I think some domains allow themselves to be modeled with enough accuracy that an abstraction allows for the useful observation of data. I would suspect that do

I think some domains allow themselves to be modeled with enough accuracy that an abstraction allows for the useful observation of data. I would suspect that domains that can be usefully modeled with a constrained set of abstractions will more practical to work with given the way scientific knowledge is growing. I hear biologists, neuroscientists talk quite often about the complexity of their domain. If these domains can not be simplified, it may mean that research becomes prohibitively expensive verse other research problems that can be simplified.