Domain: Computer Vision/Biometrics
Description: Joy Buolamwini (MIT) and Timnit Gebru's Gender Shades project exposed severe intersectional bias in commercial facial recognition systems, demonstrating these systems performed 32 times worse on dark-skinned women compared to light-skinned men.

Ethical Challenges:

• Intersectional Bias: Error rates varied dramatically: 0.8% for light-skinned men vs. up to 34.7% for dark-skinned women
• Data Representation: Training datasets with 83.5% white individuals
• Civil Rights Impact: Use in surveillance and law enforcement applications
• Platform Governance: Led to policy changes at major tech companies

Public Datasets:

• Primary: Pilot Parliaments Benchmark (PPB)
• Source: Algorithmic Justice League
• URL: Contact datasets@ajlunited.org for access
• MIT Media Lab: https://www.media.mit.edu/projects/gender-shades/faq/
• Research Paper: https://www.media.mit.edu/publications/gender-shades-intersectional-accuracy-disparities-in-commercial-gender-classification/
• Content: 1,270 unique individuals from 6 countries (3 African, 3 European)
• Features: Balanced representation, 44.39% women, 47% dark-skin representation
• Labels: Fitzpatrick Skin Type classification and gender annotations