer language for purposes of being able to “read” a digital artifact on the level of its material composition. This move is increasingly commonplace in digital studies, with examples of code literacy finding their way into recent scholarship by such figures as Ian Bogost, Rita Raley, and Noah-Wardrip Fruin. Code poetry is a particular phenemonon, but more generally scholars and theorists in new media are embracing what some term “critical code studies,” where computer languages are themselves understood as the cultural artifacts they are and exposed for serious cultural study. If a student here is working in this area then I think a case can be made for “reading knowledge” of a particular computer language along much the same lines that one makes the case for the medievalist’s ability to comprehend Latin.
2. A student working on the 19th century novel is interesting in “distance reading” after the style of Franco Moretti, in which one takes advantages of hundreds or even thousands of texts now in digital form to perform various kinds of analysis rooted in data minining and pattern recognition. She has familiarized herself with a number of the text analysis tools and software applications already available, but is unable to find one that matches her requirements. So she takes it upon herself to modify an existing program or even create a new one from scratch, regarding the process as being not only about creating a tool to meet her personal needs but contributing to the larger digital humanities community where others might in turn use (or modify) the new tool she has created. In this case, the ability to program is about self-sufficiency as a scholar: very much the sort of example we give with regard to natural languages, where we think the requirement is valuable so students don’t have to rely solely on the translations of others.
If neither of these cases displays an adequate level of imagination (or alterity) then it is unlikely de Sá Pereira or like-minded readers will be convinced. Having seen first-hand, however, how foreign language requirements are too often dealt with in various graduate programs with which I have been affiliated—-assuming that a semester of crammed coursework and contrived preparation is equipping the students for anything but *precisely* a machine (or machinic) translation, stripped of all genuine linguistic sensitivity—-I’m going to keep my mind open to some meaningful alternatives.
June 2nd, 2010 at 2:20
I disagree with nearly nothing that Kirschenbaum has written either in his original post or in this comment. We both seem to want the same things out of the future of humanities work, namely a larger role for advanced computing techniques.
I would like to restate, however, that my post emerged from the quick argument on Twitter, and I hoped the post would be enhanced by engaging with the reposting of “Hello Worlds” and from the contemporaneous HASTAC post over the long weekend, which I consider far more critical of Kirschenbaum than I was, since that post critiques Kirschenbaum’s essay from the title on down based on only a *potential, conceivable* reading of “Hello Worlds.”
The question of my post (and the argument on Twitter), which is *not* the main question of Kirschenbaum’s post, regards an either/or situation regarding human and machine languages. If a machine language were allowed to fulfill the foreign language requirement at my university, where there is only one slot, it would be a binary. Full stop. Putting the languages in conflict with each other, as they would have to be in this binary, to me, is a bad idea.
Kirschenbaum’s description of the role of programming, both in the original essay and in his comment, only reaffirms my position, which is that the languages seem so different that the binary seems unfair to both—to be avoided at all costs. I’m not sure that Kirschenbaum disagrees with this point, which I glean from his suggestion that (for now?) machine languages be allowed to fulfill requirements on a case-by-case basis only.
So, then, what’s the disagreement? Kirschenbaum writes:
“I believe proficiency in a computer language can fulfill many of the same functions — accessibility, self-reliance, heightened critical awareness — as knowledge of a traditional foreign language.”
For me, this suggests a certain kind of interchangeability, which I don’t think is right, as I see the kinds of languages as steps in a process. The foreign language helps on the way into the scholar, whereas the computer language helps on the way out.
Type B Digital Humanists benefit (more?) from knowing how to code as tools in analysis (in doing distant reading, for example), not as part of the (initial) experience of the text. The expressive potential of code, as Knuth reminds us via Kirschenbaum’s essay, helps ease the collaborative aspect of a coding project. This is, to me, still a rather different experience of a language than (pardon the clunky example) reading a chunk of code in a novel.*
So with my stated-in-the-post caveats in place—we’re assuming one slot for any kind of language fulfillment, we’re assuming a “mainstream” situation, and we’re assuming a fantasy of foreign language acquisition that is not the cramming for the exam that yields a machinic translation of dubious use—I still think the slot is better used with a foreign language. My caveats may render the point so diluted as to be useless, but my response to the original argument (again, *not* Kirschenbaum’s post) was precisely within my own situation, in which those caveats all *do* apply.
In even starker terms, to highlight what viewing the two kinds of languages as interchangeable looks like: Given the choice of either understanding a chunk of code in a novel or a note in [human language] in the novel, I would choose the human language over code.
* Thinking over this response, there is a parallel I hadn’t thought of at first. The benefit I’ve heard of knowing a foreign language for PhD-level work—that is, the semi-official one, as opposed to the one I have imagined—is not so that one can read *novels* in the foreign language, but so that one can read *criticism/theory* in the language. It is possible that, in the future, literary criticism will come to us in the form of code, in which critics attach the scripts that yielded their results. That would be a fascinating development, but I don’t know if there’s a precedent for it—do economists append their Stata files to their articles? But I doubt that we’ll ever have a journal that is nothing but code, so there will still be some sort of intervening human language for analysis—a language that will not necessarily be the language that the scholar already speaks.
June 2nd, 2010 at 3:34
“If a machine language were allowed to fulfill the foreign language requirement at my university, where there is only one slot, it would be a binary. Full stop. Putting the languages in conflict with each other, as they would have to be in this binary, to me, is a bad idea.”
The critique here seems to be that given the realities of academic programs and graduate requirements, there will inevitably be those instances where computer languages and natural languages are treated with parity–and forced into “conflict”–because if only one language is required then a student will study one to the exclusion of the other.
Fair enough. This is the situation in my own graduate program, where also only a single language is required. Here I am a pragmatist. If proficiency in a particular computer programming language will enhance a student’s education and scholarship in a meaningful way–for example, along the lines of one of the scenarios I have laid out, two out of what I would assume could be many–then I can live with that same student missing out on revisiting French in Action in order to “force a disruption [their] comfort zones (and comfort Weltanschauungs.” I have colleagues who feel otherwise, so the matter is hardly settled–we’ve not yet had a good test case here–but everyone recognizes that the language requirement only sometimes meets the lofty goals set forth for it in the admissions brochure. I’m open to creative ideas for how the requirement might be reinterpreted and reimplemented, whether they involve Ruby on Rails or not, so long as those ideas continue to serve a committed scholarship and help reduce time-to-degree.
June 3rd, 2010 at 0:27
Yeah, and I’m grateful for your response since it forced me to realize how much I was hedging to satisfy my fantasy of massive foreign language acquisition. I spent quite a bit of today playing administrator and trying to imagine what to do about this question, and I think that the best solution would be to have a more emphatic “skills and methods” requirement in coursework. Between undergrad and grad school, I took three and TAed one course that involved what I’d call “extended techniques” for a literature course–that is, more than reading from a reading list, participating in class, and writing papers / to the discussion board. And all three were some variant on “go to the library and read a bunch of stuff that isn’t on the syllabus and then present on it,” where the “bunch of stuff” was always deeply historically particular: a decade’s worth of a journal, a person’s papers in Special Collections, a school of poets.
(My program also has a reqd pedagogy/course design course in our third year and a cohort-wide colloquium fall of our first-year, but neither of these feels like a “skills and methods” course *for research*.)
There’s space for more here, and it gives me an idea for something to write about later for fun.