Data Science and Computational Public Policy at the UofC
Last Fall, I ran across a headline in one of the LinkedIn analytics groups I subscribe to that prompted a double take. It read something like “Data Science at the University of Chicago.” At first I thought it must be some new university astrophysics grant. Or perhaps economics department or B-school funding. Data science per se seemed a bit pedestrian for the UofC.
Turns out I was wrong. The University of Chicago is indeed making moves in the area. The LinkedIn article I read referenced the Eric and Wendy Schmidt Data Science for the Social Good program that “trains data scientists to tackle problems that really matter.” Summer fellowships support “aspiring data scientists to work on data mining, machine learning, big data, and data science projects with social impact. Working closely with governments and nonprofits, fellows take on real-world problems in education, health, energy, transportation, and more. For three months in Chicago they apply their coding and analytics skills, collaborate in a fast-paced atmosphere, and learn from mentors coming from industry and academia.” What would Robert Hutchins say?
There's even more UofC data science. The Urban Center for Computation and Data, established in 2012, has as its mission “to catalyze and pursue an interdisciplinary research agenda in urban sciences aimed at increasing our understanding of cities and our ability to anticipate the effects of rapid global urbanization on natural, built, and socioeconomic systems.”
And Chapin Hall, since its inception in 1985 with mission to improve “the well-being of children and youth, families, and their communities”, has a policy research focus -- “developing and testing new ideas, generating and analyzing information, and examining policies, programs, and practices across a wide range of service systems and organizations.”
I also noticed a new Masters program in Computational Analysis and Public Policy (CAPP), jointly offered by the Computer Science Department and the Harris School of Public Policy. I was immediately intrigued by the concept of a collaborative social science program involving CS.
The mission of Harris “is to empower scholars to seek impartial, policy-relevant knowledge and train leaders to put that knowledge to work for the public good.” And what's uniquely Chicago about the school “is the core, comprised of various disciplines and fields such as economics, political science, statistics, econometrics, political economy, program evaluation, and more (that help) develop the skills needed to be effective leaders in their chosen careers by honing technical and analytic skills.” Not unlike the Chicago undergrad and B-school curricula. Roughly a third of Harris grads take jobs in each of the federal, public and non-profit sectors.
The CAPP program extends the Harris mission to evidence-based policy-making. “As government decision-making becomes more data driven, issues of data use, data sharing, transparency, and accountability have become increasingly important from both a public policy and a technological perspective. Realizing the potential social benefits associated with the ability to collect, share, and analyze massive amounts of government data requires individuals trained in both public policy and computer science.”
I just had to find out more about CAPP than was available online, so I emailed Colm O'Muircheartaigh, professor and former dean of Harris. I wasn't sure what to expect. I've been interviewing students at the UofC for years without ever meeting a professor. O'Muircheartaigh, though, responded immediately and arranged to have me connect with CAPP faculty director Christopher Berry.
On our call, Berry was quick to point out that CAPP is a fully joint program between academic equals. This allayed my fear that the computational side would be soft.
Berry responded to my quip that CAPP seems the prototypical UofC “mathematical proof-based” curricula by suggesting rigorous was a more apt characterization than theoretical. And it's certainly hard to argue that. Of the 18 quarter courses required for the MS, fully nine are quant heavy 4 in CS, 3 in stats and 2 in microeconomics. Beyond the three additional core policy courses, CAPP students can add classes approved by their advisers in methodology and program evaluation along with electives from departments like economics, statistics and political science.
The four course CAPP-required computation sequence dwarfs what's seen in existing analytics masters curricula. My experience hiring recent BA and MS grads for OpenBI is that I'll trade-off math/stat courses for computational ones, finding it easier to teach computation students statistics than teach statistics students to program.
The CAPP stats/analytics core professes in both traditional parametric statistics that probe for “causal” relationships and the machine learning of CS that obsesses on predictive accuracy. Though business generally focuses on the latter, I like the balance of these disparate emphases.
Admission to such a desirable program at a top-five policy school like Harris is obviously highly-selective and CAPP cohorts will be small by design, targeted at around ten new students per year.
CAPP looks to me a solid contender as a masters for training both policy analysts and data scientists. While most graduates will go on to work in policy roles, the breadth of computational and quantitative exposure assures a solid DS getting-started background. With that, I'm pretty sure analytics and DS roles in business will be open to new grads.
I know OpenBI will be interested in CAPP students. Indeed, I look forward to interviewing one or more of the 13 originals in October, 2015.