Let's say that the interpretation of some body of facts A hinges upon the interpretation of some specific x ∈ A. 99.9_% of humans interpret x one way, and 98% of AI interpret it in another mutually exclusive way. Also A is you. You're A. & they're going to simulate you a bunch.
You'll never escape abstraction though. At a certain point it's a matter of preference, and we have models for how humans resolve differences of preference, so the outcomes are predictable.