Coleman Harris

Building on an interest in mathematics while in high school, Kristian Lum went into college set on a math degree. However, what she expected was a connection between the problems and reality—something she didn’t find in higher mathematics. It wasn’t until she took a course in statistics that she really discovered her passion, because she could use quantitative skills to tackle real-world problems.

In her last year at Rice University, Lum participated in a Research Experience for Undergraduates (REU) with one of her professors, who encouraged her to pursue a PhD. That push was what she needed to begin a PhD in statistical sciences at Duke University, again working on a real-world problem: Bayesian spatial quantile regression.

Lum pursued a short postdoctoral position in Rio de Janiero, Brazil, after her doctoral work and before taking some time off. She then joined the Virginia Bioinformatics Institute at Virginia Tech, where her research focused on microsimulations and computational epidemiology, social contagion, and disaster planning.

Lum left Virginia Tech to join the startup DataPad, founded by Python pandas creator Wes McKinney, where her work centered on customer-facing data science. After DataPad, she spent time as a consultant in the data science space and developed a part-time relationship with the Human Rights Data Analysis Group (HRDAG). However, from 2015 to early 2020, Lum was the lead statistician in a full-time role with HRDAG, where she led the group’s project on criminal justice in the United States. Her body of work includes research on algorithmic fairness and predictive policing, collectively using a variety of methods to understand the criminal justice system.

Lum joined the University of Pennsylvania CIS Department in March 2020 as a research assistant professor, including ties to UPenn’s Algowatch Initiative and the Warren Center for Network and Data Science.

Across myriad positions, Lum has explored the topics of justice, fairness, and transparency from a statistical perspective. This humanitarian approach to data science addresses the needs of many marginalized populations and embodies the ideal of Data for Good. You can find Lum on Twitter @KLdivergence.

Q&A with Kristian Lum

Why did you pursue a PhD?

To be honest, I knew I was good at school and it was more school. It wasn’t a specific plan, like someday I want a PhD in statistics or I want to be a professor. I wish I could say there was more of a master plan behind it, but it was just the natural next thing for me to take on.

I actually recommend for people to get real-world context by taking some time between undergraduate and graduate school. I think it’s important and helps to motivate the problems you can work on in grad school. People I know who did this ended up with a more rewarding graduate career, because of that added direction and experience.

What inspired you to pursue a career in the Data for Good space?

About halfway through graduate school, I wanted to make sure the topics I was pursuing were tied to reality. My dissertation was on Bayesian spatial quantile regression, which is very tied to reality, but I wanted to do something with more impact.

I cold emailed Patrick Ball, a founder of HRDAG and now a longtime mentor, with some ways I could help out. To my surprise, he responded that the ideas were great and invited me on a data collection trip to Colombia. That’s what kicked off my interest, another one of those small things that ended up having a big impact. And ever since then, I’ve stayed in the Data for Good space with HRDAG, either part-time or as a full-time statistician, doing stats through the lens of human rights.

What recommendations do you have to researchers interested in working in this area?

Find what you’re passionate about and exactly what good in the world you want to do. Try to connect with organizations that are already doing that good, because they know more about the issues you care about, and see if there are any projects to get involved with. Essentially, find a way to become involved in the areas you’re interested in, including support for the groups already working toward those goals.

What does ‘Data for Good’ mean to you?

I think the term “Data for Good” is too broad, because good is in the eye of the beholder. For example, the criminal justice area includes different ideas about what good is, depending on the perspective. I don’t think there’s this one “good,” rather it’s dependent on the perspective.

Usually, when I describe myself, I describe the areas I work in rather than Data for Good. I work in fairness, particularly in criminal justice, or I mention my work on casualty estimation (the number of people killed in various conflicts around the world). And those could be described as Data for Good, but I personally tend to describe the topic area I work in instead.

What is the most defining experience that has shaped your career?

My first summer working with HRDAG, I was working on casualty estimate—estimating the number of people who have been killed in Colombia. Until that point, I wanted to do something with impact, but, operationally, the work wasn’t that different from other statistics projects I’ve done. When we went to Colombia, I remember we stopped to speak with someone doing data collection on the ground. This woman worked at a small, somewhat rural cemetery and took names of the deceased and registered bodies that turned up without names.

It was an impactful experience to see where the data actually came from. I’m glad we didn’t see a body that day, since people often dropped them off there. The experience did, however, make the data feel so much more real and my projects more real. And our work couldn’t tell you who those people were, but at least it sort of told us the magnitude of the problem and the conflict there.

What advice do you have for people who might want to take a career path similar to yours?

Follow your interests. If there’s something you’re excited about, go with it, because you’re going to have more motivation to work on something that makes you want to throw yourself in it. I know it is a privilege to be able to follow your interests. But following them to the extent that you can defines how I’ve moved through my career so far.

If you do want to have a winding path type of career, you can’t really be worried about accolades in career progress. When you do multiple career switches, you kind of start over each time. Sometimes, it’s frustrating when you see peers get promoted or receive awards, but that happens a bit less when you’re jumping from thing to thing.

What makes a good statistician?

Someone who genuinely cares about learning about the real world. And a willingness to really get dirty with the data, to interrogate where the data comes from and what biases could have generated it.