Mapping Computational Social Science

Computational social science—the study of human behavior using new digital data and methods from computer science and other STEM fields—is one of the most rapidly growing fields in academia.

Though I've enjoyed being a part of the growth of this field, I recently realized that I know very little about where computational social science is spreading, and how it has evolved over time. This prompted me to launch what I believe may be the first comprehensive citation analysis earlier this year with several wonderful collaborators: Achim Edelman, Danielle Montagne, and Tom Wolff.

We tried to map the field of computational social science by mining the Web of Science, the largest database of academic articles currently available. We obtained all articles from this resource that mention "computational social science" or "big data" (remember that neologism?). This produced more than 10,000 journal articles, but excluded many earlier articles that came out before either of these terms existed. So I went through the top 500 cited articles by hand in order to identify some of the more major articles we might have missed.

The first thing we learned is that computational social science is growing even faster than any of us thought. The figure below describes the number of publications in our database (by field) from 2000-2018. As this figure shows, computational approaches are thriving in Business and Economics, Psychology, and (surprise) Computer Science. Unfortunately, the Web of Science assigns somewhat crude categories to its articles that prevent us from parsing out many fields of interest such as political science, but we were able to pull out sociology, which I was not surprised to see has some catching up to do (More on that below).

Our next goal was to get a sense of which academic fields are citing each other within computational social science. The next plot (below) illustrates the citation network of computational social science using the Louvain Community Detection method (and nodes are sized by in-degree). Overall, the field appears more connected than most, with dense connections between psychology, communications, sociology and geography as well as a fair amount of citation between economics/business, computer science, and information science. At the same time, there appear to be relatively isolated chunks of the field that are primarily comprised of psychologists and sociologists.

So what kind of stuff are people in these fields writing about? We are currently running lots of models that will help us get a better picture of the main substantive areas of research (many of which likely span academic fields). In the meantime, however, we've begun to analyze our own discipline, sociology.

The plot below groups computational social science articles that were classified as "sociology" by the Web of Science in terms of distinctive words employed in their abstracts. Here, we can see clear groupings around social networks, text analysis, life course and the family, social movements and media, economic sociology, medical sociology, comparative-historical sociology, good-ol' fashion inequality, and even niche fields such as food.

We will have much more to say in an article that we are writing that will come out in the Annual Review of Sociology this fall. But for now, we hope you've enjoyed this preliminary effort to map computational social science! And for those of you attending one of the Summer Institutes in Computational Social Science, I look forward to seeing you soon.