Dani Rodrik critiques the critique of economics.
He argues for a "many models" approach. This certainly happens in the experimental economics world where there are several models of behaviour, each of which fits one or some experimental situation. The downside of this approach is that you lose the point that these models contradict each other. (Either politicians are making campaign commitments in one dimension, and you get the median voter theorem, or in many dimensions and you don't.) If many models are useful, and choosing the right one is a "craft", then do we end up vulnerable to the 1-800 critique?