Papers

Papers

• Nana Ariel, Anat Weisman, Nina Beguš. 'My Electronic Psychiatrist: A Poet Hacks ELIZA.'  Under review.
  
  We reconstruct David Avidan’s 1974 conversations with ELIZA, which appear to constitute the earliest book-length literary work composed through human–machine dialogue. It develops “poetic hacking” to describe how the Hebrew poet transformed the chatbot’s fixed rules, repetitions, and misunderstandings into literary resources without changing its code. Connecting his experiment with Sheila Heti’s chatbot writing, it shows how the conversational encounter itself can become a literary form.
  
  

• Kabir Ahuja, Jivat Kaur, Yulia Tsvetkov, Nina Beguš. 'Popular Art in the Age of Generative AI: Market Incentives, Media Theory, and the Paradox of Authorial Aura.' Under review. Preprint. http://dx.doi.org/10.2139/ssrn.7590703
  
  We combine economic and media theories to examine how generative AI may transform popular art, its markets, and audience experience. Drawing on Akerlof's market-for-lemons theory, Benjamin's account of aura, and McLuhan's tetrad, we show how uncertainty about authorship can favor lower-cost AI production, while expanded access to creation and hyper-personalization reshape artistic practice. We identify a paradox of authorial aura: credible human provenance may become more valuable precisely as conventional signals of authorship become less reliable.
  
  Recommended by That Business of Meaning.
  
  

• Nina Beguš, Tessa K. J. Haining, and Gašper Beguš. 'Latent Criticism of Finnegans Wake.' Preprint. Forthcoming in 2027. http://dx.doi.org/10.2139/ssrn.7155038
  
  Introduces latent criticism, a method that treats a neural network trained on a literary work as an additional object of literary interpretation. It examines FinneganLM, trained on the text of Finnegans Wake, and FinneGAN, trained on audio derived from the novel. Combining external analysis of generated text and sound with internal analysis of learned representations, it investigates how the models reorganize Joyce’s language, ambiguity, and narrative possibilities. The approach brings literary interpretation and AI interpretability together to explore not simply what machines can say about literature, but what kinds of machines literature can make.
  
  

• Roger Beaty, Nina Beguš, Kyle Booten, Malthe Stavning Erslev, Katy Gero, Karin Kukkonen, Søren Bro Pold, Janet Rafner, and Mads Rosendahl Thomsen. 'AI and Creativity: Fields of Tension.'  Forthcoming in 2027.
  
  This interdisciplinary article examines three tensions in AI and creativity: 1) the role of friction in creative processes, 2) the measurement and attribution of human–AI contributions, and 3) the design of alternative creative tools. Bringing together comparative literature, cognitive science, and critical digital humanities, it argues that evaluating finished outputs alone cannot establish gains in human agency, learning, or creative capacity. It explores how community-specific models, deliberately challenging writing tools, and experimental uses of chatbots can foster creative practices beyond seamless automation.
  
  

• Nina Beguš. 'World Wide Models: Literary Tools for Cultural AI. ' Modern Fiction Studies, Forthcoming in Summer 2027. Preprint on ArXiv (2026) and SSRN (2025).
  
  How can literary scholarship contribute to AI development? This paper brings comparative literature, narratology, critical theory, world literature, and translation studies to the analysis and design of language models. It develops a framework for examining AI-mediated cultural encounters, including structural monolingualism, the circulation of texts, and untranslatability. Beyond treating literature only as training material for LLMs, the paper identifies literary methods, theories, and tools as resources for building and interpreting culturally informed AI systems.
  - Level 1: Experimentation and evaluation (literary testing and narrative benchmarks)
  - Level 2: Theoretical approaches (structuralism v poststructuralism, pointing to structural monolingualism of LLMs)
  - Level 3: World Literature theory (concerning canon, circulation, translation, networks, power dynamics)
  
  Presented at NYU and TEXT. Responses in Elvia Wilk's Fast Writing & Public Seminar and in Good Internet.
  
  

• MLA Group for AI and Research. 'AI and the Humanities: A Framework for Language and Literary Scholarship.' June 2026.
  
  This collaborative working paper offers a framework for research in language and literary scholarship confronted with AI. Organized around AI literacy, human-centered expertise, and responsibility and ethics, it combines thematic questions with case studies to support critical reflection on AI’s role in scholarly research and practice.
  
  

• Nina Beguš, Metahaven, and Gašper Beguš. 'Latent Spacecraft: Brains, GANs, Finnegans.' Antikythera, vol. 1, no. 2, February 2026. Pdf can be found on the journal website and eScholarship. DOI 10.1162/ANTI.5KP1

  • Listen to FiwGAN, trained on English speech 

  • Listen to FinneGAN, trained on Finnegans Wake audio

This collaborative project combines literary interpretation, machine learning, neuroscience, and design to explore the internal representations in biological and artificial neural networks, particularly speech-trained, academic-lab-made GANs. The project uses interactive interfaces and the metaphor of architectural interiors to investigate how latent spaces can become accessible to linguistic interpretation and humanistic interpretability.

