Associate Professor of Computer Science Sorelle Friedler is interested in fairness. Since 2014, she has been researching the potential discriminatory impacts of algorithms and trying to understand how to mitigate them. This work has important implications beyond the field of computer science as algorithms are increasingly used to make life-altering, everyday decisions about hiring, policing, housing, and loans.

Until now, Friedler’s work—done in collaboration with the University of Utah’s Suresh Venkatasubramanian, the University of Arizona’s Carlos Scheidegger, and generations of Haverford students—has focused on machine learning algorithms built on data “one could imagine as stored in a spreadsheet,” said Friedler. But now she and her team have joined with Haverford alum Aaron Clauset ’01 of the University of Colorado to expand their discriminatory algorithm research to social networks. The National Science Foundation (NSF) recently awarded them funding for this work.

“We know that many people learn about and get jobs by networking; a parent's friend mentions the job to you or an alum sends in your resume with a personal note,” said Friedler. “We also know that people who are demographically similar to each other are more likely to have ties to each other in a social network. Does this mean that these social network processes for distributing information about jobs are discriminatory? And how could we make sure that everyone who is qualified has equal access to information about a job? These are the types of questions we'll be exploring for this project.”

The project, which the NSF is supporting with a $128,670 grant, will include the development of novel mathematical and computational models that characterize how differences in position in social networks can amplify inequalities in access, as well as the design of techniques to change the structure of the network to increase the flow of information and reduce the overall inequities.

Clauset, an internationally recognized expert in data science and network science, is “thrilled” to be collaborating with a scientist from his alma mater for the first time. When he attended Haverford as a physics major, he only concentrated in computer science as the College didn’t yet offer a major.

“My part of the project focuses on understanding how the shape of real-world social networks structures the way information can flow across it in unequal ways, so that, for instance, some people might hear about job opportunities earlier than others simply because of where they 'sit' in the social network,” he said. “I'll be helping the team evaluate the theoretical claims about fairness and information on these realistic settings, using data science and numerical techniques. I'll also be leading an experimental section of the project, in which we set up real information flows on a social network in a laboratory setting, in order to get a clearer picture of how information inequalities emerge.”

Clauset isn’t the only Ford involved in this research. Friedler’s preliminary work on social networks and fairness included the senior thesis research of Hannah Beilinson ’20, which explored whether information access could represent an aspect of privilege. (“Hannah showed evidence in her thesis that just by looking at a social network we could find clusters of people with similar external markers of information privilege,” said Friedler.) Nasanbayar Ulzii-Orshikh '22 continued that work in Friedler’s lab this summer, and part of the NSF grant is explicitly to fund continued summer research assistance from Haverford students over the next three years.