Counterfactual Fairness

Explore counterfactual fairness in machine learning through this insightful talk by Dr. Matt Kusner from The Alan Turing Institute. Delve into the ethical implications of AI-driven decision-making in areas like insurance, lending, hiring, and predictive policing. Learn about a novel framework that uses causal inference tools to model fairness, addressing biases against specific subpopulations based on race, gender, or sexual orientation. Discover how counterfactual fairness ensures decisions remain consistent across actual and hypothetical demographic scenarios. Examine a real-world application of this framework in predicting law school success, and gain valuable insights into recent developments in algorithmic fairness, including the design of learning classifiers and construction of unfairness functions.

Introduction
Joint Work
AI Applications
Deep Fundamental Problems
Governments
Algorithmic Unfairness
Outline
Definition
Explanation
Counterfactuals
Recent work
Design
Learning classifiers
Constructing unfairness functions
Monte Carlo approximation
Compass score
Data
Thank you
Questions