Michelangelo D’Agostino: A physicist reshapes his career

After months of working 50-hour, then 70-hour, then 80-hour weeks to reelect President Obama, election day was one of the worst days of the campaign, says Michelangelo D’Agostino, who spent a year on the campaign’s digital analytics team. ‘We got into the office at 4:45 that morning, and we had this powerless feeling, because there was almost nothing left that we could do.’ The quiet, anxious day ended in jubilation, of course. And D’Agostino ended an adrenaline-driven year with the satisfaction of having applied his physics knowledge to the real world, perhaps even making a real difference.

Before going to work for the campaign in the fall of 2011, D’Agostino had been studying neutrino physics as a postdoc at Argonne National Laboratory. He earned his PhD at the University of California, Berkeley, in 2009, also in neutrino physics. All along, though, he pursued side interests, such as teaching and writing science news. On the campaign, he used his understanding of statistics to, for example, help attract people to volunteer for or donate to the cause. Having gotten a taste of the real world, he’s decided to stay there.

Physics Today‘s Toni Feder spoke to D’Agostino by phone in the weeks after the election. They discussed his physics background, his campaign experience, and his future directions.

PT: Tell me about your physics background and about some of the other things you have done.

D’AGOSTINO: One thing I should say is that I have always really loved physics but was also always interested in doing other things too. So, as an undergrad physics major [at Harvard University], I took tons of literature classes. When I graduated, I spent a year teaching physics at a boarding school in England.

As an undergrad, and when I first got to Berkeley, I was doing experimental condensed-matter physics—low-temperature stuff. But over one summer, working in the lab, I realized I wanted to work with more people in bigger collaborations. After my first year [in graduate school], I switched to IceCube, a neutrino detector [in Antarctica] that is a mixture of astrophysics and particle physics.

I applied on a whim to do an internship at the Economist in London. I didn’t even tell my adviser because I figured nothing would come of it. But I ended up getting picked, so I spent the second summer of graduate school writing about science for the Economist. That was something I was able to keep up while finishing my PhD.

PT: How did you find topics to write about?

D’AGOSTINO: On my first trip to Antarctica [for the IceCube experiment], I wrote a big, three-page report about Antarctic science. And the second time I went, I wrote a five-day online diary.

Besides that, if I came across something when I was looking at the literature as part of my research, I would ask if they were interested in a story about it. More often than not it worked out, because I learned what they liked.

PT: What drew you to the Obama campaign?

D’AGOSTINO: When I finished my PhD, I started a postdoc at Argonne—still in neutrinos, but more straight particle physics rather than astrophysics. I worked mostly on a reactor neutrino experiment in France called Double Chooz.

We used a lot of machine learning methods that come out of computer science. I realized this was something that was becoming increasingly used in industry, in business, and all sorts of places. So I read a lot of things to see what was going on outside of physics. I randomly came across this job ad for the campaign. It was on a machine learning blog. They were looking for people to be in Chicago for a year, working on the campaign, using all sorts of different large data sets to make the campaign run better.

I have always followed politics, although I wouldn’t say I was a political junky. But I really liked Obama and I thought this could be a cool opportunity to use my physics skills and do something in the outside world. So I applied.

PT: What did you end up doing for the campaign?

D’AGOSTINO: I worked in the campaign headquarters, in Chicago. Probably at its peak the office had maybe 500 people. I did analytics within the digital department. The digital department was responsible for the websites, social media, Twitter, Facebook, Tumblr, all that stuff. The big thing for the digital department was email fundraising.

We did a lot of randomized control experiments, similar to what you would do in social science or pharmaceutical drug testing. So we would have several versions of a webform—with different layouts, or sometimes different numbers of things people had to fill out. Then we would randomly direct traffic to different versions of the webpages, and then analyze the results to see if one version of a webpage or form was better at getting people to actually go through and complete a form compared to a different version.

PT: Did they vary much?

