There is this interpretation of Bayesian networks that the parents are the true causes of a node and to predict what happens if you change a variable you need to remove the edges from the parents to that variable from the network. And then I study what you can do with that method