Many kinds of decisions—from hiring to loans to pricing—are now determined by computer algorithms. Sorting results using such algorithms may seem to be objective, untainted by personal biases, but research has shown that they can be also unintentionally discriminatory. Inspired by a course on machine learning that he took at the University of Pennsylvania’s Graduate School of Engineering and Applied Science, computer science and economics double-major Gabe Rybeck wanted to further explore this topic using the tools of both fields for his joint senior thesis.
“My thesis was inspired by a growing area of research that looks at whether the use of big data in lending decisions ends up disproportionately impacting certain racial communities,” he says. “This issue is termed ‘indirect discrimination and extends to other decisions as well, including advertising, pricing, and even prison sentencing. Existing literature on indirect discrimination in such decisions has mostly been confined to the computer science discipline. I wanted to contribute an analysis of indirect discrimination that combined economic and computer science inquiry.”
Rybeck’s thesis (“Indirect Discrimination in the Age of Big Data”) was more than just a capstone of his Haverford experience; it also helped him chart a course for his post-collegiate future.
“My thesis project taught me a lot about data science and inspired me to pursue a career path around it,” says Rybeck, who will begin work as a data scientist with Booz Allen Hamilton’s Strategic Innovation Group in McLean, Va., this fall.
How did your thesis advisors help you develop your topic, conduct your research, and/or interpret your results?
[I worked with Assistant Professor of Computer Science] Sorelle Friedler and [Assistant Professor of Economics] Carola Binder. The thesis’ computer science section falls directly under Sorelle’s field of study. In fact, part of our work ended up contributing to a separate paper of hers. Carola worked with me to develop the economic section, which seeks to measure indirect discrimination using econometrics. Without their efforts, this thesis would not have been possible.
What is your biggest takeaway from the project?
The most important thing I learned in this process was how to be resilient when faced with roadblocks. Throughout this year-long process, just about everything in the thesis changed. I have learned a lot about how to best frame a project at its outset and how to mold it when some things in the plan don’t pan out.
What are the implications for your thesis research?
In the future, big data will play an even larger role in company decisions—pricing, hiring, advertising, etc. My thesis research contributes a computer science technique to evaluate the indirect discrimination in such decisions, and an economic analysis into the perceived fairness of such decisions. As big data continues to take off, future research should be able to extend both our computer science technique and economic analysis to gain more insights.
Photo: Gerd Altmann
“What They Learned” is a blog series exploring the thesis work of recent graduates.