Keynote: Digital Echoes: Understanding Mass Violence with Data and Statistics
summary
Argues that raw statistical patterns reflect how data were collected not reality, and shows concrete fixes and failures: using daily situation reports and an estimated crude mortality of 0.62 per 100 per day to demonstrate extraordinarily high prison deaths and win conviction evidence; using machine learning on county features to predict hidden graves in Mexico and prioritize searches (green-dot counties proved accurate and drove advocacy); reimplementing a predictive‑policing algorithm to show feedback loops that amplify racialized overpolicing rather than reduce crime.
AI-written summary · about
chapters (11)
- 0:00Intro
- 3:38Prisoners of War
- 4:18Mortality
- 6:21Hidden Graves
- 11:20Predictive Policing
- 12:55Drug Crimes
- 14:25Bias
- 15:38Feedback Loop
- 18:07Why are ML models wrong
- 19:24Cost of being wrong
- 20:32Edgar Garcia case
related talks
01h 04mData Hemorrhage, Inequality, and You...
31mEthics in technology: History, practice, and mitigation
26mSridhar Telidevara: Professor, Great Lakes Institute of Management at Cypher 2017 - Part 1/2
13mKeynote Speech
28mAlgorithmic Bias: Data as a Weapon with Ayodele Odubela
55mKeynote - Big Data's Ethical Drought
47mEthical Machine Learning
24mData science, apps and civic responsibility