Naïve Bayes versus causal model

Naïve Bayes classifiers are very common machine learning models. Some of the most well known and celebrated applications of Bayesian nets in practice turn out to by just naïve Bayes models such as are spam filters and recommendation systems. This vide explains such models, but also shows (using a simple example in AgenaRisk) why a naïve Bayes net is almost never the correct causal structure for a decision problem.