Ethics in technology: History, practice, and mitigation
summary
Argues that supposedly neutral systems often amplify biased, incomplete data and gives a concrete case: PredPol/Getica repurposes the ETAS algorithm on police call-and-observation records, predicts hot squares, and is right under 0.4% accuracy in Wired’s 24,000-prediction audit. Shows how predictions feed back into patrol data to create self-reinforcing feedback loops, over‑surveilling communities of color. Recommends concrete mitigations: fine‑tuning, smaller domain models, safety prompts, red‑team testing, and surfacing grounded sources and knowledge bases.
AI-written summary · about
chapters (10)
- 0:00Introduction and speaker intro
- 2:13AI excitement and concerns
- 4:43Predictive policing case study
- 8:58Problems with predictive policing
- 12:55Historical surveillance patterns
- 13:16Self-reinforcing feedback loops
- 17:09Good intentions bad outcomes
- 18:35Mitigation strategies overview
- 23:51Meta prompts and safety
- 27:03Technology neutrality and responsibility
related talks
16mExistential Engineering or How Ethics Precedes Error Handling
47mEthical Machine Learning
47mAI Ethics, Strategic Decisioning and Game Theory
31mEthics of algorithms - from theoretical discussion to practical application
45mBias in Artificial Intelligence
55mEthical Machine Learning
38mSocial Implications of Bias in Machine Learning
01h 01mDoing Data Science with Ethics