Reporting in Harvard GSAS, Berkeley Social Sciences, and Who's Afraid of Finnegans Wake documentary. Responses on Kat Zhang's Poetic Engineering, Rick Moss's Further Thinking. Andrej Tomažin's LUD Literatura, Outsider, John Rapko.
Presentations at OpenAI (shorter video), Beaux-Arts de Paris, Modes of Reading (Recording), AI and the Creative Condition, and Venice Biennale (journal launch), Warsaw MoMa, MoMa Mexico, among other venues.
The American Vandal podcast.
Exhibitions at Venice Biennale and Worth Ryder Art Gallery.

• Nina Beguš. Experimental Narratives: A Comparison of Human Crowdsourced Storytelling and AI Storytelling. Humanities and Social Sciences Communications October 2024. https://doi.org/10.1057/s41599-024-03868-8
  arXiv preprint from October 2023.
  
  This study compares 250 human-written stories with 80 stories generated by GPT-3.5 and GPT-4 in response to identical prompts about creating and falling in love with an artificial human. Combining narratological analysis with statistical methods, it examines how cultural narratives (the Pygmalion myth) and social biases (testing for gender, sexuality, race, ethnicity) appear in human and AI storytelling. Under the tested conditions in both experiments, AI-generated stories offered less imaginative scenarios and rhetoric than human-written stories, despite occasional innovative plot twists. The behavioral experiment ran in June 2019, and the AI experiment in March 2023 (its findings thus concern those models and conditions).
  
  The paper has been used in AI pedagogy (English 101 at U of Louisiana at Lafayette) and AI and writing courses (Future Text at Stanford U, Writing with Robots at U of Illinois Urbana-Champaign). It was also used for practical experimentation with LLMs.
  The paper was presented at the LMU München workshop, the Conference on AI, Science, and Society at the École Polytechnique, and at the Center for Digital Narrative at the University of Bergen.
  Reporting in The Financial Times, The Independent, Berkeley News, Tech Xplore, Incredibble, NPR, KCBS, and KNX Radio, MBR video, The Scholarly Kitchen, Illumination (English), PresseText (German), Haaretz (Hebrew), The Independent Turkçe (Turkish), ICT Global (Hungarian), LaSexta (Spanish), LeBigData (French), Scientias (Dutch), CNN Brasil (Portuguese), AIGeist (Italian), ICT Global (Hungarian), 续航教育/Forward Pathway (Chinese/English).
  
  

• Nina Beguš. Engaging Comparative Literature in AI Development. OSF preprint. July 2024. Updated version forthcoming in Modern Fiction Studies's issue on Cultural AI.
  
  This essay argues for integrating comparative literature into AI development rather than restricting literary scholarship to retrospective critique. It identifies cross-cultural expertise and literary analysis as resources for understanding AI’s ethical, philosophical, and sociotechnical implications. It calls for interdisciplinary collaboration that places diverse cultural and ethical perspectives within technological development.
  
  

• Nina Beguš. A Typology of the Pygmalion Paradigm. Collected Papers of the 21st Congress of the ICLA: The Rhetoric of Topics and Forms 4 (2021): 319-30.
  
  This article develops a two-part typology of the Pygmalion paradigm by focusing on the comparison between Ovid’s Pygmalion narrative, Hoffmann’s The Sandman, and the Silk Road tale The Painter and the Mechanical Maiden. It distinguishes the Pygmalionesque type, in which the creator is also the lover, from the agalmatophiliac type, in which creator and lover are separate characters. The comparison provides a framework for studying stories of human attraction to humanlike creations across literary traditions.
  
  

• Nina Beguš. “The Identity Problem” in Prenatal Testing. Voices in Bioethics 6 (2020): 1-4.
  Winner, Reproductive Ethics category, 2019 Voices in Bioethics Essay Contest, Columbia University.
  
  This essay develops the “identity problem” in prenatal testing as a response to the "non-identity problem" in ethics. The "identity problem" centers on genetic information changing our understanding of, and relationship to, an embryo or fetus. Reading Gattaca and Brave New World alongside reproductive ethics, it examines how medical prognoses turn into judgments about identity, disability, and desirable futures. It argues that fictional and autobiographical narratives should inform decisions about reproductive technologies, bringing individual experiences and human diversity into ethical deliberation.
  
  

• Nina Beguš. A Tocharian tale from the Silk Road: A philological account of The Painter and the Mechanical Maiden and its resonances with the Western canon. Journal of the Royal Asiatic Society 30/4 (2020): 681-706.
  
  This article examines six Sanskrit, Tibetan, Tocharian, and Chinese versions of The Painter and the Mechanical Maiden, a tale about a craftsman’s attraction to an artificial woman, set as a competition between two artisans, a painter and a mechanic. It identifies a close relationship between the Tocharian and older Chinese versions, investigates connections with Asian literary genres, and compares the tale’s motifs with Western renditions of the Pygmalion myth. The analysis traces how the Silk Road story of humanlike relation change through translation and adaptation as well as the archetypal presence of the Pygmalion story across folklores.
  
  The paper was taught in SANSKRIT 198 at Harvard U.