can you provide a bit more context? Why do you have to answer this?
In the meantime: I never saw a paper that uses them and I personally would never use them. They intend to reward parsimony but at the end of the day, all what counts is the real test of the model implications. What does a parsimony yet misspecified model bring you. In my point of few, these are problematic transfers of statistical / descriptive modeling ideas ("keep models simple and number of parameters small").
And yes, fitting non-parsimonious models provide the illusion of an informative model but that cannot be turned around that parsimony is an goal in itself and can outbalance model fit.