Att dikta med data

Generativ AI utmanar kreativt och litterärt skrivande

Abstract

Writing with Data: Generative AI and the Transformation of Literary Creativity

Literary studies have traditionally treated the finished text as the primary object of analysis. The emergence of generative AI in literary production challenges this approach, as writing increasingly unfolds through distributed processes in which textual outcomes are shaped across iterative interactions between humans and AI agents. This article argues that analysing AI-mediated literary writing requires a shift in attention from textual output to processes of textual becoming, and asks how creative agency, authorship, and stylistic decision-making unfold across human–AI collaborations in literary practices.

Drawing on postdigital and posthumanist concepts of diffraction and intra-action, we analyse three contemporary Swedish ‘testbeds’ of different AI-assisted literary practices: a poetry collection (Ammaseus horisont), a short story collection (Mening och motstånd i en föränderlig värld), and future-oriented narratives produced in a school intervention (Futures Day). Using process documentation from these sites, we examine how writing activities are distributed across successive phases of preprocessing, coprocessing, and postprocessing.

The analysis shows that the creative centre of gravity shifts across these phases depending on the configuration of humans and AI agents. In Mening och motstånd i en föränderlig värld, creativity is concentrated in prompt-based editorial dialogues during coprocessing, whereas in Ammaseus horisont it is located primarily in the preprocessing phase through iterative model design. Futures Day distributes creative decision-making across design, collaborative writing, and reflective postprocessing. Diffractive co-creation renders stylistic similarities and deviations between human- and machine-generated text perceptible, inviting critical interpretative practices that shape textual outcomes.

These distributed writing practices produce ambiguity regarding authorship and responsibility, as human and AI roles emerge and dissolve throughout the process. We argue that attending to textual becoming offers an interpretative approach for analysing AI-mediated literary works and their reconfigurations of aesthetic form.