Distant Reading after Moretti
What follows is the text of a talk delivered at the 2018 MLA Annual Convention for a panel, “Varieties of Digital Humanities,” organized by Alison Booth and Miriam Posner. Marisa Parham, Alan Liu, and Ted Underwood were the other speakers. (Howard Ramsby was also scheduled to present, but he was unable to attend because of the blizzard). Ted’s remarks can also be found on his website.
I don’t have much time, so I’ll get right to the point—
The question I want to explore today is this: what do we do about distant reading, now that we know that Franco Moretti, the man who coined the phrase “distant reading,” and who remains its most famous exemplar, is among the men named as a result of the #MeToo movement.
I feel deeply for his victims. But given the context of this panel, what I want to focus on, today, is how his actions might prompt us to revisit some more longstanding issues regarding gender, power, and distant reading (which, following Andrew Goldstone, I’ll use in the lowercase-d lowercase-r sense to refer to the subset of computational methods that derive from statistical modeling and computational linguistics that are most commonly applied to analyze texts at scale).
Because sexual harassment is a structural, as well as personal problem, as Sara Ahmed has recently observed. By describing it a structural problem, Ahmed calls attention to how sexual harassment is sustained not only by the harassers themselves, but also by the institutions that shelter them. She explains how the confidential nature of most institutional inquiries ensures that “people remain, networks stay alive, and structures and processes are not put under investigation.” This is in large part because no one outside of the individual actors gets to know what happened, and as a result, the structural nature of the problem never becomes visible.
Ahmed’s work focuses on institutional structures, and academic institutional structures in particular. But the problems associated with ending harassment are not limited to academic structures alone. They also derive from flaws in cultural and conceptual structures as well—
So it’s here that I want to try to apply Ahmed’s lessons of structural power to the problems of power we face—still—with respect to lowercase-d lowercase-r distant reading in DH. Because as surprising as it might have been to some, when the allegations against Moretti surfaced, I don’t think it would be surprising to anyone in this room to bring up the many critiques that have been levied over the years at distant reading, and about how that particular field is, we might say, unwelcoming to women. These include critiques from the early 2010s by Moya Bailey, Miriam Posner, and the #transformDH collective about issues of representation in the field; critiques from around that same time by Bethany Nowviskie and others on Twitter that called out distant reading for its unduly masculinized rhetorical positioning; more recent work by Lisa Rhody, Tanya Clement and Jessica Marie Johnson that points out its failure to engage with the conceptual issues that relate to women—most obviously gender, but also sexuality, race, class, and ability, among many others. And then, in recent work by Laura Mandell, we see a critique of the actual computational models of gender that are often deployed when applying distant reading approaches to texts.
(And I would be remiss if I did not take a moment to acknowledge the critiques that have been levied against the #MeToo movement itself, by which I refer to its erasure of the voices of the Black women who did the earliest and most difficult work).
In fact, Ahmed’s structural critique of harassment helps show us how these issues are all related. Generalized racism and sexism, as well as the more specific issue of sexual harassment, each result from the same disparities of power, and other structural inequalities, that enable larger cultures of violence and oppression. There are many ways that this interrelation can be manifested, and structural power reinforced. Not all of these ways are easily explainable, or even traceable to a single source. But here is one example that can be quickly (if somewhat essentially) described: that flaw in prosecution that I mentioned just a minute ago, in which the actors, networks, and systems that enable harassment remain in place? This flaw leads to workplace environments that are unwelcoming (if not outright hostile) to women and other minoritized groups. But it’s those very same people who would otherwise be best positioned to identify and challenge the instances of sexism, or racism, or other forms of oppression that they see—not only in their institutional environments, but also in their scholarly work. Without those voices, conceptual structures, as well as institutional ones, remain securely in place, unchallenged and unchanged.
To put the problem another way: it’s not a coincidence that distant reading does not deal well with gender, or with sexuality, or with race. Gender and sexuality and race are precisely the sorts of concepts that have been exposed and interrogated by attending to non-dominant subject positions. And like literary world systems, or “the great unread,” the problems associated with these concepts, like sexism or racism, are also problems of scale, but they require an increased attention to, rather than a passing over, of the subject positions that are too easily (if at times unwittingly) occluded when taking a distant view.
So, then, what to do about it.
I think we need to start with our corpora. We need to assemble more corpora—more accessible corpora—that perform the work of recovery or resistance. An example: the corpus created by the Colored Conventions Project, which seeks to recover and aggregate evidence that documents the Colored Conventions of the nineteenth-century United States; these were organizing meetings in which Black Americans, both fugitive and free, came together to strategize about how to achieve social and legal justice. By making this corpus available for others to download, the CCP opens up the project of distant reading to texts beyond quote “representative” samples, which tend to reproduce the same inequities of representation that affect our cultural record as a whole.
We also need to rethink how we formulate our models. Instead of first asking what can be modeled—what phenomena we can track at scale—we might instead ask: what might be hidden in this corpus? And are there methods we might use to bring out significant texts, or clusters of words, that the eye cannot see? Another example: for the past several years, I’ve been working on a project that applies a set of distant reading techniques to a corpus of nineteenth-century abolitionist newspapers. I’ve been focused on the issue of gender, and on how the influence of key men and women can be tracked across the corpus in terms of both content and tone. But the limits of named entity recognition, which I used to develop my initial set of actors and events, required that I begin with the actors and events that could be automatically detected—rather than with the influences that could not be reduced to a single, computationally-tractable source. But it’s precisely the forms of influence that cannot be traced to a single point of origin that best reflect the distributed nature of structural power. And modeling those forms of influence is far harder, as I can personally attest. But those models that are increasingly necessary—lest we inadvertently reinscribe the same power relations that we intend to critique.
There are, of course, other things we can do better to connect the project of distant reading to the project of structural critique. But my time is nearly up, so I’ll end with this: it’s not that distant reading can’t do this work—it’s that it’s yet to sufficiently do so. But if we re-commit ourselves to the project of exposing and interrogating power, we could arrive at a form of distant reading that is much more capacious, and much more inclusive, than what we have at the present. Because the view from a distance, is, of course, as much of a view from a particular place as a view from up close. And it may very well be that a distant view that is trained on power, and that is self-reflexive about the forces that enable it—cultural and conceptual as well as computational—can contribute, significantly, to the project of dismantling structural power. Indeed, this project of critiquing power and working towards justice is the most pressing project of our time.