Big data and cumulation in the social sciences

Research using big data has become popular in the social sciences, raising many new questions. This essay focuses on the question of cumulation, and why the kind of cumulation that is characteristic of social data science is more akin to cumulation in the natural sciences. The reasons for this include how research teams are organized and how they compete to exploit certain data sets to improve upon the work of other teams. There are other factors, however, that mitigate against cumulation, including the lack of access to certain datasets and a lack of building on existing findings in the social sciences. Some of these factors pertain to fundamental philosophical issues in social science, including new ideas about the workings of causal explanation. Others relate to the collaboration or absence of collaboration between different disciplines and to the difference between more applied and more academic research. The essay reviews these factors and develops an account of cumulation anchored in a realist philosophy of science and in the practices and tasks of social science research. It concludes with a call for big data research to be more integrated with already ongoing cumulative findings in the social sciences while recognizing that there are several obstacles to such an integration