This chapter explores the integration of machine learning and causal inference, emphasizing how machine learning overcomes the limitations of traditional causal inference, which often suffers from restrictive model assumptions, limited covariate selection, and difficulties in handling non-linearity and heterogeneity. Machine learning expands these methods through flexible modeling, high-dimensional variable selection, and data-driven discovery. The integration of machine learning with propensity score matching, instrumental variables, and quasi-experimental design demonstrates this combination’s capacity to address confounding, reveal heterogeneous causal effects, and refine counterfactual analysis. Within computational social science, this approach complements traditional causal inference, though its use should be determined by the research context. This chapter further provides methodological guidance for choosing appropriate techniques based on data structure and analytical objectives.