Bayesian Additive Regression Trees (BART) seem like a robust way to do causal inference; easily-queried black boxes with built-in uncertainty quantification of causal effects. Anyone using BART in practice?