Causal modelling and the logic of science