Pearl’s Causal Hierarchy, Experiments vs Observational Research, and Potential Outcomes vs SCMs
The workshop delves into the logic that underlies causal inference, introducing them to Pearl’s Causal Hierarchy or ‘ladder of causation’ and establishing why it is necessary to make stronger assumptions if researchers want to rise above purely descriptive model fitting. I talk about the three rungs of the ladder—associational, interventional, and counterfactual—and how we can travel up the ladder by making assumptions. In the second part, I compare experiments and observational approaches to causal inference, focusing on a) the pros and cons of each; and b) why we might choose one over the other. The final part provides a high-level overview of two prominent approaches to causal inference found in the literature, potential outcomes and structural causal models. He covers how each approaches the issue of causal ‘identification’, their core assumptions, and concludes with a brief discussion of how they can serve as mutually reinforcing tool sets or paradigms for researchers working on questions of causal import.