309 McNeil Building, 3718 Locust Walk
Risk assessment tools are widely used around the country to inform decision making within the criminal justice system. Recently, considerable attention has been paid to whether such tools may suffer from predictive racial bias, and whether their use may result in racially disparate impact. Evaluating a tool for predictive bias typically entails a comparison of different predictive accuracy metrics across racial groups. Problematically, such evaluations are conducted with respect to target variables that may represent biased measurements of an unobserved outcome of more central interest. For instance, while it would be desirable to predict whether an individual will commit a future crime (reoffend), we only observe proxy outcomes such as rearrest and reconviction. In the first part of this talk I will discuss how this issue of "target variable bias" affects evaluations of predictive bias. In the second part of the talk I will turn to the question of disparate impact in the context of risk-based sentencing. I will discuss how certain hypothetical risk-based sentencing policies might operate to widen existing observed racial disparities in sentence duration.
Alexandra is an Assistant Professor of Statistics and Public Policy at Carnegie Mellon University's Heinz College of Informations Systems and Public Policy.
She received her B.Sc. from the University of Toronto in 2009, and in 2014 she completed her Ph.D. in Statistics at Stanford University. While at Stanford, she also worked at Google and Symantec on developing statistical assessment methods for information retrieval systems.