Historical perspective in causal structure learning from observational and experimental data

I will briefly introduce graphical causal models, the basic formal representation of causation that is becoming standard in computer science and elsewhere, and then use them to explain how causal discovery depends on the "data collection regime," especially with regards to whether data is collected from passive observation or experimental control. I will then use graphical models and time series to make vivid the serious challenges that still beset causal inference even in "randomized clinical trials," for example blinding and non-compliance.