It’s “Technically Fair” But You May Not Like It: Algorithmic Curation, Filtering, and Prediction Wrestles With Ethics and Public Opinion

30 Apr 2018
4:10 pm - 5:30 pm

210 South Hall

Event Type

Christian Sandvig
Professor, School of Information and the Department of Communication Studies, University of Michigan

As the filtering and curation of everything has been taken over by computers, “fair algorithms” has become both a legal problem and a rallying cry. Researchers in machine learning are now trying to explicitly incorporate fairness into their conceptualization of the algorithmic systems that curate today’s job applicants, predict recidivism, offer housing, find rides, and filter social media. Commercial platforms that operate these systems have, belatedly and after a series of scandals, started to recognize that fairness is a problem. Yet the fairnesses addressed so far have mostly been limited to an arid definition where “fair” means statistical fairness or compliance with certain US laws. This is kind of fairness is relatively clearly defined and largely uncontroversial. But a technically-legal algorithm that will still be widely perceived as unfair is no solution to algorithmic fairness. This paper argues that these platforms now need to grapple with the more expansive meanings of fairness, even if this entangles computing with the morass of applied ethics, philosophy, and public opinion. To that end, the paper proposes a list of the kinds of fairness that are relevant for people who operate algorithmic platforms that curate, filter, or predict. It also argues that these kinds of fairness are already present in other “technical” engineering work although they have been resisted by software engineering.

This is a part of the Algorithmic Fairness Lectures.

Christian Sandvig is Professor in both the School of Information and the Department of Communication Studies at the University of Michigan. He specializes in the design of Internet infrastructure and social computing. His current work focuses on the implications of algorithmic systems that curate and organize curate culture, especially social media. He has also written about social media, wireless systems, broadband Internet, online video, domain names, and Internet policy.

Before moving to Michigan, Sandvig was a faculty member at the University of Illinois at Urbana-Champaign (where he founded the Center for People & Infrastructures) and Oxford University. Sandvig has also been a visiting scholar at McGill University, the Oxford Internet Institute, the Centre for Socio-Legal Studies at Oxford, Intel Research, Microsoft Research, the Sloan School of Management at MIT, and the Berkman Klein Center for Internet & Society at Harvard. His work has been funded by the US National Science Foundation, the Social Science Research Council of New York, the MacArthur Foundation, the Economic and Social Research Council of the United Kingdom, and the Internet Society. Sandvig’s research has appeared in The Economist, The New York Times, Le Monde, National Public Radio, CBS News, and other media outlets.

This event is sponsored by CSTMS.
Additional sponsorship comes from:  Algorithmic Fairness and Opacity working group (AFOG)


Other Events of Interest