From Algorithmic Fairness to Outcome Indistinguishability

This talk is part of the UNIVERSITY MESSENGER LECTURE series and the DATA SCIENCE DISTINGUISHED LECTURE series.

Abstract:

The past fifteen years have witnessed the emergence of algorithmic fairness as a new field in theoretical computer science, statistics, and AI.  In the course of developing the definitional and structural foundations for the area, a philosophical question emerged.  Predictive models, also known as scoring functions, assign to each individual, or individual instance, a number between 0 and 1 that is often interpreted as a probability:  "How likely is this tumor to metastasize?" "What is the probability this individual will commit a violent crime in the next two years?"  But what is the probability of a non-repeatable event?  Without an understanding of what an algorithm is supposed to be producing, how can we evaluate the algorithms we build?  Outcome Indistinguishability, a notion with roots in fairness and complexity theory, provides an avenue of attack. 

This talk describes key milestones in the theory of algorithmic fairness and introduces Outcome Indistinguishability.

Bio:  
Cynthia Dwork is the Gordon McKay Professor of Computer Science at the Harvard University John A. Paulson School of Engineering and Applied Sciences and Affiliated Faculty at Harvard Law School. She uses theoretical computer science to place societal problems on a firm mathematical foundation. Her awards and honors include the National Medal of Science, the IEEE Hamming Medal, the RSA Award for Excellence in Mathematics, the Dijkstra, Gödel, and Knuth Prizes, and the ACM Paris Kanellakis Theory and Practice Award. She is a member of the US National Academy of Sciences and the US National Academy of Engineering, and is a Fellow of the American Academy of Arts and Sciences and the American Philosophical Society.