Competent Men and Warm Women: Gender Stereotypes and Backlash in Image Search Results
summary
Searches run for 68 trait queries plus the baseline "person" across four Anglophone markets collect the top 1000 images per query from Bing and run them through Clarifai with LUKE-based tag parsing to infer gender (95% of images processed). Overall images for "person" show about 42% men and 18% women, with the top 100 far more skewed (~61% men, ~11% women). Traits cluster as intelligent/competent/rational = masculinized and sensitive/warm = feminized, and a backlash penalizing stereotype-incongruent images appears (competent women more than warm men).
AI-written summary · about
chapters (16)
- 0:00Introduction
- 0:34Gender bias in image search
- 2:11Personality traits
- 2:41Automatic analysis
- 3:35Theory
- 4:50The trait adjectives checklist method
- 5:55Experiments
- 6:30Research Questions
- 8:10Pilot Study
- 9:41Clarify
- 10:59Clarification
- 11:41Results
- 12:29Gender Distribution
- 14:40Summary
- 15:39Conclusion
- 16:51Question
related talks
01h 40mAfternoon Session #1: Science, Pseudo-science, Canon and Belief
31mGender Bias in Tech: Examining Evolution & Persistence of Stereotypes
45mThe Myth of Neutrality: How AI is widening social divides
27mThe Limitations and Danger of Facial Recognition
28mEthical AI
30mWhy the Future of AI Is Feminist
38mSocial Implications of Bias in Machine Learning
37mThe AI 'Black Box'