Joint paper presentation with my dear Copenhagen colleague Mick With Berland at the conference “Culture and Technics”, University of Ljubljana, Faculty of Arts, December 3-5, 2025
The paper proposes a triangulation of potential artistic approaches to working critically with, through, and on AI, resulting in three coordinates:
AI imaginaries – practices of conceiving and creating imagery through machine learning applications, as well as imaginaries about the limitations and potentials of AI;
AI deconstructions – studies of the effects and affects fostered by AI and the relation to the technology’s cultural history, including the deconstruction of the technology itself;
AI exposures – examinations revealing the technological and systemic conditions of AI and its potential to facilitate investigative practices into realms of power and control.
The aim of these coordinates is to provide a flexible framework supporting the analysis of artistic practices engaging with machine learning (ML), be it on the level of automated tools assisting in the creation of the work, or as research object the work critically engages with. The proposed framework challenges the assumption that AI generators merely reproduce characteristics inherently present in the underlying datasets. Instead, it provokes us to investigate the question of how aesthetic practices may critically engage with generative AI-technologies.
The paper, co-authored with Mick and Tanya Ravn Ag, is based on insights developed in a research seminar we organized at the University of Copenhagen’s Department for Arts and Cultural Studies. Organized by scholars from media, art, and literary studies, it hosted artists and researchers working with and investigating the effects of ML technologies. The differing aesthetic and analytical approaches of the participants are discussed as case studies to explore questions relating to artistic engagement with ML: How can generative AI technologies be imagined, deconstructed, and exposed through critical aesthetic practices? How can such practices reveal the conditions and effects of ML technologies, i.e., issues of aesthetics, extraction, and autonomy? The proposed triangulation provides a framework for this endeavor while it is itself tested for its analytical viability and operational flexibility.
