Increasingly, algorithms rule our world. They guide doctors toward our medical treatments, advise bankers on whether to give us a home loan, help judges decide whether to release us on bail. They’re often hidden and mysterious, guiding our lives in ways we don’t understand. Are they doing a good job? In particular, are they fair, or are they treating some groups of people better than others?
In March, SFI brought together experts from a range of disciplines, including computer science, law, philosophy, and the social sciences to discuss the following question: Can algorithms bend the arc toward justice? The workshop was organized by SFI External Professors Melanie Moses (University of New Mexico) and Tina Eliassi-Rad (Northeastern University) and UNM Law Professor Sonia Gipson Rankin, who are all members of the SFI & UNM Interdisciplinary Working Group on Algorithmic Justice. The workshop, part of a series on the foundations of natural and artificial intelligence, is part of an NSF Artificial Intelligence Research Institute planning grant led by SFI’s Davis Professor of Complexity Melanie Mitchell and Melanie Moses.
To begin, the group analyzed the notion of justice itself, which tends to be understood very differently by computer scientists, ethicists, and lawyers. Computer scientists tend to have a narrow but precisely defined view of fairness — a view that is useful for writing or analyzing algorithms, but often too utilitarian to capture what social scientists, philosophers, lawyers, and everyday people mean by “justice.” One challenge is to find practical ways of deepening algorithmic justice to incorporate broader definitions.
Working group participants also discussed the regulations or incentives needed to ensure that algorithms are working in our best interest; they worked to build a comprehensive theory that would lead algorithms to be more robust and adaptive; and they looked for ways that algorithms could be designed with feedback loops that would break down existing biases rather than reinforce them.
“What I think is unique about SFI,” Moses says, “is the ability to bring together scholars from different disciplines to have a productive discussion. We learn from one another and chart a path forward where artificial intelligence advances justice rather than exacerbates or accelerates injustices.”
Read more about the March 30–April 1 working group Can Algorithms Bend the Arc Toward Justice?