Which is fine, as long as your assumptions are good models of reality. When your assumptions are essentially articles of faith your models tend to be about as good as the guys who interpret patterns in chicken bones.
You need a model to design an experiment.
This is how all science is done. You hypothesize. Then experiment.