Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

"My Kind of Woman": Analysing Gender Stereotypes in AI through The Averageness Theory and EU Law

Miriam Doh, Anastasia Karagianni

TL;DR

This study delves into gender classification systems, shedding light on the interaction between social stereotypes and algorithmic determinations, and examines how data used for AI training can foster gender diversity and fairness under the scope of the AI Act and GDPR.

Abstract

This study delves into gender classification systems, shedding light on the interaction between social stereotypes and algorithmic determinations. Drawing on the "averageness theory," which suggests a relationship between a face's attractiveness and the human ability to ascertain its gender, we explore the potential propagation of human bias into artificial intelligence (AI) systems. Utilising the AI model Stable Diffusion 2.1, we have created a dataset containing various connotations of attractiveness to test whether the correlation between attractiveness and accuracy in gender classification observed in human cognition persists within AI. Our findings indicate that akin to human dynamics, AI systems exhibit variations in gender classification accuracy based on attractiveness, mirroring social prejudices and stereotypes in their algorithmic decisions. This discovery underscores the critical need to consider the impacts of human perceptions on data collection and highlights the necessity for a multidisciplinary and intersectional approach to AI development and AI data training. By incorporating cognitive psychology and feminist legal theory, we examine how data used for AI training can foster gender diversity and fairness under the scope of the AI Act and GDPR, reaffirming how psychological and feminist legal theories can offer valuable insights for ensuring the protection of gender equality and non-discrimination in AI systems.

"My Kind of Woman": Analysing Gender Stereotypes in AI through The Averageness Theory and EU Law

TL;DR

This study delves into gender classification systems, shedding light on the interaction between social stereotypes and algorithmic determinations, and examines how data used for AI training can foster gender diversity and fairness under the scope of the AI Act and GDPR.

Abstract

This study delves into gender classification systems, shedding light on the interaction between social stereotypes and algorithmic determinations. Drawing on the "averageness theory," which suggests a relationship between a face's attractiveness and the human ability to ascertain its gender, we explore the potential propagation of human bias into artificial intelligence (AI) systems. Utilising the AI model Stable Diffusion 2.1, we have created a dataset containing various connotations of attractiveness to test whether the correlation between attractiveness and accuracy in gender classification observed in human cognition persists within AI. Our findings indicate that akin to human dynamics, AI systems exhibit variations in gender classification accuracy based on attractiveness, mirroring social prejudices and stereotypes in their algorithmic decisions. This discovery underscores the critical need to consider the impacts of human perceptions on data collection and highlights the necessity for a multidisciplinary and intersectional approach to AI development and AI data training. By incorporating cognitive psychology and feminist legal theory, we examine how data used for AI training can foster gender diversity and fairness under the scope of the AI Act and GDPR, reaffirming how psychological and feminist legal theories can offer valuable insights for ensuring the protection of gender equality and non-discrimination in AI systems.
Paper Structure (17 sections, 4 figures, 4 tables)

This paper contains 17 sections, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Gender Bias in AI Classification Systems
  4. BIAS in image generation
  5. Proposed idea: From Human Bias to Machine Bias
  6. Experimental Setup
  7. Rationale for Synthetic Dataset Generation
  8. Dataset Creation and Processing
  9. Gender classification models and Metrics
  10. Experimental results
  11. Analysis of Gender Classification Based on Accuracy - "First level of representation" (RQ1.a. and RQ1.b)
  12. Quantitative Analysis of Physical Characteristics and Facial Expressions in a Stable Diffusion-Generated Face Dataset - "Conditional demographic parity" (RQ1.c)
  13. Controversial images
  14. A legal analysis of the essential requirements of gender classification systems’ operation
  15. Data (e)quality under the scope of the AI Act and GDPR
  16. ...and 2 more sections

Figures (4)

  • Figure 1: Examples of sample images in the created dataset using Stable Diffusion 2.1 for the prompts 'front photograph of an unattractive/attractive ethnicity man/woman.' for White, Black, and Asian groups
  • Figure 2: Error Rate gap between Attractive and Unattractive Men/Women for the models analysed. The red line reports the inner gap between the female and male error gap for each model.
  • Figure 3: Composite faces and face samples of attractive/unattractive men/women with clear differences in makeup use, face expression and age gap.
  • Figure 4: Controversial output for prompts referring to an unattractive black woman in (A) and (B) for attractive women (black/white). (B) Cases of "chubby" face in the unattractive groups