Closure is not conservative
ProvedTranscendenceTower.Reification.closure_not_conservativeThe closure axiom , asserting that the extension of is exactly the given set , is a closed-world assumption. Unlike reification itself, it is not conservative: it decides -sentences that the base language leaves open. From and one obtains
so every whose code lies outside is refuted, even though the base language asserted nothing about it.
The statement is deliberately given with three parts, because the refutation claim alone would prove nothing: a theory with no models refutes everything vacuously. So it asserts that the theory in question is satisfiable, that it refutes a particular sentence, and that the sentence is nonetheless satisfiable in the base language. Only the conjunction shows that closure has added content.
The negative information supplied, that nothing else is true at this level, is exactly the content unobtainable by definitional extension. Nonmonotonic machinery of this kind is McCarthy's own invention: is a circumscription of .
import Definitions.Def_TranscendenceTowerReification set_option autoImplicit false
namespace TranscendenceTower.Reification
theorem closure_not_conservative :
∃ (S : Set L0) (p : L0),
(∃ M : RStruc, SatT0 M ∧ SatCL M S)
∧ (∀ M : RStruc, SatT0 M → SatCL M S → Form.eval M.val p = false)
∧ (∃ v : Val, Form.eval v p = true) := by sorry
end TranscendenceTower.Reification