Event Description:
There's widespread concern that high-stakes decisions -- made both by humans and by algorithms -- are biased against groups defined by race, gender, and other protected traits. In a series of two talks, Goel will describe several interrelated threads of research that seek to define, detect, and combat bias in human and machine decisions, drawing on new and old ideas from statistics, computer science, law, and economics. In the first talk, Professor Goel will discuss bias in human decisions and demonstrate that the most popular statistical tests for discrimination can, in practice, yield misleading results. To address this issue, he proposes two new methods. The first method is the threshold test, and is designed to circumvent the problem of "infra-marginality". The second method is a risk-adjusted regression that mitigates the problem of "included-variable bias". Professor Goel illustrates these techniques on large-scale datasets of police interactions with the public. In the second talk, he’ll turn then to bias in machine decisions, and similarly show that the most popular measures of algorithmic fairness suffer from deep statistical flaws. Algorithms designed to satisfy those measures can, perversely, harm the very groups they were designed to protect. To demonstrate these ideas, he’ll propose a class of risk-assessment algorithms used by judges nationwide when setting bail.
Event Speaker:
Sharad Goel, Assistant Professor, Department of Management Science & Engineering, Stanford University
Event Information:
Free and open to the public. Reserve your seat on the event webpage. Email the Data Science Institute with any questions. This event is hosted by The Data Science Institute Working Group on Computational Social Science in conjunction with The Columbia Population Research Center.
