Yep, definitely needs some clarification there. Humans don't distinguish between the utility for different microscopic states of the world. Nobody cares if air molecule 12445 is shifted 3 microns to the right, since that doesn't have any noticable effects on our experiences. As such, a state (at least for the purposes of that definition of utility) is a macroscopic state. "~X" means, as in logic, "not X". Since we're interested in the negative utility of the floor being clear, in the above case X is "the airplane's floor being clear" and ~X is "the airplane's floor being opaque but otherwise identical to a human observer". In reality, you probably aren't going to get a material that is exactly the same structurally as the clear floor, but that shouldn't stop you from applying the idea in principle. After all, you could probably get reasonably close by spray painting the floor. To steal from Hofstadter, we're interested in the positive utility derived from whatever substrate level changes would result in an inversion of our mind's symbol level understanding of the property or object in question.

How is this defined? If an airplane has a clear floor such that its passengers vomit whenever they look down, removing the floor would put them in an even worse position. We want to remove only the property of transparency, which would involve replacing the clear material with an entirely different opaque material that had all other properties identical. What's troubling to me about the counterfactual is that it doesn't seem to have an objective baseline, a single thing that is ~X, so we are left comparing Y(X) with Y(Z), the utility of thing X instead of thing Z. I'm not sure how valid it is to talk about simply removing properties because the set of higher level properties depends on the arrangement of atoms. It seems like properties are their own thing that can be individually mixed and matched separate from material but they really can't be.