Gender Glitch Matrix: Queer Aesthetics and the Politics of Error in Digital Media
Author(s)
Akdoğan, Merve
DownloadThesis PDF (23.85Mb)
Advisor
Akšamija, Azra
Sass, Lawrence
Terms of use
Metadata
Show full item recordAbstract
Situated at the intersection of digital media studies, queer theory, and glitch art, this thesis critically examines the normative biases and centralization in artificial intelligence (AI) and, more specifically, machine learning systems as they relate to marginalized identities. Unlike conventional approaches that prioritize optimization and polishing of AI, this thesis introduces the notion of a glitch—a short-lived digital error—as both a metaphorical and an artistic technique that critically subverts societal norms. The thesis interrogates AI’s structure, dissecting it to reveal “black box” complexities to question the vulnerability of computational systems. It proposes an alternative approach that embraces error as a means of resistance, developing a critical commentary on technology production through artistic interventions. Grounded in Judith Butler’s “Matrix of Intelligibility,” the artistic interventions introduced in this thesis aim to craft a glitch aesthetic that integrates queer theoretical perspectives with practical machine learning applications. This thesis interrogates how AI models can potentially propagate entrenched societal norms about gender, what the political errors made by AI systems are and what can be the activist potential of technology in challenging these cisheteronormative renderings. Aiming to develop and test machine learning models for identifying bias in digital media, this research is organized into four sections, beginning with the development of a theoretical framework and a review of relevant literature on AI errors and glitch art. Subsequently, the thesis explores the design of glitch prototypes through training and testing machine learning models. Finally, through experiments using these methodologies, including archival work, media manipulation, and attribution studies with AI models, this thesis reveals the AI systems’ deficiencies as they relate to queer identities. This work underscores the transformative potential of integrating artistic techniques to subvert and reveal technological development. It envisions technology not merely as a mechanism for perfecting systems but as a powerful conduit for advocating a more inclusive future.
Date issued
2024-05Department
Massachusetts Institute of Technology. Department of ArchitecturePublisher
Massachusetts Institute of Technology