Fellows' Presentation

Finding Fairness

Photo by Kevin Grady/Radcliffe Institute

Part of the 2018–2019 Fellows' Presentation Series

Lecture by Cynthia Dwork RI '19

Free and open to the public.

Cynthia Dwork’s research focuses on applying computer science theory to societal problems. Examples include developing a theory of and algorithmic tools for privacy-preserving data analysis, developing universal techniques for ensuring statistical validity in exploratory data analysis, and defining and ensuring fairness in classification algorithms.

The three problems of privacy-preserving data analysis, resilience to exploratory data analysis, and fairness in classification are deeply related, but placing fairness on a rigorous foundation is exceptionally challenging. This is due, in part, to the existence of many competing—and mutually exclusive—notions and measures of fairness, ensuring that any model of fairness will fail on some reasonable measures. The task is further complicated by a lack of access to “ground truth.”