Like all social science disciplines, much of political science is devoted to making causal inference. But the quest to make causal inference in the discipline is undermined by dominant and long-standing philosophical traditions, which hold that causation is not a genuinely meaningful concept. Thus, in the tradition of Humean empiricism, causal relationships in the discipline are sometimes treated as unexplained correlations; and when mechanisms are offered to explain correlations, it is often done in the tradition of instrumentalism empiricism: mechanisms are relatively simplistic and treated as fictions that do not possess genuine explanatory power.; In short, political scientists often take a black box approach to causation, emphasizing empirical regularities over theoretical mechanisms. When political scientists do attempt to identify accurate and detailed theoretical mechanisms that reflect the complexity of real world social processes---when the black box is opened and explored---their work is often conducted without regard to correlational evidence. This too is a result of the powerful philosophical traditions that favour regularities over mechanisms: correlational methods are often considered a distinct sphere of activity, separate from, and superior to, non-correlational methods that are geared primarily toward identifying social mechanisms.; This thesis attempts to develop a philosophical basis that complements, rather than clashes with, the desire to make causal inference. Building on scientific realist literature, I develop a unique approach to making causal inference: an approach applicable to political science and, more generally, the social sciences. The social scientific realist (SSR) approach that I develop holds that theoretical mechanisms are indispensable to establishing causation: thinking inside the black box is critical. But my approach also holds that theoretical mechanisms cannot be firmly established in isolation from correlational evidence: to think seriously inside the black box, it is necessary to embrace rather than eschew correlational methods.
展开▼