Big data and machine learning in sociology

The dawn of the digital age, aptly characterized by “computers everywhere” (Salganik, 2018, p. 3), has shaped modern societies and, thus, the lives of individuals worldwide in unique ways. The ubiquity of the internet, in conjunction with the mass distribution of a variety of affordable internet-enabled digital devices, has created new possibilities for collecting, storing, linking, sharing, and exchanging information. Also, the massive progress in computer performance regarding processing capacities and computational speed has paved the way for advances in programming which culminated in the recent progress in artificial intelligence (AI) research, referred to as the recent AI spring (for a brief outline of the history of AI research, see, e.g., Mitchell, 2019). Its results are—among others—the deep-learning-induced successes in speech and object recognition that enable processes as complex as simultaneous translation or autonomous driving. The societal consequences range from the emergence of new professions, business fields, leisure activities, behavioral cultures, and associated lifestyles to new social inequalities (digital divide), dependencies (digital and data literacy gaining relevance as key competencies), and forms of deviant/criminal activity (e.g., cyberbullying and -crime, online hate speech, crimes organized/executed through the internet). This digital revolution affects the social sciences in various ways. First, social processes experience fundamental change and adaption that require extensive scientific elaboration. Second, the steadily increasing application of digital technologies generates an enormous mass of finely granulated data in various forms and formats. It is not just that enormous amounts of data can now be easily accessed and analyzed. Digital innovations have allowed the collection of data in various formats that were previously difficult to compile (e.g., georeferenced data, tracking or process data, intensive longitudinal data, social media text data; Golder and Macy, 2014; Leitgöb and Wolbring, 2021). This digitization and datafication of society have shaped empirical social science research fundamentally in recent years and will continue to do so. Third, the increasing computational power and the maturation of software environments have promoted the development of algorithmic solutions for complex statistical problems. It paved the way for the nascent field of computational social science (CSS; e.g., Lazer et al., 2009, 2020; Edelmann et al., 2020; Engel et al., 2022a,b) at the intersection of the social sciences, statistics, informatics, and mathematics.