Sociological Foundations of Computational Social Science

This chapter delves into the intersection of computational social science (CSS) and sociology, illustrating how digital data analysis and agent-based modeling enhance sociological research. I highlight the strengths of CSS, like its ability to handle large-scale, continuous, and nonreactive data, which traditional methods cannot. This capability is exemplified in Chang et al. (2021), where mobile data analysis provided new insights into social inequality and infection rates. However, I also point out CSS's limitations, particularly its focus on behavior rather than the underlying meaning and beliefs. Thus, I emphasize that to bridge the gap between CSS and sociology, it is essential to incorporate actors' beliefs and interpretations into CSS models. This approach aligns with sociological traditions, as seen in Berger and Luckmann's (1966) theory of social construction of reality. I discuss the model by Goldberg and Stein (2018) as an example where agents in an agent-based model are endowed with an associative matrix to interpret and give meaning to behaviors, showcasing a more sociologically nuanced approach to CSS. The chapter also underscores the significance of starting with strong sociological questions and theories in digital data analysis. This approach is exemplified by DiMaggio, Nag, and Blei's (2013) study on newspaper coverage of U.S. government arts funding, where the application of topic modeling to text analysis revealed shifts in media portrayal over time. The chapter concludes that for CSS to substantially contribute to sociology and vice versa, research should begin with sociological theories and concepts. This approach allows CSS methods to rigorously test these theories, potentially leading to novel findings and the development of new theories, thus advancing both sociology and CSS. This productive collaboration could pave the way for new interdisciplinary fields, enriching both domains.