Casual inference is the bottom most rug of what gives hard sciences its power. It is that we understand the objects we are manipulating at a much deeper level so we don't sound like idiots.

Suppose that we take:

g = ma

A perfectly valid way to find experimental values for gravity at a location. But that doesn't mean that if we push an object really hard we increase the gravity in that location, or decrease it if we pull on the object. Just because symbol manipulation gives as an answer doesn't mean that the answer makes sense, you need to keep track of all the implicit state of the universe.