Eli Ben-Michael is an assistant professor in Dietrich College of Humanities and Social Sciences’ Department of Statistics & Data Science and Heinz College of Information Systems and Public Policy. His research focuses on developing statistical and computational methods that can be used to solve practical issues, such as how to reduce crime, improve education, or create more equitable healthcare systems.
Tell me about your scholarly work.
My research focuses on developing statistical methods for the social sciences. Often this involves questions about cause and effect. For instance, how do changes to policing in a city affect crime, or how does new technology in schools affect student learning? Understanding causal impacts is particularly important for evidence-informed public policy: we want to be able to tell whether policies are having their intended effect. Disentangling correlation and causation is a difficult problem because we need to isolate the attribute that we're interested in studying and make sure that we're making comparisons between individuals, cities, etc. that are otherwise similar. My work uses tools from optimization and machine learning to design procedures that guarantee that we're making proper comparisons. Recently, I've been particularly interested in how modern AI tools can help us to answer more complex causal questions, such as understanding the effects of text, images and video on people's behavior, and how we can use causal reasoning to understand the impacts that AI tools have in practice.
How is your scholarly work adding to the greater field?
The old saw is that statisticians get to play in everyone's backyard, meaning that we are scientific generalists who can work with just about anybody. My work is rooted in collaborations with researchers studying criminal justice, economics, demography and public health, where statistical challenges often come up. Typically, many other researchers are running into similar challenges, and so I develop statistical methods to help solve them, both for my collaborators and for the broader community. An important part of that is making sure that the methods I develop are easy to use and come with open-source software.
How did you become interested in this topic?
I actually started out as a physics major in college, but after a summer spent in a lab, I found out I was more interested in the spreadsheets than in the lasers. At the same time, I have always been interested in the social sciences. When I got to grad school, I saw that studying causal inference would let me combine my two interests in a way that could also have a broader impact by helping to answer some of our biggest public policy questions.
What are you most excited to accomplish as a faculty member at CMU?
One of my favorite things about CMU is just how interdisciplinary it is. We have many different people from all sorts of backgrounds all working together to solve big problems. I've had the chance to work with and learn from some amazing people here, and I'm excited to continue interdisciplinary, collaborative work at CMU.
What are your goals for the next generation of scholars?
We live in a time where people are inundated with information, some of it dodgy. This is especially true for data analysis and statistics, where many of our debates involve people supporting their arguments with quantitative analyses. My goal is for those coming up to feel confident that they have the skillset to identify high quality analyses and to do them themselves.