The Feedback Loop of Statistical Discrimination
Talkat Bridging Prediction and Intervention Problems in Social Systems (Simons Institute, 2026) · 2026-01-12..01-18·
summary
Model of repeated hiring where the firm's evaluation noise for a group improves only when that group supplied hires last period, so the initial noise level determines long-run fate: if initial evaluation sits below a threshold for initial evaluation ability the group's hiring rate goes to zero over time, while starting above it yields a positive steady hiring share. Simulations use linear regression updates, show smaller groups need a higher starting accuracy, and drastic short-term interventions beat mild long-term ones.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 0:46Algorithmic impacts study
- 3:19Feedback loops in hiring
- 7:55Dynamic discrimination model
- 12:55Noise variance over time
- 18:01Steady states explained
- 20:33Historical disadvantages
- 21:43Breaking feedback loops
- 24:34Group size and thresholds
- 25:17Model validation
- 27:29Implications for systems
- 27:50Q&A
related talks
42mUnderstanding the nature of Bias in Natural & AI Systems
01h 31mHow To Measure And Mitigate Unfair Bias in Machine Learning Models
30mIndividual Statistical Fairness
01h 19mMorning Session: Causality
15mExploring the math of thresholds
01h 30mFairness Tales: How To Measure And Mitigate Unfair Bias in Machine Learning Models
37mSelection bias: The elephant in the room
47mEthical Machine Learning