Does This Algorithm Have a Soul?
summary
Examples and numbers decide: Pro Publica COMPAS finding shows risk scores wrongly label Black defendants as higher risk about twice as often, facial recognition systems hit 99% vs 35% facial recognition accuracy gap for white men versus dark-skinned women, predictive policing creates self‑reinforcing arrest feedback loops, Singapore uses ML for treatment plans, Operation Serenata jamol audits politicians, and companies retrain models (IBM Microsoft). Recommendations are concrete: admit bias, open black boxes, diversify teams, and intentionally adjust training data or models to counteract skew.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:10Machine learning's impact
- 2:22Bias in algorithms
- 4:17Compas algorithm bias
- 7:37Crime prediction feedback loops
- 9:33Facial recognition bias
- 11:28Google image search bias
- 13:42Fixing algorithmic bias
- 14:25Acknowledging bias as experts
- 15:50Spreading awareness beyond tech
- 16:55Transparency in algorithms
- 17:38Changing crime prediction
- 18:22Diverse teams and fairness
- 19:54Algorithms reflect prejudice
- 20:39Do algorithms have souls
- 21:05Q&A
related talks
45mBias in Artificial Intelligence
38mSocial Implications of Bias in Machine Learning
31mAvoiding Digital Bias
47mEthical Machine Learning
01h 00mThe Elephant in your Dataset: Addressing Bias in Machine Learning
29mResponsible AI Challenges and Recommendations
55mEthical Machine Learning
41mMeasures and Mismeasures of Algorithmic Fairness