Three Degrees of Representation: How GenAI Filters, Retrieves, and Remixes Lesbian Film Data

Author

• Maria G. Baker, CUNY Graduate CenterFollow

Date of Degree

6-2026

Document Type

Master's Capstone Project

Degree Name

Master of Arts

Program

Digital Humanities

Advisor

Michael Mandiberg

Subject Categories

Digital Humanities | Feminist, Gender, and Sexuality Studies | Film and Media Studies

Keywords

Generative AI bias, Critical AI Studies, Lesbian Cinema, Data Auditing, Queer Media, Cultural Representation

Abstract

This project investigates how generative AI systems assemble, filter, and represent Lesbian film data. Through auditing GenAI outputs across multiple dimensions (text, images, metadata), 3°R explores what gets included, excluded, and transformed in the process.

—

Three Degrees of Representation is a public website (https://mgb513.github.io/lesfilm-meets-AI/) featuring an investigation of AI-generated images and texts in response to Lesbian film data. Filtering simple prompts based on Lesbian film titles through the algorithms of AI image and text generators makes visible how these systems digest and remix existing (training) data, and how aspects of Lesbian representation are either amplified or suppressed and erased in the process. The films at the core of the database themselves are representations of queer women and constitute a stable data set from which AI images can be derived and to which these outputs can then be compared. The proposed systematic engagement with image and text generators (in this case, DALL-E 3 and ChatGPT-5) builds on established methods of algorithm auditing.

An analysis of the AI-generated data reveals the homogenizing drift and the biases inherent in “the coded gaze” (Buolumwini). Furthermore, the featured data comparisons and analyses crystallize the difference between human-generated and AI-generated representations of Lesbians in images and text. An awareness of these differences becomes especially salient at a time when ever more artistic choices inherent in producing a cultural artifact like a film, from scriptwriting to actual video production, are outsourced to or co-created with generative AI. The project serves as a call for critical awareness of and skeptical engagement with generative AI systems because: The representations of our representations will be our new representations.

Comments

Online component: https://mgb513.github.io/lesfilm-meets-AI/

Recommended Citation

Baker, Maria G., "Three Degrees of Representation: How GenAI Filters, Retrieves, and Remixes Lesbian Film Data" (2026). CUNY Academic Works.
https://academicworks.cuny.edu/gc_etds/6762