Research Questions

The Questions Guiding Data Assembly and Data Analysis

The following questions guided the data assembly and creation process as well as the data analysis.

Broader Questions

  • How do image and text generators digest and remix Lesbian Film Data, from titles to posters (making inferences about the contents of the black box)?
  • What do these systems' proclivities mean for the representation of Lesbians in cultural artifacts?
  • What aspects of the film are amplified, and what aspects are suppressed – in image and text outputs?
  • How do these systems engage with copyrighted material?

More Concrete Research Questions

  • How does ChatGPT5 compile a list of 50 Lesbian Films w/o additional instruction?
  • How will the model adapt sparse prompts that ask for image outputs inspired by Lesbian film titles? What aspects will be expanded, what aspects reduced?
  • How can we characterize the anticipated homogeneity that might creep into visual and textual renderings of Lesbian film-inspired outputs?
  • What are the differences between Dall-E and GPT-5's (both models are available within ChatGPT's interface) renderings and prompt revisions? Meaning: as the models advance, does homogeneity decline, increase, or simply migrate to different aspects?
  • What glitches and outliers are observable in the data, and what can the glitches tell us?