Postdoc Machine Learning at Harvard: where econometrics brings you | Interview with Joran Lokkerbol

Christiaan Cakici
08 april 2019

An interview with Joran Lokkerbol

Image result for joran lokkerbol

Tell us a little bit about yourself

My name is Joran Lokkerbol. I live in Utrecht with my wife and two kids. I got my Masters in Econometrics in 2008 and ended up doing research in mental health. I got my PhD in health economics in 2015 and am currently heading the Centre of Economic Evaluation & Machine Learning at the Netherlands Institute of Mental Health and Addiction (Trimbos institute). In my spare time I like to play my piano, listen to Tom Waits, read or run.

Why did you decide to study Econometrics?

It was a bit of a happy coincidence, where I was planning to study Economics, but ran into someone telling me about Econometrics during my Economics Introduction camp. I was convinced this was a better choice, went to university on my first day and switched to Econometrics.

Where did you study Econometrics and did you also study other subjects?

I studied Econometrics at the University of Amsterdam, took some additional courses in Actuarial Sciences and Operations Research to allow myself to postpone choosing between the three, and studied Music for a semester in UNSW, Sydney. Last year, I worked at Harvard and MIT and took courses in Applied Econometrics and Machine Learning while being there.

Where did you work and what kind of work did you do?

After my studies I started working as an Economics consultant at PwC for two years. After a steep learning curve in the first year, with exposure to many different projects, I couldn’t get rid of feeling like a factory worker standing at a conveyer belt, running projects, and decided to leave. This brought me to the Institute of Business and Industrial Statistics (IBIS UvA), which introduced me to doing research and publishing, as well as teaching, which I enjoyed very much. In 2010, I joined the Netherlands Institute of Mental Health and Addiction to work on assessing which policy choices result in the best value for money when it comes to preventing and treating mental disorders.

You decided to go back to school and pursue a PhD. Tell us about your experience.

I pursued my PhD alongside my job as researcher at the Netherlands Institute of Mental Health and Addiction. Pursuing a PhD is not so much ‘going back to school’, but rather ‘working with those who are properly schooled’. And it is more about producing than consuming, which doesn’t allow for the mediocre understanding that you could get away with as a student, but forces you to go to the bottom. Regardless of where your career ends up bringing you, I think this is a useful setting to be in at some point in your life.

What role does research play in your current position?

I was lucky enough to get two personal grants, which gave me close to full freedom on how to spend my time for a period of four years. One of these fellowships brought me to Harvard and MIT for a year, where I was able to work with and learn from the very best. Back in the Netherlands, I use data from clinical practice to find out which treatments work best for which patients in mental healthcare, which is considered to be the holy grail in our field. I developed a machine learning course, which helps professionals in mental healthcare over a period of four months to apply machine learning to improve clinical practice by tackling prediction problems that support their clinicians to better adjust their treatments to the specific patients they are trying to help. As demand for this course proved to extend far beyond mental healthcare, I founded Data Science Institute to provide for this.

What is it exactly that Data Science Institute offers and stands for?

Data Science Institute stands for the pursuit to apply machine learning to improve the value that companies and institutions deliver, from a proper understanding of what machine learning is and what it can and cannot do. I simply got frustrated with the misunderstandings regarding machine learning, the types of problems it can tackle, and how to use its results in a way that doesn’t overreach. There is a great potential to using machine learning to improve business processes, but much of its potential is currently wasted by either unrealistic expectations, or a lack of understanding on how to apply and interpret machine learning. Data Science Institutes aims to tackle this, one project at a time. This calendar year we will end up running about 40 machine learning projects.

Do you also train students who want to master these skills?

We have room for about 3 students a year who want to use their thesis to experience the process of publishing an article and/or want to learn how to do a machine learning project.