Multi-level causation generates serious methodological issues that are not always appreciated in contemporary social research. This chapter uses the dynamics of HIV/AIDS to illustrate three such issues. First, the failure to take both individual-level and group- or context-level forces into account leads to systematic bias in the statistical analysis of observed data. We decompose a fully specified cross-level regression model into separate individual- and group-level components to illustrate the resulting biases in data analysis. Second, we look at the application of fully specified cross-level regression models to processes that are not in equilibrium. Static cross-level regression models cannot properly estimate multi-level cause-and-effect when there are non-linear feedback effects among independent and dependent variables over time. Finally, we explore how computational modeling can be used to study these feedback dynamics in multi-level causal processes. We illustrate two computational methods that help researchers unravel such complex causal environments: counterfactuals and process decomposition.