My friend Sarah Wasserman, who is finishing up a brilliant dissertation on the idea and instances of ephemerality in 20th century American literature, recently sent me a couple of online columns written by Stanley Fish a few months back on the subject on the subject of blogs and the digital humanities.
Ephemerality has something to do with why Sarah sent these columns to me, but let me keep the precise reason on hold for a moment so that I can get on the table the three basic claims Fish makes about the digital humanities. These have made me reflect on my experiences this past year with the Foreign Literatures in America project, as well as my more general introduction to the ideas and culture of the digital humanities that have come through the consistently and remarkably stimulating intellectual environment, personnel, and speakers I’ve encountered at MITH.
The first argument that Fish makes is that, entwined with its own sense of excitement, there is a “religious” quality to the digital humanities—one that digital practitioners themselves might be reluctant to admit, but that manifests itself in a consistent self-descriptive discourse of revelation: one in which solitary authors and critics, finite texts, and definitive and authoritative understandings are radically displaced by digital horizons of the intellectually infinite and indefinite as well as the socially collective and cooperative.
Fish likes Milton, and so it is in the Miltonian image of the “all in all” that Fish describes this “theological” digital self-conception that “promises to liberate us from the confines of the linear, temporal medium in the context of which knowledge is discrete, partial and situated… and deliver us into a spatial universe where knowledge is everywhere available in a full and immediate presence to which everyone has access as a node or relay in the meaning-producing system.”
I like Emerson, and so, when Fish describes a digital vision in which “
And I think Fish is right to suggest that this theological (and very American, if you ask me, as an Americanist) narrativity does tend to inform digital humanist self-presentation to some degree that is worth thinking about.
For Fish, this “expansive” inclination in the digital humanities runs the risk of denying the finite, linear, and mortal conditions through which, much of the time at least, individuals actually think and certain forms of knowledge can only become generated. Here I think Fish has a valid point: digital humanists seem to me frequently more comfortable operating in iconoclastic mode, i.e., liberating humanists from prior dogmas such as traditional conceptions of the library, traditional models, valuations, and understandings of textuality, and the traditional divorce between humanistic and scientific/empiricist methods (all subjects I’ve heard several talks on this year, and good ones), than they are at convincingly explaining what can and should be done with the New Freedom.
As with Emerson, there is a kind of voracious ambiguity by which the digital humanities sometimes rhetorically position themselves—through alternatively abstract/visionary and narrowly technical vocabularies—to swallow up a world whose actual conditions of happiness, insight, and self-control may in important respects be contingent, cumulative, and time-bound, subject to disciplining restraints and material limitations not referable to, or at least in need of checked relation with, digital totality.
Yet I must confess to some ultimate sympathy with this religious vision: first, because I have been genuinely surprised in my brief time at MITH at how easy it has been to let go of certain formerly sacred ideas (like the traditional library: not books, but the monolithic model of the library as physically browsable warehouse), which, if not exactly a religious conversion on my part, is nevertheless the rejection of a kind of orthodoxy that I have tended to take as seriously as any other; second, because I am not surprised by but am very impressed with the transformative possibilities of productive knowledge that lie at the intersection of humanistic and scientific/empiricist method and also collaborative interchange between parties of different expertise (more on which in a moment); and third, because the digital humanities’ revolution and revelation, like all things in the Gutenberg continuum, is here to stay—the right question to ask is how to administer it in the interests of insight and justice.
This leads to Fish’s second point about the digital humanities: that it is predicated not simply on theological revolution but a sense of political revolution as well—one that seems to intertwine an insurgent and open-ended sense of who and what is and should be a part of academia with a more broadly democratic commitment to knowledge dissemination, empowerment, production, and discussion generally. In this conception of the digital humanities, Fish presents its practitioners styling themselves in genealogical descent from the 1960’s/70’s cyber- and counter-culture, and, though not inherently pegged to any particular politics in itself, the digital humanities implicitly find themselves embracing a “left agenda” that “self-identifies with civil liberties, the elimination of boundaries, a strong First Amendment, the weakening or end of copyright and Facebook/YouTube revolutions that have swept across the Arab world.”
There is too much to bite off on here that is really not germane to this blog, so let me simply advance two quick thoughts. First, though I’ve become convinced of the genuinely revolutionary possibilities (not all aspects of which, I should clarify, I am thrilled with: I am not convinced that digitality has produced a culture of better readers—though I’m not exactly convinced of the reverse either) the digital world offers to transform not only the substance but also the methods of education, it is not really clear to me that the “left agenda” described above really is predicated on political “revolution” so much as a ratification of current and dominant America/Western/neoliberal conventionalities and momentum; these values are the extension of democratic capitalism, not really—as a matter of inherent structure—affording any critical nor certainly any revolutionary counter-balance or upheaval of those processes.
This leads to a second point: that the challenge for projects like FLA, and as actually practiced in important progressive digital initiatives like the Digital Public Library of America and Carla Peterson’s Black Gotham archive and the national digital education initiatives advanced by the National Humanities Center, is not simply to assume but to think carefully about the kind of transformations the digital world can effect, and with what liabilities too. For FLA, whose archives are beginning to assume substantial proportions: I am both heartened but also concerned about who and what will benefit most from the tools it offers to analyze political manipulations of non-American-authored novels in the U.S. over time and in different political contexts.
