Machine learning: Economics and computer science converge
Today’s digital economy is blurring the boundaries between computer science and economics — in Silicon Valley, on Wall Street, and increasingly on university campuses.
Yale undergraduates interested in both fields can pursue the Computer Science and Economics (CSEC) interdepartmental degree program, which launched in fall 2019, with coursework covering topics such as machine learning and computational finance.
Philipp Strack, CSEC’s inaugural director of undergraduate studies, is comfortable straddling multiple disciplines. With an academic background in economics and mathematics, his research reflects this broad and interdisciplinary outlook — ranging from behavioral economics and neuroscience to auction design, market design, optimization, and pure probability theory.
Strack, an associate professor of economics in the Faculty of Arts and Sciences, recently spoke to YaleNews about the real-world implications of this work, what the CSEC program offers students, and how it bridges these critical fields.
What is your research focus?
Phillipp Strack: I’m quite unfocused — I work on all types of things, from auctions to models of learning. But most of my work falls into institutional or mechanism design and behavioral economics. I find it fascinating to apply quantitative methods — or just theoretical reasoning — to social problems. The design perspective on institutions is very interesting, and I enjoy teaching it. How can you find institutions within a given set of constraints that maximizes welfare for a given group or society? It’s a very intriguing engineering question.
What are some of the real-world implications of your work?
Strack: Recently I’ve been researching how to explain prejudice in society. In a working paper with Paul Heidhues and Botond Köszegi, we explore the idea that overestimating yourself can generate prejudice. Suppose I think I’m great, even though I’m not; when I look at my outcomes in life, I am unhappy and think I deserve more. To explain why I’m not receiving what I think I deserve, I may believe there is discrimination against me. This leads me to overestimate discrimination against the groups I’m a member of, while underestimating discrimination against groups with which I compete.
This can create prejudice and specific patterns in people’s beliefs. For example, if you ask Black and white people which group is discriminated against, there are huge gaps between the perception of different groups. The same is true for men and women. But standard theories of discrimination — that people discriminate because they don’t like certain groups, or believe them to be less qualified — don’t predict these large differences in beliefs about discrimination across groups.
What drew you to Yale and how has your experience been?
Strack: The economics department is incredibly intellectually vibrant. I interact most with the theory group, which is large and very good. Most other universities have much less focus on theory. It’s just an amazing place — there are very few places like it.
What is student-faculty collaboration like?
Strack: There are a lot of institutionalized opportunities for interaction between students and faculty. The theory group runs a reading group, a lunch, a seminar, a breakfast where students present, and a lot of other activities. In general, it’s a very supportive and collaborative environment; the faculty really cares about engaging with students and discussing their work.
What gets students most excited about the Computer Science and Economics (CSEC) major?
Strack: For many students, the benchmark was double majoring in economics and computer science — so CSEC frees up space and gives students more flexibility.
Also, I think a lot of students really want to learn the newest methods. It’s incredibly exciting, for instance, that we can teach a computer to recognize a cat or a dog in a picture. That’s almost magical! I think the combination of learning new quantitative methods while applying them to social issues is a very attractive combination for students. Machine learning, for example, is a very popular course that a lot of CSEC students take as an elective.
By the way, the major was very much established when I arrived in 2019. All the credit for designing CSEC goes to Joan Feigenbaum and Dirk Bergemann. Their work set us up for success.
When will the first CSEC majors graduate?
Strack: Actually, a student graduated last year as Yale’s first CSEC major: Musab Javed. He wrote the first CSEC thesis, as well, focused on online learning algorithms, which predict answers to a sequence of questions using knowledge of past correct answers, a huge area in computer science. Applying these algorithms to crypto and stock markets, he explored whether they could predict future market behavior and volatility.
What types of careers are CSEC majors most interested in?
Strack: Many students are interested in tech companies and quantitative finance firms, like hedge fund-type companies. These types of work align relatively well with what they learn in the major. Students have taken summer internships at tech companies — a lot of exciting research in CSEC areas happens at large companies like Amazon or Google. A lot of machine learning methods, for instance, were developed inside these companies for very concrete applications.
Do CSEC majors engage with debates over the potential negative consequences of Big Tech?
Strack: It’s a huge issue. In the academic communities, there’s a lot of awareness and research on these topics. In economics, there’s debate over whether or how tech companies should be regulated. In computer science, there’s debate about whether machine learning methods pick up, replicate, and potentially amplify biases already present in society. There’s no explicit CSEC class on these issues, which are very multifaceted, but there are Yale working groups focusing on them.
How does CSEC balance the two fields, and has your involvement with the major impacted your own research agenda?
Strack: There is a real intersection of computer scientists and economists working on the same questions. For example, game theory and auction design — two big ideas in economics — now have many computer scientists working on them, for the simple reason that they became important to Google AdWords and the digital advertising sector. Recently, Jacob Leshno and I worked on the question of what Bitcoin and other decentralized proof-of-work protocols can and can not achieve, which is a question mainly studied in computer science.
These areas are converging academically, as well. Some of my research was already at the intersection: Mechanism design, an area of economics related to game theory and the design of institutional incentives, is an area in computer science as well. From time to time, I attend computer science conferences, and I once wrote a research article with computer scientists Hu Fu, Nicole Immorlica, and Brendan Lucier that was published. I don’t think there’s such a clear divide anymore — computer scientists are picking up tools from economics, and vice versa.
Greg Larson is a freelance consultant and writer focused on economics, public policy, and social impact.