Keynote: Using Data Science for Social Good: Examples, Opportunities, and Challenges
summary
Shows concrete deployments and numbers: off‑the‑shelf machine learning on police records cuts false positives by about 50% fewer false positives while finding 15–20% more true positives, a lead‑risk model prioritizes prebirth inspections and links into EMRs in pilot, dropout models find the same at‑risk students three years earlier, EPA/city inspection ranking improves inspection efficiency, and a jail‑risk rank puts top 200 -> 100 jailed next year. Code and datasets live on public GitHub for replication.
AI-written summary · about
chapters (13)
- 0:00Data science for social good
- 1:04Police misconduct systems
- 7:05Lead poisoning prevention
- 12:43High school dropout risk
- 17:15Hazardous waste inspections
- 20:07Urban blight prevention
- 22:14Criminal justice strategies
- 26:09Public policy templates
- 30:02Social impact challenges
- 33:32Skills for social good
- 35:16Resources and involvement
- 36:40Q&A interpretability
- 43:40Data collection validation
related talks
34mKeynote: Building a Data Driven Dialogue From Filling Potholes to Disrupting the Cycle
37mKeynote: Ethical Machine Learning: Creating Fair Models in an Unjust World
31mKeynote: Become a Data Superhero How Data Can Change the World
55mEthical Machine Learning
33mData Science in Healthcare: Beyond the Hype
36mKeynote: Ethics in AI: Building Fair and Just Intelligence Systems
47mEthical Machine Learning
31mDoes This Algorithm Have a Soul?