Fish’s third point is his most important point, and the only one, based on my experiences at MITH, that I fundamentally differ with him on. Here, Fish essentially argues that the methods of the digital humanities are fundamentally distinct from, and not capable of supplanting or even meaningfully, in any originating way, contributing to, the kind of insights that derive from traditional humanities, as epitomized by learned, cumulative, nuanced practices of close reading.
For Fish, the big dividing line between the traditional humanities and the digital humanities is that whereas the former presuppose an interpreter/reader who originates a hypothesis, which only then becomes tested and explained and elaborated through a process of engaged reading and pattern demonstration, the digital humanities reverse this process:
“First you run the numbers, and then you see if they prompt an interpretive hypothesis. The method, if it can be called that, is dictated by the capability of the tool. You have at your disposal an incredible computing power that can bring to analytical attention patterns of sameness and difference undetectable by the eye of the human reader. Because the patterns are undetectable, you don’t know in advance what they are and you cannot begin your computer-aided search (called text-mining) in a motivated — that is, interpretively directed — way. You don’t know what you’re looking for or why you’re looking for it. How then do you proceed?”
But of course you can begin pattern searches in “motivated ways.” One of the tasks we’re engaged in at FLA is trying to train computers to read big piles of data—concerning the American reception of non-American authors: we’re trying to figure out what the politics of international literary reception in the US is—based on the precedent of annotation questions we humans have asked of a smaller set of data, which the computer can project forward. Six of us, for instance, have asked of 100 American magazine articles that reference Dostoevsky and Tolstoy in the early 1900’s questions like: who is/are the principal literary author(s) discussed in this article; is a generally positive, generally negative, indeterminate, or no opinion expressed of that author; is gender an issue explicitly raised in the article; is race; is socioeconomic class; is radical politics; is America or the West invoked as a point of similarity; or contrast?
Now Fish would be amusedly impatient with these questions because he knows how difficult, loaded, and ambiguous all of the key terms involved are.
But we know this too, and we spent a series of months refining our questionnaire and developing individualized conventions for answering our questions based on specifically applying the questions to small subsets of data, week by week refining our conventions in light of problems and cruxes arising from application of those questions to data ultimately representative of a larger sample, all the while actually really proceeding on the model of legal analysis—whereby you have a legal rule or principle or statutory provision and you make your decisions case by case developing a body of hopefully consistent precedent.
The proof is in the pudding here: with our annotation questionnaire, applied blindly by two or three different team members, we have now achieved remarkable consistency across nearly all of the 17 questions we’ve sought to apply and use to annotate the data, for replication (hopefully) by a machine when patterns are extended by it to larger stockpiles of texts to analyze.
We came armed with plenty of “motivating” hypotheses—that’s why we picked the questions we did in the first place—but we’re pursuing them in tandem with the machine: first, by annotating our data in a systematic fashion whose patterns (invisible to us, and/or more extensive across great big fields of data than we can read) we hope the machine will be able to read; and second, by then being able to scrutinize keyword and collocation patterns based on the categories of annotation the machine will sort things into.
Fish I believe would respond—and the traditional humanist in me would agree with him—: but don’t you then actually have to go back and read (closely) yourself whatever articles sorted in whatever categories featuring whatever keywords—to understand the true meaning of those words and that article? But the answer is that the computer can help identify patterns to explore in relation to one’s hypotheses about meaning that one didn’t know existed in the first place.
The same thing is actually true of the kind of topics modeling procedures where—unsupervised by any “annotating” training or human intervention—tools like MALLET or WEKA identify clusters of words, themselves presumably organized in terms of “subject” or “topic,” in, say, a field of data that presents itself over time. I see no problem with actually letting the computer generate “topic” word strings in and of itself (though it’s more complicated than this: a human being is always playing a role in establishing the number of topics) and then trying to come up with hypotheses based on that as to what the word strings mean, how they fit together and why, But even more than this, new topic modeling procedures allow “motivated” human beings to intervene after the initial topic model generation, either to filter out or to emphasize specific words a human “expert” thinks is either very relevant or noise, so as then to generate more refined and revealing topic models.
Again, the process is a kind of dance (though I hate the sentimentality of that word) between human and machine, between technological specialist and humanist, between a substantive expert armed with hypotheses and also open to questions and puzzles and an inhumanly large field of data to which such experts have to make themselves answerable if they want to ask questions relevant to that domain in the first place. (This last point is important: if, as a humanist, you are making claims about a big field of historical textuality, and you can’t legitimate your larger claims in confirming relation to what distant reading models suggest, then maybe you are wrong—indeed, I think the burden is then on you to explain how you are actually right.)
All of which returns me to my friend Sarah, a gifted literary critic and philosopher of the ephemeral, who sent me Fish’s columns in the first place. She did this because she knows that I have deep sympathies with the note on which Fish begins his first article described above: grumbling (masterfully!) about blogs, which for him is a kind of synecdoche for the digital humanities as a whole, on the grounds of their self-consciously ephemeral character!
Fish has this ability to get one to argue with oneself; MITH does too. I’m very fortunate to have come into contact with them both.
Peter Lancelot Mallios is a MITH Faculty Fellow. He directs the Foreign Literatures in America Project and is a professor in English at the University of Maryland.