Causal graphs and causal inference on top of Bayesian networks: do-calculus criteria (front-door/back-door), causal impact estimation and counterfactual reasoning (module CM, depends on BN).