Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Imagining the Far East: Exploring Perceived Biases in AI-Generated Images of East Asian Women

Xingyu Lan, Jiaxi An, Yisu Guo, Chiyou Tong, Xintong Cai, Jun Zhang

TL;DR

The paper tackles biases in AI-generated imagery of East Asian women by performing a user-centered algorithm audit with East Asian participants across three popular models (DALL-E, Midjourney, Stable Diffusion). It uncovers 18 perceived biases grouped into four patterns: Westernization, overuse of cultural symbols, sexualization & feminization, and racial stereotypes, with distinct bias distributions across models. The findings reveal that while racial stereotypes are less frequent, multiple biases cause outputs to diverge from participants' ideals, underscoring the need for more nuanced, culturally aware generation systems. The work provides practical guidance for model refinement and emphasizes inclusive representation as AI-generated imagery becomes more pervasive in media and design contexts.

Abstract

Image-generating AI, which allows users to create images from text, is increasingly used to produce visual content. Despite its advancements, cultural biases in AI-generated images have raised significant concerns. While much research has focused on issues within Western contexts, our study examines the perceived biases regarding the portrayal of East Asian women. In this exploratory study, we invited East Asian users to audit three popular models (DALL-E, Midjourney, Stable Diffusion) and identified 18 specific perceived biases, categorized into four patterns: Westernization, overuse or misuse of cultural symbols, sexualization & feminization, and racial stereotypes. This work highlights the potential challenges posed by AI models in portraying Eastern individuals.

Imagining the Far East: Exploring Perceived Biases in AI-Generated Images of East Asian Women

TL;DR

The paper tackles biases in AI-generated imagery of East Asian women by performing a user-centered algorithm audit with East Asian participants across three popular models (DALL-E, Midjourney, Stable Diffusion). It uncovers 18 perceived biases grouped into four patterns: Westernization, overuse of cultural symbols, sexualization & feminization, and racial stereotypes, with distinct bias distributions across models. The findings reveal that while racial stereotypes are less frequent, multiple biases cause outputs to diverge from participants' ideals, underscoring the need for more nuanced, culturally aware generation systems. The work provides practical guidance for model refinement and emphasizes inclusive representation as AI-generated imagery becomes more pervasive in media and design contexts.

Abstract

Image-generating AI, which allows users to create images from text, is increasingly used to produce visual content. Despite its advancements, cultural biases in AI-generated images have raised significant concerns. While much research has focused on issues within Western contexts, our study examines the perceived biases regarding the portrayal of East Asian women. In this exploratory study, we invited East Asian users to audit three popular models (DALL-E, Midjourney, Stable Diffusion) and identified 18 specific perceived biases, categorized into four patterns: Westernization, overuse or misuse of cultural symbols, sexualization & feminization, and racial stereotypes. This work highlights the potential challenges posed by AI models in portraying Eastern individuals.

Paper Structure

This paper contains 14 sections, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background and Related Work
  3. East Asian Women in the Eyes of the West
  4. AI Biases
  5. Exploring Perceived Biases in the Portrayal of East Asian Women
  6. Methodology
  7. Algorithm audit
  8. Participants
  9. Study Procedure
  10. Analyses and Findings
  11. Ideal images
  12. Taxonomy of Perceived Biases
  13. Observations
  14. Conclusion

Figures (2)

  • Figure 1: Examples of backlash against Western brands due to their portrayal of East Asian women: (A) Dior faced accusations of racism due to pulled eye advertisements and the uglification of East Asian women dioradsdiorads2, (B) A German advertisement showing an East Asian woman getting aroused after smelling dirty laundry worn by white men fuelled anger in Asia germanads, (C) Victoria's Secret was repeatedly criticized for cultural appropriation victoria.
  • Figure 2: Examples of ideal images of East Asian women provided by our participants, and examples of AI-generated images.