PhD defence by Mirabelle Jones

Follow this link to participate on Zoom

Title

The Data is of Us But Not Us: Exploring Relations to Responsive AI Systems Through Speculative Participatory Research

Abstract

As artificial intelligence becomes increasingly embedded into society, invisibly and visibly intertwined in people’s daily lives and identities, it becomes crucial to consider the ethical implications of the resulting technological systems. While several researchers in HCI have pursued mitigating the potential harms and risks related to AI systems from a technological perspective through the development of AI and ML tools, others have considered these issues from a socio-technical lens, addressing the complexities of human-AI relations as deeply situated in the social, calling for interdisciplinary approaches. However, much of this research fails to consider knowledge formation as situated and embodied, discarding the importance of emotional, cognitive, and aesthetic knowledge as valuable findings in human-AI research. With the latter objective in mind, this dissertation explores human-AI relations as construed through how humans make sense of responsive AI systems by implementing an art-based participatory approach that highlights the importance of situated knowledge. I propose SPAR (Speculative Participatory Art-Based Research) as a method to further understandings regarding how people relate to emerging AI technologies through lived and felt experiences. SPAR adopts an inductive process of speculation, making, and scaffolding participatory artworks. I demonstrate this method in three speculative participatory art-based research projects focusing on generative AI systems that respond to and mimic human identities. In three papers and an annotated portfolio, I demonstrate the impact of: large language models on bodies using performance art, encounters with deep fakes of speculative alternative selves, and normativity and personalization in fine-tuned large language models. I conclude by speculating on the method’s application towards future research in an emerging fourth wave of HCI.

Supervisors

Principal Supervisor Jakob Grue Simonsen
Co-Supervisor Irina Shklovski
Co-Supervisor Christina Neumayer

Assessment Committee

Associate Professor Tariq Osman Andersen, DIKU
Professor Kristina Höök, Royal Institute of Technology (KTH)
Professor Steve Benford, University of Nottingham

Moderator: Tenure-Track Assistant Professor Valkyrie Savage, DIKU

For an electronic copy of the thesis, please visit the PhD Programme page.