D’AGOSTINO: It’s tremendous. Putting images in versus not putting them in, or how much text you have on a page, or how many things you ask people to input can have a dramatic impact on whether or not people fill out a form. If it’s a form asking people to make a donation, it has a sizeable impact on how much money you get in. Small improvements can translate into large increases in actual dollars.

PT: Where did the ideas come from?

D’AGOSTINO: The campaign had a very experimental approach. It was a pretty collaborative environment. A lot of things we thought would win didn’t win, and a lot of things we thought would lose did really well. So we got to the point where we would rely on data and not on people’s intuition.

PT: Can you give an example of something you thought would work and flopped, or the other way around?

D’AGOSTINO: No. We’re not really supposed to talk too much about specifics.

PT: How long before you knew the results from a given test and could decide what to move to a larger scale?

D’AGOSTINO: Sometimes we could tell within a couple of hours, and sometimes we would run tests for a couple of days. We even did things longitudinally over many weeks. It depended on the size of the effect we were trying to measure.

PT: How did your physics background help?

D’AGOSTINO: Just being really comfortable with data and statistics, and really understanding when a difference in something is statistically significant.

One class of problems used statistical knowledge. Another big class of problems that the analytics people worked on was modeling problems, where you would use historical data to build a model to predict future behavior. This basically uses the same machine learning techniques that we use in particle physics. You have an event, something happens in your detector, and using historical or simulated data, you know what a signal looks like and you know what a background event looks like. You can use a model to score every future event and tell you how likely it is to be a signal. Those same techniques that we use in particle physics translate directly into other modeling problems.

So you can build a model using historical data. You use the data to predict which people are likely to respond to an email that asks them, for example, to volunteer. You can get better use of email resources, so you are not constantly emailing everybody all the time. Instead, you can email people who are very likely to respond.

PT: Did you feel like your efforts made a difference?

D’AGOSTINO: We always knew from the beginning that it would be a close election. So everyone felt like even the small things we were working on could have an impact. If you tweak a form, or help select a better email, and you raise some extra money that pays someone’s salary to go organize in a swing state and go knock on some doors, that can have actual impact.

I do feel that all the hours I put in were worth it, and definitely helped raise money, and get more volunteers. And made some small difference.

PT: Did you meet the president?

D’AGOSTINO: Yes. He came to the office on two occasions. And the day after the election, we were supposed to go in and turn in our laptops and say goodbye to people. I don’t know if it was planned or a total surprise, but he came by the office. He gave this amazing speech. He looked out at this giant room full of people and said how much it meant to him that we had all worked so hard on the campaign. He said that when things got terrible on the campaign, he thought about all the great things the young people on the campaign were going to go out and do, and that was what gave him hope. He was choked up, and it was interesting to see him be so heartfelt and emotional. He stayed for two hours and hugged everybody in the office.

PT: What’s next for you?

D’AGOSTINO: My work on the campaign opened my eyes to a lot of industry applications of data analysis. The first thing was realizing that the physics had tremendous carryover to the campaign. It was both finding out that my skills had real-world applications and, from talking to other people, I realized that there are lots of opportunities.

I’ve accepted a job as a data scientist at a startup called Braintree in Chicago. They allow tech startups to take credit card payments on their website. There are lots of data issues—detecting fraud as it is occurring so it doesn’t cost the merchant, quantifying and understanding risk, providing intelligence back to merchants about when traffic on their websites peaks. I don’t know exactly what I’ll be doing since I haven’t started yet.

PT: Why did you choose that over staying in physics?

D’AGOSTINO: I always have such weird feelings about leaving physics. I don’t know that I would say I feel guilty, but other academic disciplines, like computer science and economics, are more comfortable with people going back and forth or leaving to take a job in industry. In physics, at least in particle physics and astrophysics, there is less of that. I love physics, but these other experiences have been really great. And the time scale suits me better. The campaign was hyper, too fast. But in particle physics it takes a very long time to build and design experiments.

And, in physics, I felt I didn’t have much control over my destiny. If I wanted to stay in physics, I would pretty much have to go wherever I could get a job. When you get married and start thinking about raising a family, it becomes harder to think about moving around like that.