As a language scientist ( a linguist) I know that you are coming from the hard sciences with an argument that I honor for its simplicity, its integrity and its beauty. Unfortunately it struggles in the real world of complex systems. In simple experiments we can shave away all the complementary and supplementary functions that are connected to any simple act or action. But in the real world we are constrained by reality and complexity, because if we remove attendant functions, we lose the reality of the system itself. Things/actions do not exist in simplicity, they grow, change, advance or retreat through complexity and this is where the hard sciences fail to recognize that complexity is the nature of reality and that the empirical function, of reducing to a single simple element or action, does not reflect the nature of the real universe.
Thus we scorn "modeling" because if we make a change in an attendant function we get a different answer even though the simple question ("Are we changing the world through pollution?") has not changed. We get different answers because of the complexity of the system, not because of the inability of the practitioners to define the component functions (or parts) carefully enough.