Behind the Research: Elisabeth Paulson – Harvard Business School

Elisabeth Paulson, an assistant professor in the Technology and Operations Management Unit at Harvard Business School, grew up discussing math at the dinner table. Now, she uses algorithms and machine learning tools to make a difference in people’s lives. We talked with Elisabeth about her research, her academic path, and why she loves reality television shows.

What’s your area of research?
I study how to efficiently and fairly design and allocate resources, especially limited resources in public sector or nonprofit organizations, with the goal of improving social well-being. I primarily work in two areas. One is refugee and asylum seeker resettlement, where the limited resources are the availability of resettlement locations in different host countries. The other is allocating food subsidies to increase fruit and vegetable consumption among low-income communities. Two very different application areas, but actually kind of similar when you get down to the nuts and bolts of the math, the machine learning models, and the algorithms that underlie both of these problems.
Tell us more about that—how do you go about this work?
For the first one, I work on designing new machine learning and algorithmic tools to help resettlement organizations find the best spot to resettle new families when they arrive in a host country. I collaborate with the Stanford Immigration Policy Lab, and we have partnerships with various resettlement organizations in Europe, the US—although that program is kind of shutting down right now—and in Canada. We work with our partner agencies to understand the specific context in a country and then come up with tailored algorithms and machine learning models that those organizations can use to help with the process.
For increasing fruit and vegetable consumption, I work on designing financial incentives and thinking through how they can be combined with other programs like nutrition education to try to improve the availability of healthy food and improve diets, especially in food deserts or low-income communities.
What’s an example of the refugee work?
We have an algorithmic tool called GeoMatch that takes historical administrative data from refugee or asylum seeker resettlement programs. We use that to train machine learning models to try to understand long-term outcomes and who is more likely to thrive in different locations. We use those predictions in an algorithm to recommend specific locations when a new family arrives. The main idea is that we can both speed up the process that agencies go through when making these placements—which has always been a very long manual process—and find better matches between people and places.
What’s an example of the nutrition work?
One project is with the Massachusetts Department of Transitional Assistance, which runs the state’s Supplemental Nutrition Assistance Program (SNAP) and a supplemental program that provides rebates when SNAP dollars are spent on fruits and vegetables at certain vendors. It’s designed to increase fruit and vegetable consumption among low-income households while also supporting local farmers since the vendors are typically farmers’ markets, CSAs, and farm stands.
The challenge is determining where to expand the program with a fixed budget while ensuring equal access and utilization across different groups. It’s very costly to the government because they’re essentially giving out produce with a 100 percent rebate. The first step was assessing the impact of adding vendors in specific neighborhoods—how much would a new farmer’s market in Quincy, for example, increase SNAP households’ fruit and vegetable consumption? We recently completed a paper on that first step with faculty and a PhD student at MIT.
How did you get interested in these applications of machine learning and algorithms?
I’ve always been interested in using math and data science to solve problems and help make decisions. I knew that to stay motivated long-term, I wanted to do something that had the potential to help in some way—to improve social outcomes and also to help organizations that are doing the work on the ground to try to improve those outcomes.
What are you working on at the moment?
A lot of my current research focuses on improving the resettlement algorithm or the predictions, because our algorithm will only be as good as the predictions underlying it. One complicated issue I’m working on now with a PhD student is that we have very little data for some locations–those that resettle only 20 refugees a year or that are very new and have no data.
There are a lot of technical complexities that make this problem really challenging—both to generate the predictions and then to build an algorithm. It’s a cool example where the forefront of machine learning and online decision-making meets an important social problem. That’s exactly the type of problem I like to work on.
Where did you grow up and what’s your background?
I grew up mostly in State College, Pennsylvania, where Penn State is. My mom is retired now, but she was a statistics professor. I come from a very nerdy family, so I’ve always been interested in math, statistics, and computer science. At the dinner table, my older brother would constantly ask me math logic questions. That was just our family dinner dynamic. I remember my friends in high school being nervous to come over for dinner because they knew we’d be talking about math.
My mom always seemed to really enjoy what she was doing, and she had a lot of flexibility. That opened the door to me seeing academia as an exciting career path. I started doing research when I was in high school and carried that on throughout college. I’ve always really enjoyed being in an academic environment, doing research, and spending months or even years just thinking about an interesting problem.
I went to Penn State for undergrad—so, very close to home—and studied math and statistics there. Once I graduated, I took an industry job just to make sure that I wanted to go to grad school. Within a couple of months, I applied to PhD programs.
I chose to go into a field called operations research, which is using math, statistics, computer science, machine learning—all of these analytical tools—to try to make better decisions in practical settings. I went to grad school at MIT, did my postdoc out at Stanford, and have been here for three years.
What do you like to do in your spare time?

I love to try out new restaurants with my husband. I hate the term “foodie,” but I guess that’s the best term. I also have a dog, so I spend a lot of time with her—playing fetch, taking her on hikes. I also play on an intramural volleyball team once a week. I’m very invested in that and definitely get overly competitive. And I watch a lot of TV—that’s really my main way of relaxing. Right now my favorite shows are Severance, White Lotus, and Shrinking. But honestly, I’m most passionate about reality TV. I love just the brainless aspect of it, they’re great to watch while I’m cooking. I just think it’s fascinating to see people being filmed and watch what they do and the decisions they make. I can’t get enough of it.
Read more about Elisabeth Paulson in Working Knowledge. For updates on HBS faculty research, sign up for Working Knowledge’s weekly e-mail newsletter.
Comments must be on-topic and civil in tone (with no name calling or personal attacks). Any promotional language or urls will be removed immediately. Your comment may be edited for clarity and length.
→  Visit Newsroom

source

About The Author

admin

See author's posts