Gordon McKay Professor of Computer Science
Fairness, Randomness, and the Crystal Ball
Prediction algorithms score individuals, or individual instances, assigning to each one a number in the range from 0 to 1. That score is often interpreted as a probability: What are the chances that this loan will be repaid? How likely is this tumor to metastasize? What is the likelihood that this person will commit a violent crime in the next two years? A key question lingers: What is the probability of a non-repeatable event? Without a satisfactory answer, how can we even specify what we want from an ideal algorithm?
In this talk, we will introduce ‘outcome indistinguishability’ — a desideratum with roots in complexity theory. The talk will also situate the concept within the 10-year history of the theory of algorithmic fairness and the four-decade literature on forecasting.