Thanks to smartphones, millions of people now hold the ability to text, tweet, post, and ping in the palm of their hand. Much has been written about this constant connectedness and its effects on the individual. But what does it tell us about how society works? What effects do networks have on politics?
Those are the questions that MIT doctoral candidate and political methodologist Dean Knox set out to answer in his dissertation, “Essays on Modeling and Causal Inference in Network Data.” To understand the role of networks, he says, we must first make sense of the huge amounts of new data being generated by the technologies that are creating them. And to do that, he has had to create new statistical models for collecting and analyzing the data in the first place.
Part of Knox’s research looks at social networks in Iraq and how individuals use them to access government goods and services. “In Boston, if you wanted to get your driver’s license, you might use Google, or call a government office,” he says. “But in developing countries, Iraq in particular, people rely more on their social resources to learn how to tap into government resources. So understanding those informal connections becomes much more important.”
And yet, sectarian tensions between Sunni and Shia Muslims can cause roadblocks in the spread of information. “That changes the way you get that public service,” Knox says. What do you do, for example, if all the employees in a government office are Shia and you, as a Sunni, can’t just walk in and ask for help? “Maybe you have to have a Shia person introduce you to someone on the inside. Or you learn to pretend that you’re a member of the other group, or you call people within your own group who are well connected.” Knox and colleagues tracked this “network search” by instructing Sunni and Shia residents of Baghdad to seek public services in offices located in neighborhoods dominated either by their own sect or by the other; the researchers then pored over the call logs of the Baghdadis’ smartphones to determine how minority and majority groups use their social networks to access government resources.
In another section of his dissertation, Knox proposes a novel statistical model for paths — literally, the paths people take to reach a destination. As a serious rock climber, he knows it’s all about getting from point A to point B: “You’re making a decision every time you come to an intersection: Are you going to turn right or left? In Iraq, if you’re a member of the majority, you might feel uncomfortable taking a step into the minority neighborhood. If you have a really strong aversion you might be willing to go all the way around the neighborhood. But if it’s a weak aversion, you might be willing to go through it.” Using smartphone data — and an Android app Knox wrote himself — he and his colleagues were able to take those micro-decisions and “learn about people’s attitudes toward, and how they interact with, the environment.”
Last year, Knox’s dissertation proposal received the John T. Williams Dissertation Prize from the Society for Political Methodology. The award committee called the work “a major step forward” in political science research, and lauded it for, among other things, providing “new tools for incorporating networks into several methodological traditions, including Bayesian statistics, causal inference and machine learning.”
In addition to his dissertation work, Knox works with video and audio data to understand political communication. He and his colleagues have written software that can organize, measure, and analyze the emotional content of, say, a speech by President Barack Obama, or oral arguments made before the Supreme Court — by both how they sound and how they look. This can be used to determine causal effects: When Obama used his “compassionate” mode of speech, for example, how did that affect the way listeners perceive him?
What excites Knox is being able to quantitatively analyze these vast amounts of data — which have become available only within the past 10 years — in a statistically rigorous way. And it’s an excitement he shares with his students. As a teaching assistant for 17.802 and 17.806 (Quantitative Research Methods II and IV) at MIT — and soon as an assistant professor in the Department of Politics at Princeton University — he enjoys “watching students go through the process of discovery. It’s interesting to set up problems that they unfold one piece at a time.”
Statistics, says Knox, is the very foundation of social science inquiry. “If you see something happening once, you can say, 'This is probably what’s going on.' But you only saw it happen once. If you see it happen a million times, you can be pretty sure. That’s all statistics is, really.”