Developing an intersectional framework to analyze biases in artificial intelligence and deep neural
summary
Argues that bias comes from missing data, baked-in human prejudice and feature choices, and that adding targeted synthetic data with contextual labeling and metadata teaches models about those causal and social factors. Describes four data sources (public celebrity sets, crowdsourcing, web scraping, in‑house collection), cites real harms (soap dispensers and COMPAS recidivism) and deepfake stats (96% pornographic, 99% female), and shows synthetic augmentation as a cheaper, iterative mitigation that improves recognition for underrepresented groups while raising ethical and governance tradeoffs.
AI-written summary · about
chapters (9)
- 0:00Intro
- 1:28Sources of Bias
- 6:35Bias and Facial
- 9:02Sourcing Data
- 11:50Retracing Bias
- 14:52Conceptual Framework using Synthetic Data and Critical Theories
- 17:13Contextual Labelling
- 18:19Labelling Examples
- 19:50Challenges
related talks
38mSocial Implications of Bias in Machine Learning
40mEthical AI: Addressing Bias & Algorithmic Fairness
47mEthics & AI: Identifying & preventing bias in predictive models
38mSocial Implications of Bias in Machine Learning
27mCrouching Harm, Hidden Bias—Addressing issues of fairness....
21mFairness and bias in AI
41mWhen AI Failures Reflect Our Own Social Biases
56mAn AI with an Agenda: How Our Cognitive Biases Leak Into Machine Learning