The North American Dostoevsky Society stands with all the people of Ukraine, Russia, and the rest of the world who condemn the Russian invasion of Ukraine. See our statement here.
by Fiona Bell
Dostoevsky’s relationship to revolution may be ambivalent, but speakers and attendees at this year’s NADS-sponsored ASEEES panel seemed to agree: he’d have hated the AI revolution. A comforting thought, maybe, but it doesn’t help much with a more important question: what are we going to do about it? How do these technological shifts affect our teaching and research, our communities, and, for crying out loud, our readings of Dostoevsky?
A continuation of a discussion begun at a roundtable on the “Digital Dostoevsky” computational text analysis project at last year’s conference, this year’s North American Dostoevsky Society panel featured the author’s reception history on Wikipedia, a digital humanities methodology called stylometry, and meditations on language and humanity as theorized, respectively, by Dostoevsky and ChatGPT. Lynn Patyk offered generous discussant comments, asking panelists to reflect more on the “loss and promise” of the digital age, how to navigate this “ménage à trois between author, human, and machine.”
Lindsay Ceballos charted the short history of the Russian and English Wikipedia pages devoted to the author. Reading the editing history of each page, along with the “talk” fora, in which volunteer editors discuss revisions, she critiqued the “polyphonic” quality of this site, whose author-editors are required to “pursue a possibility of neutrality to an absurd degree.” Ceballos paid particular attention to treatments of Dostoevsky’s antisemitism, tracking when this section appeared and disappeared, and closely reading the language of censure and shrugging that reported on this aspect of the writer’s politics. She concluded with a call to action: for scholars to make our best knowledge more publicly accessible through Wikipedia volunteerism. After all, she noted, large language models are sourcing information from Wikipedia. The least (or most?) we can do is make this site more reliable.
Katherine Bowers, presenting on the work of herself and absent coauthor Kate Holland, reported on their use of stylometry to compare narrative voice across Dostoevsky’s major novels. As “formalism nerds,” Holland and Bowers were interested in determining whether the stylistic deviation of the epilogue of Crime and Punishment from the rest of the novel could be represented or explained through stylometry. This computational method compares the relative frequency of specific character sets to determine similarity: as Bowers noted, it was recently used in attempts to unmask Elena Ferrate. Stylometry, in Bowers’s and Holland’s presentation, seemed to be a tool less suited to answering questions, but extremely productive in generating new ones. For instance, their analysis of the epilogue actually suggested that Parts I and II of Crime and Punishment differ significantly from Parts III-VI. Now, with more traditional reading techniques, we can determine why this might be. Bowers emphasized the defamiliarizing effect of stylometry on her and Holland’s reading practices: “This makes us uncomfortable,” she noted, but “it exposes things we can study with different tools.”
Suddenly, “the Golyadkin Jr. of writing,” as Chloë Kitzinger put it, appeared before us: ChatGPT. If you read op-eds, you’re familiar with the criticism that generative AI corrupts liberal notions of the human. The mark of humanity, these arguments go, is being an idiosyncratic individual, saying things in your own words, expressing and creating yourself simultaneously. ChatGPT robs students, in particular, of the opportunity to be and become themselves. In her paper, Kitzinger queried Dostoevsky’s “sources of the self,” particularly in Notes from Underground, indicating how the writer departs from this brand of liberal humanism while also warning against an eerily disembodied positivist dystopia. Original language, Kitzinger argued, is not necessarily an indicator of humanity in Dostoevsky’s world. The most ethical state, for him, may be silence. Yet, Kitzinger wondered how we might appreciate Dostoevsky’s critique of positivism without endorsing his “coercive collectivism,” one based on religious and national affinity.
Who does language belong to? This question has occupied our Bakhtin-inflected subfield for some time, priming Dostoevsky specialists to weigh in on artificial intelligence and language. At first glance, and with horror, we might ask: what’s more polyphonic than a large language model? If Bakhtin praises Dostoevsky for dissolving into his characters, then why aren’t we praising our students for boldly renouncing the bourgeois values of original authorship and intellectual property, academic honesty and honor codes, and dipping their cups in the sea of collective language? ChatGPT, I think, reveals the cracks in the Bakhtin industrial complex. The renunciation of individualism is not inherently good; what matters, of course, is the nature of the collective one enters. In Dostoevsky’s case, it was Russian ethnonationalism. In the case of AI, what appears to be “free language” is owned by corporations and sourced from the labor of others. Len von Morzé, writing for Radical Teacher, likens ChatGPT to a blood bank in which the generated text takes a transfusion from the “donations” (unwilling and unpaid) of thousands of flesh-and-blood writers from centuries past. Who “gave blood” so that we may generate text effortlessly today?
Last year, Eric Naiman suggested that Dostoevskyrequires “absolute individual responsibility” to avoid complicity with another’s evil. To use ChatGPT, he memorably concluded, is, for this reason, to burn Christ at the stake. To write po-Dostoevskomu, then, is to write in your own words, Devil may care. Or better yet (on Kitzinger’s reading) to write nothing at all. But Ceballos offers another way out: sometimes, writing is our best shot at redemptive forms of collectivity. Clearly, this panel is one of many conversations to come about artificial intelligence in our field. How fortunate that, rather than adopting a Dostoevskian brand of masochism—why does such a technology live?—the panelists modeled urgent, imaginative, and critical approaches to digital tools and the people who use them.
Fiona Bell is an Assistant Professor of Russian at the University of Utah and a member of the North American Dostoevsky Society’s Readers Advisory Board. Learn more about her work here.