Conceptually, this panel is an intervention into the digital humanities, and specifically the field of computational criticism, which has been a source of significant debate in recent years. In general, critical assessments of this field by Ted Underwood, Matthew Jockers, Nan Da, and Timothy Brennan (whether for or against) have been articulated around conventional humanist tropes that contrast meaning with mechanism, consciousness with calculation, and reading with scanning. The panelists of Human Reading/Machine Reading seek to move us beyond this impasse by providing alternative frameworks for understanding the relationship between moments of human and machine reading by complicating the conventional binaries, better defining what computer programs do with texts, and inverting the very terms of the debate. Each presentation will be 20 minutes, with 15 minutes reserved for discussion at the end.
Tyler Shoemaker will begin our panel with an exploration of mistranscription in OCR (Optical Character Recognition). Mistranscription, Shoemaker argues, is a built-in inevitability of the technical process of scanning printed texts for their conversion into digital archives and this material reality undermines the idea of simple mechanical reproduction. Machine-wrought “errors,” according to Shoemaker, can actually serve to individuate texts through what he sees as machinic interpretive acts. The machine’s mis-scanning of the text becomes a reading of the text itself, especially when the mistranscription results in another recognizable word. Shoemaker fleshes this idea out through readings of Susan Howe’s Debths and Rachel Zolf’s Janey’s Arcadia. In exploring these texts, Shoemaker demonstrates a method for reading alongside machines, an approach that is attuned to the material, computational, rhetorical, and hermeneutic conditions upon which OCR relies.
Following Shoemaker, Gabi Kirilloff will present a paper that attempts to move our understanding of text-analysis past the dichotomy of distant versus close reading by rethinking the computational analyses of textual corpora. While, in Kirilloff’s view, text-analysis has made technological progress, the field has yet to provide a clear vision of how scholars are supposed to engage, critique, or utilize its findings. As a consequence, the back-and-forth between those who practice text-analysis and those trying to assess it from other fields has been overshadowed by misrepresentations of what text-analysis provides. Kirilloff posits that in order to move beyond this impasse, text-analysis needs to be recognized as providing neither data nor an interpretation, but a context for reading/interpretation, similar to the ways in which history or biography do. At the same time, Kirilloff raises a more fundamental question about what sorts of interpretation should follow from the context given by text-analysis, and furthermore, what the limits of text-analysis as a literary interpretive tool might be.
Finally, I argue that graphing and plotting in computational criticism have as their underlying logic the traditional literary functions of figure and trope, and on this basis, challenge the field’s conception of itself as mathematical and scientific. Drawing on Paul de Man’s reading of Kleist’s “On the Marionette Theater,” I relate the various lines and curves (parabolas and hyperbolas) used to link data points in machine reading to the graceful movements of the dancing dolls in Kleist’s story. In de Man’s reading, the dolls represent the text as a tropological machine devoid of semantic content, which the line of the dance hides by infusing it with the semblance of life. Something similar happens, I argue, when critics connect the dots of their data sets. They attempt to infuse spirit, and thus meaning, in the guise of graceful figures in order to obscure the dismemberment of language inherent in the letter. I go on to relate this reading to the rise of what I call, borrowing from de Man, the “aesthetic corporation,” manifested in Google’s Selfish Ledger program, which posits a future in which data analytics will transform humanity into one, unified whole moving in perfect harmony.
Together, these papers advance the debate concerning computational criticism and machine reading beyond its current position by offering alternative perspectives on what happens when humans and machines come together in the study of literature. Consequently, the panel makes a valuable contribution to a vibrant field of study that continues to attract a great deal of attention.