HCII PhD Thesis Proposal: Jordan Taylor
When
-
Description
Queer Computing: Rethinking Users, Uses and Utility Functions
Jordan Taylor
HCII PhD Thesis Proposal
Date & Time: Thursday, June 12th @ 10:00 a.m. EST
Location: Newell Simon Hall (NSH) 4305
Remote: Zoom Link
Committee:
Haiyi Zhu (co-chair), Carnegie Mellon University
Sarah Fox (co-chair), Carnegie Mellon University
Sherry Wu, Carnegie Mellon University
Oliver Haimson, University of Michigan
Amy Bruckman, Georgia Institute of Technology
Abstract:
Human-computer interaction (HCI) scholars are increasingly engaging in research about marginalized groups, such as LGBTQ+ people. This research often focuses on technological harms in service of including marginalized communities in the design of mainstream technology. Some have critiqued harm-centered research for reducing marginalized communities to their experiences of marginalization. In the first part of my proposal, I build on this work by examining the representation of LGBTQ+ people in HCI research through a systematic literature review. Then, I consider the implications of different ways of framing of LGBTQ+ people in HCI research through a diffractive analysis of a gay online community. Across these studies, I identify opportunities for HCI researchers to contend with the totality of LGBTQ+ people's lives—including and beyond experiences of harm—and to recognize LGBTQ+ people's agency to resist and re-imagine technology. In doing so, I contribute to both Queer HCI scholarship and ongoing debates over how to conduct research with marginalized communities in HCI.
In recent years, there has also been a sharp increase in HCI research on Artificial Intelligence (AI). Like HCI research on marginalized communities, critical computing scholarship on GenAI typically focuses on harms, such as representational harms against queer people and allocative harms against artists. However, queer artists' agency to make use of or resist these technologies is under-considered. In the second part of my proposal, I examine queer artists' perceptions of and material engagements with GenAI. First, I conduct a medium-term workshop study, providing 13 queer artists with access to two major corporate generative text (GPT-4) and image (DALL-E 3) models. Participants struggle to use these models due to various normative values embedded in their designs, such as hyper-positivity and anti-sexuality, but develop strategies to work around these limitations. Then, I perform an interview study with a group of 15 queer artists to see how they are perceiving and responding to the proliferation of GenAI. I detail how my interviewees are individually and collectively resisting the development and deployment of GenAI models that misalign with their aesthetic and moral values.
My prior work demonstrates the power of queer artists to re-appropriate, resist, and refuse GenAI. However, their agency is still limited by GenAI developers' design choices, with participants in both my interview and workshop studies critiquing the style of text-to-image models. This begs the question: What do developers aesthetically value in the production of these models? In my proposed work, I aim to explore the aesthetic values of generative image model developers through a critical discourse analysis of trace data (e.g., blogs, technical reports, and developer Discord servers) from GenAI model and data providers. As a case study, I will then examine how aesthetic values are materialized in a prominent open-source dataset used to train text-to-image models: LAION-Aesthetics. Finally, I will complement this work by interviewing those researching and developing generative image models to understand how they make sense of aesthetic quality in training data and model behavior. My proposed work will uncover the implicit utility functions undergirding contemporary generative image model development, opening new opportunities for queer aesthetic critique.
Best,
Jordan