[baddox]

The point is that the models and their accompanying scientific descriptions provide explanatory power that is a valid source of knowledge even though it is not empirical (empiricism is a false epistemology, but that's a topic for another time). If you have a bridge design produced according to good models, and another bridge design produced without using any such models or engineering knowledge, the epistemological status of the claims "this bridge design is safe" is different between the two bridge designs. That is true even before each bridge is actually built, and if it weren't true, then there would be no reason to have laws requiring bridges to be built according to certain standards (since "we wouldn't know if it's safe until we build it").