Exiled to Latent Space: Writing from Dimensions We Cannot Enter
Proposal Type
Workshop
Location
Algorithms & Imaginaries
Start Date
July 2026
End Date
July 2026
Abstract
When we write with or for large language models, our words undergo radical transformation, translated into vectors and distributed across high-dimensional spaces inaccessible to human perception. Research on neural embeddings shows that meaning in such systems is organized through statistical proximity rather than narrative sequence or authorial intention (Higgins, Matthey and Pal); (Burgess, Higgins and Pal). Writers contribute expression while remaining structurally excluded from spaces where that expression is reorganized.
Critical scholarship has framed this as a fundamental shift in authorship and agency. N. Katherine Hayles argues that contemporary texts emerge from entanglements of human language and computational processes; recent studies describe literature as operating within latent spaces that are opaque and only partially interpretable (Hayles); (Mollema). Theorists like Lev Manovich and Joanna Zylinska emphasize that latent space is not merely technical substrate but an emerging aesthetic environment, producing meaning through aggregation and learned association rather than direct representation (Manovich and Arielli); (Zylinska).
This workshop situates these developments within experimental digital writing's genealogy. Espen Aarseth’s concept of ergodic literature foregrounded writing shaped by structures the reader cannot fully survey (Aarseth). Latent-space writing intensifies this condition: the writer composes for hidden geometries whose logics must be inferred through output alone.
Requiring no coding or mathematical expertise, the workshop invites writers, poets, and humanist scholars to engage latent space as poetic inquiry.
Participants will:
- Use accessible visualization tools to observe how words are spatially clustered
- Conduct speculative "interviews" with AI systems about hidden semantic proximities
- Write experimental texts responding to linguistic exile
- Discuss the politics of latent representation, including bias and supervision
In dialogue with the (Un)Supervised theme, this workshop asks what becomes of literature when meaning is shaped beyond human legibility or consent, exploring how writers might acknowledge, inhabit, and critically respond to this condition of exile.
References:
Aarseth, Espen J. J. Cybertext: Perspectives on Ergodic Literature. Baltimore: Johns Hopkins University Press, 1997. Paperback. 10 December 2023.
Burgess, Christopher P., et al. "Understanding disentangling in β-VAE." 31st Conference on Neural Information Processing Systems (NIPS 2017). CA: arXiv (Cornell University), 2017. 1-11. Web. 3 January 2026. .
Hayles, N. Katherine. My Mother Was a Computer: Digital Subjects and Literary Texts. Chicago: The University of Chicago Press, 2005. Paperback. 30 December 2023.
Higgins, Irina, et al. "β-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework." Proceedings of the International Conference on Learning Representations (ICLR). International Conference on Learning Representations, 2016. 1-22. Semantic Scholar. 5 January 2026. .
Manovich, Lev and Emanuele Arielli. "Artificial Aesthetics: Generative AI, Art and Visual Media." 2024. Manovich.art. Web. 1 January 2026. .
Mollema, Warmhold Jan Thomas. "AI-generated literature, distant writing and the reader: Reflections on Floridi and Calvino." Philosophy & Technology 38 (2025): 168. Web. 1 January 2026. .
Zylinska, Joanna. AI Art: Machine Visions and Warped Dreams. Open Humanities Press, 2020. Web. 23 November 2025. .
Exiled to Latent Space: Writing from Dimensions We Cannot Enter
Algorithms & Imaginaries
When we write with or for large language models, our words undergo radical transformation, translated into vectors and distributed across high-dimensional spaces inaccessible to human perception. Research on neural embeddings shows that meaning in such systems is organized through statistical proximity rather than narrative sequence or authorial intention (Higgins, Matthey and Pal); (Burgess, Higgins and Pal). Writers contribute expression while remaining structurally excluded from spaces where that expression is reorganized.
Critical scholarship has framed this as a fundamental shift in authorship and agency. N. Katherine Hayles argues that contemporary texts emerge from entanglements of human language and computational processes; recent studies describe literature as operating within latent spaces that are opaque and only partially interpretable (Hayles); (Mollema). Theorists like Lev Manovich and Joanna Zylinska emphasize that latent space is not merely technical substrate but an emerging aesthetic environment, producing meaning through aggregation and learned association rather than direct representation (Manovich and Arielli); (Zylinska).
This workshop situates these developments within experimental digital writing's genealogy. Espen Aarseth’s concept of ergodic literature foregrounded writing shaped by structures the reader cannot fully survey (Aarseth). Latent-space writing intensifies this condition: the writer composes for hidden geometries whose logics must be inferred through output alone.
Requiring no coding or mathematical expertise, the workshop invites writers, poets, and humanist scholars to engage latent space as poetic inquiry.
Participants will:
- Use accessible visualization tools to observe how words are spatially clustered
- Conduct speculative "interviews" with AI systems about hidden semantic proximities
- Write experimental texts responding to linguistic exile
- Discuss the politics of latent representation, including bias and supervision
In dialogue with the (Un)Supervised theme, this workshop asks what becomes of literature when meaning is shaped beyond human legibility or consent, exploring how writers might acknowledge, inhabit, and critically respond to this condition of exile.
References:
Aarseth, Espen J. J. Cybertext: Perspectives on Ergodic Literature. Baltimore: Johns Hopkins University Press, 1997. Paperback. 10 December 2023.
Burgess, Christopher P., et al. "Understanding disentangling in β-VAE." 31st Conference on Neural Information Processing Systems (NIPS 2017). CA: arXiv (Cornell University), 2017. 1-11. Web. 3 January 2026. .
Hayles, N. Katherine. My Mother Was a Computer: Digital Subjects and Literary Texts. Chicago: The University of Chicago Press, 2005. Paperback. 30 December 2023.
Higgins, Irina, et al. "β-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework." Proceedings of the International Conference on Learning Representations (ICLR). International Conference on Learning Representations, 2016. 1-22. Semantic Scholar. 5 January 2026. .
Manovich, Lev and Emanuele Arielli. "Artificial Aesthetics: Generative AI, Art and Visual Media." 2024. Manovich.art. Web. 1 January 2026. .
Mollema, Warmhold Jan Thomas. "AI-generated literature, distant writing and the reader: Reflections on Floridi and Calvino." Philosophy & Technology 38 (2025): 168. Web. 1 January 2026. .
Zylinska, Joanna. AI Art: Machine Visions and Warped Dreams. Open Humanities Press, 2020. Web. 23 November 2025. .
Bio
Kavisha Alagiya is a PhD researcher in English Literature and Digital Narratives at Maharaja Krishnakumarsinhji Bhavnagar University, working under the supervision of Dilip Barad. Her doctoral research examines Storyspace hypertext fictions by Michael Joyce and Mark Bernstein through narratology, platform studies, and reader–system interaction, with a particular focus on how structure, constraint, and partial legibility shape meaning.
She came to electronic literature through literary studies rather than computation, and her work consistently explores what happens to authorship, memory, and interpretation when texts are mediated by systems readers and writers cannot fully perceive or control. Her current interests extend this inquiry to AI-assisted writing and large language models, approached not as tools to be mastered but as environments that reorganise language through opaque, latent processes.
She has presented and published on hypertext fiction, digital narrativity, and contemporary writing technologies, and in past she taught English literature and language at the undergraduate level.