ABSTRACT
The chapter narrows the scope of computational social science (CSS) to the intersection of data science and social science. It reviews current perspectives on CSS methodology and discusses causal and predictive inference as its main objectives. The chapter highlights causal inference as part of the counterfactual model for observational data analysis, discusses major threats to the validity of inference in CSS, and draws particular attention to validation through proving the replicability of results. The chapter discusses this principle as practiced in data science to extend this training/test sample comparison to the replicability of results across the outcomes of decisions to be made in the conceptualization, modeling, and validation steps of research. The chapter covers a case study of political extremism that uses pooled data from rounds 1 to 9 of the European Social Survey to exemplify the merits of combining causal inference with the generalized idea of replication in data science.
