Journalist-in-Residence Talk: The Limits of Fairness
summary
Argues that fairness metrics and better models often miss the real harms: algorithms reshape norms (example: Gmail autocomplete nudges acceptable replies), simplify human complexity into brittle probabilities, and can crowd out plural, value-based judgment. Calls out overselling of precise gains (the 24.8% crime reduction claim) as misleading, highlights the legibility versus black‑box catch‑22, and warns that widespread algorithmic judgment pushes institutions toward local optima and erodes professional craftsmanship and democratic accountability.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:08Critique of algorithmic fairness
- 10:16Oversimplified social science models
- 21:07Incentives to oversell results
- 30:00Structural effects of algorithmic judgment
- 43:10Interdisciplinary gaps in fairness
- 55:09Public discourse on fairness
- 1:07:20Journalism shaping the debate
- 1:09:50Closing remarks
related talks
50mThe Measure and Mismeasure of Fairness
42mFairness in Machine Learning: From Theory to Practice
37mKeynote: Ethical Machine Learning: Creating Fair Models in an Unjust World
21mFairness and Accountability Design Needs for Algorithmic Support in High-Stakes Public Sector ...
40mQuestions and Practices to Make Algorithmic Decision Making More Fair
38mSocial Implications of Bias in Machine Learning
54mKeynote - Fairness and Diversity in Online Social Systems
16mExistential Engineering or How Ethics Precedes Error Handling