Warren Center Affiliate Profile- Duncan Watts

Networks and Collective Dynamics

What happens when groups of people interact and how does the structure of those interactions lead to better or worse collective outcomes? This fundamental question has driven Duncan Watts’ work since he was a PhD student at Cornell. Over the years, he has moved between disciplines: from a Physics undergrad degree to an Engineering PhD to becoming a Sociology professor at Columbia to Yahoo! and Microsoft Research, where he worked mostly with computer scientists, and now to Penn; but through all this change, he’s continued to be interested the same question. “Networks and collective dynamics have been a theme of mine for basically my whole career,” says Watts.

The technological landscape has changed dramatically since Watts’ career began in the mid 1990s. This was before Amazon and Google existed and long before Facebook and Twitter; no one had cell phones, and email was still a novelty. Back then, large-scale social network data was extremely scare, so he mostly relied on simple mathematical models and computer simulations. The revolution in digital communications that has unfolded over the past twenty years has not only transformed society, it has also generated vast troves of new data that are highly relevant to the questions of social science.

As the available data have changed, so have Watts’ research methods: in place of models and simulation, his work is now mostly empirical and experimental. In fact, the whole field of computational social science evolved more or less in response to the growing availability of digital data. So even where the questions he’s asking have stayed the same over the years, the ways in which he goes about answering them are very different than when he started.

Current Projects

In the course of his work on networks and collective dynamics, Watts has added a couple of “big” questions to his agenda: First, how can we collect the kind of data that we need to answer the question of what happens when groups of people interact, and what infrastructure might we need to build in order to get that data? And second, what does it mean to explain social phenomena, and how should evaluate our explanations? To further explore these questions, Watts is currently working on three main research projects.

The first project is directly related to collective social dynamics, but in the specific context of team performance. This problem has attracted a great deal of attention over the past 60 years from researchers in a variety of disciplines, and is also of obvious relevance to many people who run or work in large organizations, most of which rely on teams to build products or manage projects. In collaboration with the new Analytics program at Wharton, Watts is hoping to develop “high throughput” virtual lab experiments to explore the many ways in which different elements of teams (the attributes of team members, team size, the nature of the task, the problem solving environment, etc.) can impact performance.

The second project is related to Watts’ interest in data collection and infrastructure. Motivated by recent concerns about the effects of misinformation and bias in the media ecosystem (encompassing both traditional and social media), Watts and his collaborators have been constructing several large datasets and data processing systems to study the production and consumption of news. In addition to pursuing a number of their own research projects, their hope is to build a shared data infrastructure that will enable many research groups to collaborate, thereby generating a more replicable and more cumulative scientific understanding of misinformation and its societal consequences.

The third project is more abstract, but also of practical importance. Watts explains, “in science we often use the words “explanation” and “understanding” as if (a) we know what they mean, and (b) we all mean the same thing by them.” Over the past decade or so, however, he has come to believe that researchers are often very sloppy in how they use these terms, invoking conceptually distinct meanings (e.g. interpretive sense making vs predictive accuracy vs. causal identification) without being clear which is being referred to and why. “It is my opinion that this unexamined conflation of distinct concepts has created tremendous confusion about the goals of science, and how we evaluate its progress,” says Watts. If true, then clarifying exactly what people mean when they claim to have “explained” or “understood” something is critical to making science—and especially social science—more useful and relevant to society.

Computational social science is, by design and by necessity, highly interdisciplinary, and Watts’ current collaborators include computer scientists, sociologists, economists, psychologists, historians, and even a bioethicist. He’s excited to collaborate with fellow Warren Center faculty affiliates, a group he describes as practically a “who’s who” of people at Penn. “It’s a truly impressive and diverse community of researchers that I’m excited to join,” says Watts.

The Future of Data

Looking to the future, Watts hopes to see a few developments.

First, he hopes for better collaborations between industry and academia around data sharing. Thanks to a series of scandals in recent years, the public’s trust in tech companies has plummeted, and they are under increasing pressure not to protect their users’ data. Academics, meanwhile, have not made a convincing case either to the companies or to the general public that more data sharing for research purposes would be good for anyone other than academics. The result is a system for sharing data that is extremely ad-hoc, nontransparent, and inequitable. “It is bad for science, bad for society, and probably bad even for the tech companies,” says Watts. He thinks that an entirely different approach to data sharing is needed: one that is systematic, transparent, ethical, and geared toward creating public value while respecting individual privacy. “We are currently a long way from such a system but many people are increasingly interested in creating it, so I am hopeful that it can be done.”

Second, Watts hopes the computational social science community designs and builds its own data generating infrastructure. No matter how much data is ultimately received from corporations, it will always be limited in its value to science for the simple reason that it wasn’t collected with scientific questions in mind. The systems that spin off all this digital data were designed to have certain properties that make them work for consumers or for advertisers or for their system administrators, but when a social scientist comes along and tries to repurpose the data to answer some research question it’s almost always problematic in some ways.

An alternative approach is to start with the research question and then design the data collection platform to collect the data of interest. This approach requires a lot more work up front but can pay dividends in the analysis stage. By analogy, many of the biggest breakthroughs in physics and astronomy in recent years have come from new instruments—LIGO, the Large Hadron Collider, the Hubble Telescope–that were designed to solve specific problems (detecting gravity waves, the Higgs Boson, extraterrestrial planets, etc.). A big challenge is that these instruments are also very expensive, and social scientists—unlike physicists—don’t have much of a history of banding together to argue for major investments in shared resources. “That’s something that I think will need to change for computational social science to reach its full potential,” says Watts. “Fortunately, it’s a discussion that an increasing number of us are starting to have, so again I am hopeful that it might happen.”

Watts’ Background

Watts joined Penn in July 2019 as the twenty-third Penn Integrates Knowledge University Professor, with faculty appointments at the Annenberg School, in the Department of Computer and Information Science in the School of Engineering and Applied Science, and in the Department of Operations, Information and Decisions in the Wharton School, where he is the inaugural Rowan Fellow.

Before coming to Penn, Watts was a principal researcher at Microsoft Research (MSR) and a founding member of the MSR-NYC lab. He was also an Andrew D. White Professor-at-Large at Cornell University. Prior to joining MSR in 2012, he was a professor of Sociology at Columbia University, and then a principal research scientist at Yahoo! Research, where he directed the Human Social Dynamics group.

Watts is the author of three books—Everything is Obvious: Once You Know The Answer (Crown Business, 2011); Six Degrees: The Science of a Connected Age (W.W. Norton, 2003); and Small Worlds: The Dynamics of Networks between Order and Randomness (Princeton University Press, 1999)—as well as articles in such leading publications as Nature, Science, the American Journal of Sociology, the Columbia Journalism Review, and the Harvard Business Review.

He has received major awards for his contributions to network science and computational social science, including the Young Scientist Award for Socio- and Econophysics from the German Physical Society, the Lagrange-CRT Foundation Prize in complexity science, and the Everett Rogers Award from the Annenberg School at the University of Southern California. In 2018, he was elected an inaugural fellow of the Network Science Society.

Watts has a two-year old son, so much of his free time is devoted to being a dad. Whenever possible, he and his wife love to go skiing, especially in the west. They just moved from Manhattan to Philadelphia, so he’s looking forward to getting back into some activities—like cycling, squash and swimming—that are much easier to do in Philly. “In short, I’m delighted to be here!” says Watts.