Charisma (Embodiment): a Response to Tess McNulty

1:

In fall 2018, I made the trek across the Bay to Berkeley, where Andrew Piper was giving a talk drawn from his newly-published Enumerations. He presented results from chapter six, "Corpus (Vulnerability)," in which he modeled Edward Said's concept of "late style" on the oeuvre of several poets. Before revealing his results, he passed out slips of paper and instructed the audience to write down our predictions: how much does a poet's style change from earlier to later periods? We paused to reflect on the features in Piper's model (lexical, semantic, syntactic, and phonetic) and wrote our predictions for each feature's likely change. As he shared his results, one feature at a time, the audience responded audibly: surprise! consternation! jubilation! incredulity! The exercise elicited its intended response while preemptively defusing the most common dismissal of computational research: "We already knew that." There, in writing, it was clear: we didn't.

I loved it. So much so that, when preparing a talk of my own a few months later, I thought I'd take a cue from Piper (as I often do) and incorporate his exercise. I wanted an engaged audience, I wanted to convey confidence, and I wanted to give my results sufficient gravity. I tried it out in a run-through with some friends and they told me, emphatically but not unkindly, that it did not work. I lack Piper's je ne sais quoi — call it seniority, masculinity, renown, height. Call it charisma.

2

In her review of Enumerations, Tess directs our attention to Piper's use of models to challenge the supremacy of scholarly charisma. Charisma is at the heart of Piper's distinction between close reading and computational modeling in this science of generalization. Because the procedures of close reading are often hidden from view, readers must trust the critic's authority. The close reader is a sort of a magician, performing a feat of "critical insight," extracting meaning from a passage like a rabbit from a hat.1 Against this, Piper argues that in cultural analytics, "models replace charisma as the guiding vehicle for generalization."2 Not magicians but craftsmen, model builders make their process transparent, accessible, and replicable. Models enable more equitable, constructive dissent, because "the tools and information for mediating disagreement have been made mutually available."3 Tess hears "a democratic note" in Piper's defense of modeling. I do, too, but not without reservation.

"Charisma" is an evaluative term that is not applied or perceived equally; it's gendered, and racialized. To wit: the OED's definition of "charisma" ("A gift or power of leadership or authority; aura. Hence, the capacity to inspire devotion or enthusiasm") includes ten examples of usage, from 1930 to 1981. Not a single illustrative quotation refers to a woman, or uses female pronouns, when describing a person who has charisma; the charismatic leaders listed are all white.4 And so I applaud the rejection of "charisma" as a force that governs scholarly engagement. In proposing modeling as an antidote, Piper lays the groundwork for a more equitable approach to computational analysis and to literary studies broadly. This is especially clear in the conclusion, as Tess discussed, when Piper briefly introduces his research (conducted with Chad Wellmon) on the inequality of prestige journal publications. See Tess's review for a discussion, or read the longer piece at Critical Inquiry online (still, for my money, the best piece of computational criticism published by CI). "What would it mean to have a more inclusive academic system, one that is more responsive to a broader range of voices, including those outside the academy?"5 Piper asks. In addition to revealing how inequality works, Piper argues that modeling can also help us fix the problem by "generating more diverse ecosystems of knowledge," and "creat[ing] smarter, more adaptable, and more inclusive publishing platforms."6 In part, Piper asks us to reject charisma in order to embrace equity.

I find the rejection of charisma the most thrilling and promising aspect of the science of generalization that Piper advocates in Enumerations. In it, there's a real radical potential to upset academic hierarchies and institutional inequalities. Yet, much of this potential remains unrealized. Ultimately, I agree with Tess when she argues that Enumerations "makes too much of close reading's reliance on charisma [. . .] And it makes too little of model-building's." Modeling relies on charisma, not only because, as Tess argues, the average literary scholar is not well-versed in statistical methods and must put a great deal of trust in the cultural analyst, but also because Piper frames modeling as a personal and embodied project.

3

To Piper's credit, he does not embrace modeling as a mode of impersonal or apolitical empiricism. Rather, he argues that modeling "run[s] strongly counter to much of the language of empiricism that has surrounded the initial rise of the field. We are always present in our models." Piper insists on the researcher's presence in the model that she constructs. He repeats variations of the refrain throughout the book: we are "implicated in" our models, we "cede ourselves, fold ourselves into" our models.7 There are personal, sometimes physical, consequences to the recognition that we are implicated in the models we construct: models are "situated in time and space," and so modeling "slows us down" 8; as a "practice of mirroring" modeling asks us to be "self-reflective."9 Likewise, modeling has interpersonal stakes: it is a "social and collective" process.10 All of this leads to Piper's ultimate call to action: that we make ourselves "even more explicit in the models we build."11 This is a decidedly humanistic approach to data science, one that rejects the false neutrality of for-profit, algorithmic approaches. It is the most polemical aspect of Enumerations.

In discussing our implication in our models, Piper relies on bodily metaphors about how we look at and move through the world. Yet he stops short of acknowledging the ways different bodies embody differently. Perhaps Piper invokes these metaphors to describe the effect of modeling on our cognitive processes; he's not talking about actual bodies. And that is precisely the problem. I do not know how to think about myself without thinking about my body. I can't make myself explicit in the models that I build without explicitly acknowledging my (white, cis, straight, female) body. My cognitive processes — the way that I think and write and build models — are circumscribed by my experiences as a body in the world, and the products of my thinking and writing are received as an extension of that body. The two cannot be disentangled.

I learned this well when I tried to replicate Piper's model. In my talk, I used a similar method (linear discriminant analysis) to determine whether there is an "agent signature" in contemporary prizewinning American literature; in other words, can we tell who has represented any given novel? (There is, and we can.) My model is no less compelling, my feature set no less valid, and my results no less accurate than Piper's. And yet. My attempt at replicating his scholarly theatrics yielded entirely different results. Piper is right when he says that models mediate "between ourselves and our observations."12 But so do bodies. Because I am implicated in my model — because I am a woman and a junior scholar; because I am blonde and small; because I have a high-pitched voice and I speak quickly — it is received differently. This is why, though I want to believe Piper when he says that modeling might upset scholarly charisma, I simply can't.

---

Laura B. McGrath is Associate Director of the Stanford University Literary Lab and, beginning in 2020, Assistant Professor of English at Temple University.

---

References