Auditory Ethics

The Future of Humanities Research in the Era of AI

A few weeks ago a post by Ally Louks, a freshly minted Cambridge University Literature PhD garnered a lot of attention on X:

Most people drawing attention to this post were doing so in order to point out the apparent vacuousness of the ideas that were discussed in the thesis. They seemed to conform all the stereotypes about the modern academia. More charitable interpretations conceded that this would potentially be a very interesting topic of discussion, but the framing was too anchored in the tired and tested “victimology” overarching framework to be able to offer any genuinely deep and novel insights.

Many of us felt the deja vu, and the whole incident brought back flashbacks of the Sokal Affair and the Grievance studies affair.

My own first thoughts were that it would be possible to replicate the above thesis with modern AI tools. At the time I used ChatGPT to write a very plausible alternative abstract for the thesis. A few weeks later ChatGPT o1 and o1 pro were released, and o1 pro’s own reasoning and writing skills were in my opinion strong enough to handle larger quantities of writing. However, they were still not reliable and good enough to do their own primary research, and reliably gather all the references. I toyed with an idea to set a day or two aside to try to do it all by hand, but I could not justify taking that much time off from all of my other projects and obligations.

Finally, this past week OpenAI released Deep Search, a new tool that can do in-depth automatic research on your behalf. The first few impressions were extremely favorable. I had it write longish in-depth report on a couple of topics that I am right now very interested in exploring. The reports were both thorough and in depth, and did not seem to have any obvious hallucinations.

And that’s when I decided to test if the system is good enough to handle a whole literature PhD thesis in one shot. I decided to change the emphasis from olfactory ethics to auditory type. First I generated an abstract using o1 pro and the title of the thesis: “Auditory Ethics: The Politics of Sound in Modern and Contemporary Prose”. This is what I got:

This thesis investigates the intertwined phenomena of sound, ethics, and politics in modern and contemporary prose, arguing that auditory dimensions are central to understanding both the aesthetic and sociopolitical stakes of literary production. Drawing on a diverse corpus of twentieth- and twenty-first-century writers—including but not limited to James Joyce, Virginia Woolf, Toni Morrison, and Maxine Hong Kingston—this project interrogates how authors craft soundscapes as pivotal arenas for negotiating power, identity, and moral responsibility. Although the critical study of literature has often privileged visual or textual features, this study contends that the resonant spaces of hearing, silence, and vocalization constitute an equally rich field of inquiry, revealing hidden ethical imperatives that shape readers’ engagements with the text.

Building on theoretical foundations from sound studies, narrative ethics, and postcolonial theory, the thesis advances the concept of “auditory ethics”: a framework that underscores the moral ramifications of the way sound is depicted, perceived, and politicized in literary works. Central to this argument is the premise that sound—encompassing noise, voice, silence, and ambient resonance—acts as more than mere background detail; it both structures narrative form and encodes broader social tensions, including race, gender, and class. Through close readings paired with computational text analysis, each chapter examines how specific literary techniques (interior monologue, free indirect discourse, typographical experimentation, rhythmic repetition) shape readers’ auditory imagination while simultaneously reinforcing, contesting, or subverting prevailing power structures.

The research methodology combines traditional hermeneutics with evidence-based approaches in the digital humanities. By employing corpus-linguistic tools, it systematically surveys sonic motifs—such as whispers, echoes, and silences—across a large dataset of modern and contemporary prose. This mixed-method approach not only identifies recurrent auditory patterns but also isolates statistically significant deviations that highlight moments of creative subversion. Archival materials, including authors’ letters and original manuscripts, further illuminate the deliberate craft decisions behind the manipulation of sound. Interviews with contemporary novelists and data from historical sound archives complement textual analysis, positioning the project within an interdisciplinary dialogue that spans literature, cultural history, and acoustic ecology.

Ultimately, the thesis posits that listening ethically—attuning to how voices, noises, and silences structure the moral architecture of a text—sheds new light on longstanding debates about the social function of literature. In foregrounding the politics of sound, it challenges scholars to reconsider the experiential and ethical dimensions of reading, suggesting that future literary critique must account for the sonorous underpinnings of narrative if it is to grasp the full complexity of modern and contemporary prose.

I used the above abstract and lots of detailed instructions to get ChatGPT Deep Research o3-mini-high to write a full 60,000 word report, references and all. Deep Research asked me five more follow up questions, but then it was off to the races. It took about twenty minutes to produce the final artifact. The problem was that I had requested it to be formatted in LaTeX, but I was unable to convert the LaTeX file to pdf. So I redid the whole project, and this time I ended up with a moderately simpler Markdown formatted text. Even then it took me a couple of hours of negotiating with various mobile, web, and desktop tools and utilities before I was able to get anything humanly readable.

The final document seemed highly plausible and erudite to me. However, I have no academic background in literature or humanities, so I have no way of ascertaining how good of a thesis it is. But I did notice that even this document is mired in victimology cliches. It’s probably not something that we can easily exorcise from the AI’s thinking processes at this stage. Unfortunately this mindset has taken a deep and hard-to-expunge hold on most of the academic writing, especially in humanities and social sciences, over the past hundred years or so. We might need to wait for an AGI Zero to come around before we have a truly novel deep thinker that goes beyond our recent cultural baggage.

If you want to read the final document you can get the pdf version her. I’ve put all the artifacts, including the earlier LaTeX version into a GitHub repo that you can access here.