The Hauntology of Generative AI in Design
Paper i proceeding, 2026
Generative Artificial Intelligence (GenAI) is a fundamentally hauntological technology. This paper interrogates that character of contemporary AI systems and explores why this feature undermines the goals of design research. We argue the need for an exorcism of GenAI, or at least a constructive rebellion against the ghosts it systematically produces. GenAI is convergence-driven and nuance-averse. It aims for reasonableness as a first-order output. Its training data reflects a profound anthropocentrism that extinguishes plurality. As a result, GenAI accelerates the foreclosure of possible worlds by preferentially reproducing that which is statistically probable and culturally dominant. By contrast, design relies upon idiosyncrasy, rupture, serendipity, and the cultivation of plurality. In short, the very qualities systematically eroded by the probabilistic logic of generative systems. Through a lens of glitch feminism, we propose error, derangement, and an enthusiastic embrace of the uncomputable as resistance strategies to retain what is fundamentally human within inhuman systems.
design research
genAI
hauntology
glitch feminism
Författare
Michael Heron
Chalmers, Data- och informationsteknik, Interaktionsdesign och Software Engineering
Göteborgs universitet
Klara Aune
Göteborgs universitet
Mafalda Samuelsson-Gamboa
Göteborgs universitet
Chalmers, Data- och informationsteknik, Interaktionsdesign och Software Engineering
Sofia Thunberg
Chalmers, Data- och informationsteknik, Interaktionsdesign och Software Engineering
Göteborgs universitet
Uppsala universitet
Negin Hashmati
Göteborgs universitet
Chalmers, Data- och informationsteknik, Interaktionsdesign och Software Engineering
Pauline Belford
Göteborgs universitet
Chalmers, Data- och informationsteknik, Interaktionsdesign och Software Engineering
Proceedings of Drs
23983132 (eISSN)
Vol. DRS2026 119
Design Research Society, DRS 2026
Edinburgh, United Kingdom,
Edinburgh, United Kingdom,
Ämneskategorier (SSIF 2025)
Människa-datorinteraktion (interaktionsdesign)
DOI
10.21606/drs.2026.786