- The structural strategy involves looking for mechanisms to explain correlations, building statistical models to represent hypothesized mechanisms, testing models through statistical tests and background knowledge, and evaluating the overall validity of the model to make causal inferences. - The interventionist strategy focuses on invariance under intervention as a definition of causation, but this lacks a clear methodological application to observational data. Weak invariance, stability across subpopulations, provides one way to test causal claims with observational data. - Overall, the best approach is to consider causal modeling as evaluating the validity of a whole explanatory model, not just testing a single condition, to make sense of correlations.