Documentation

Studies.AloniVanOrmondt2023

Aloni and van Ormondt (2023): modified numerals and split disjunction #

Superlative modifiers (at least n, at most n) generate ignorance inferences that comparative modifiers (more than n, fewer than n) do not, and the inferences are obviated under universal quantifiers and modals, distribute under universals, license free choice under modals, and vanish under negation — exactly the profile of plain disjunction. Following Büring, superlative modifiers are split disjunctions, at least n ↦ n ∨ more and at most n ↦ n ∨ less, and the whole profile falls out of [Alo22]'s neglect-zero enrichment [·]⁺ once BSML is raised to the first-order QBSML.

The QBSML facts of §5 are the universal theorems of Logic/Team/QBSML/FreeChoice; here the denotations (14) and (16) are the split of the ≥ and ≤ intervals of Degree.Comparison, the results (56)–(61) and (63) are those facts for the paper's three/more predicates in every model, under the paper's side conditions, and the obviation claim (57) is the Fig. 14 countermodel. Its worlds are, as throughout the paper, the sets of atomic facts true at them: w_{Pa} is "the world w in which some object a has property P" (Example 3.1, p. 549). The example rows record the inference profile the analysis answers to.

References #

Superlative modifiers as disjunctions #

theorem AloniVanOrmondt2023.modifier_rows (row : Datum) :
row ∈ Examples.all → ∀ m ∈ row.feature? "modifier", ∀ i ∈ row.feature? "inference", (m = "comparative" → i = "none") ∧ (m = "superlative" → row.feature? "embedding" = none → i = "ignorance")

Superlative rows carry the ignorance inference in unembedded position and comparative rows never do (the contrast of (2)–(7)).

The predicates #

three and more are the paper's numeral predicates; they instantiate the P and Q of the §5 facts.

Instances For
    @[instance_reducible]
    Equations
    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      @[instance_reducible]
      Equations
      • One or more equations did not get rendered due to their size.
      Instances For
        @[instance_reducible]
        Equations
        def AloniVanOrmondt2023.instReprVar.repr :
        Var → ℕ → Std.Format
        Equations
        • One or more equations did not get rendered due to their size.
        Instances For
          @[instance_reducible]
          Equations
          • One or more equations did not get rendered due to their size.
          theorem AloniVanOrmondt2023.classicality_univ {W : Type u_1} {Domain : Type u_2} {Const : Type u_3} [DecidableEq W] [DecidableEq Domain] [Fintype Domain] {M : QBSML.Model W Domain Const Predicate} {s : Finset (QBSML.Index W Var Domain)} {v : QBSML.Index W Var Domain → Var → Domain} (hv : ∀ i ∈ s, ∀ (y : Var), i.assign y = some (v i y)) :
          QBSML.support M (QBSML.Formula.univ Var.x (three.disj more)) s ↔ ∀ i ∈ s, (FirstOrder.Language.Formula.all₁ Var.x ((QBSML.predSymb Predicate.three).formula₁ (FirstOrder.Language.var Var.x) ⊔ (QBSML.predSymb Predicate.more).formula₁ (FirstOrder.Language.var Var.x))).RealizeAt M.interp i.world (v i)

          By Proposition 4.1, a state supports the NE-free ∀x(three(x) ∨ more(x)) iff its first-order translation, computed by rfl, is true at every index.

          The results (56)–(61) #

          theorem AloniVanOrmondt2023.ignorance {W : Type u_1} {Domain : Type u_2} {Const : Type u_3} [DecidableEq W] [DecidableEq Domain] [Fintype Domain] {M : QBSML.Model W Domain Const Predicate} {s : Finset (QBSML.Index W Var Domain)} {c : Const} (hSB : Team.IsStateBased M.access (QBSML.State.worldProj s)) (h : QBSML.support M (threeOrMore c).enrich s) :

          Ignorance (56), Fact 3, is [three ∨ more]⁺ ⊨ ◇three ∧ ◇more on a state-based accessibility relation, the paper's epistemic reading.

          theorem AloniVanOrmondt2023.distribution {W : Type u_1} {Domain : Type u_2} {Const : Type u_3} [DecidableEq W] [DecidableEq Domain] [Fintype Domain] {M : QBSML.Model W Domain Const Predicate} {i : QBSML.Index W Var Domain} (h : QBSML.support M (QBSML.Formula.univ Var.x (three.disj more)).enrich {i}) :

          At a state of maximal information, a single index, [∀x(three(x) ∨ more(x))]⁺ supports ∃x three(x) ∧ ∃x more(x), the distribution of (51) (Fact 5).

          theorem AloniVanOrmondt2023.distributionEpi {W : Type u_1} {Domain : Type u_2} {Const : Type u_3} [DecidableEq W] [DecidableEq Domain] [Fintype Domain] {M : QBSML.Model W Domain Const Predicate} {s : Finset (QBSML.Index W Var Domain)} (hSB : Team.IsStateBased M.access (QBSML.State.worldProj s)) (h : QBSML.support M (QBSML.Formula.univ Var.x (three.disj more)).enrich s) :

          Distribution under partial information (58), Fact 6, yields the modalized conclusion ∃x◇three(x) ∧ ∃x◇more(x) on a state-based accessibility relation.

          theorem AloniVanOrmondt2023.boxFreeChoice {W : Type u_1} {Domain : Type u_2} {Const : Type u_3} [DecidableEq W] [DecidableEq Domain] [Fintype Domain] {M : QBSML.Model W Domain Const Predicate} {s : Finset (QBSML.Index W Var Domain)} {c : Const} (h : QBSML.support M (threeOrMore c).nec.enrich s) :

          Free choice under necessity (59), Fact 7, is [□(three ∨ more)]⁺ ⊨ ◇three ∧ ◇more.

          theorem AloniVanOrmondt2023.diamondFreeChoice {W : Type u_1} {Domain : Type u_2} {Const : Type u_3} [DecidableEq W] [DecidableEq Domain] [Fintype Domain] {M : QBSML.Model W Domain Const Predicate} {s : Finset (QBSML.Index W Var Domain)} {c : Const} (h : QBSML.support M (threeOrMore c).poss.enrich s) :

          Free choice under possibility (60), Fact 8, is [◇(three ∨ more)]⁺ ⊨ ◇three ∧ ◇more.

          theorem AloniVanOrmondt2023.universalFreeChoice {W : Type u_1} {Domain : Type u_2} {Const : Type u_3} [DecidableEq W] [DecidableEq Domain] [Fintype Domain] {M : QBSML.Model W Domain Const Predicate} {s : Finset (QBSML.Index W Var Domain)} (h : QBSML.support M (QBSML.Formula.univ Var.x (three.disj more).poss).enrich s) :

          Universal free choice (63), Fact 9, is [∀x◇(three(x) ∨ more(x))]⁺ ⊨ ∀x◇three(x) ∧ ∀x◇more(x), the inference [Che09b] attests.

          theorem AloniVanOrmondt2023.negation {W : Type u_1} {Domain : Type u_2} {Const : Type u_3} [DecidableEq W] [DecidableEq Domain] [Fintype Domain] {M : QBSML.Model W Domain Const Predicate} {s : Finset (QBSML.Index W Var Domain)} {c : Const} (h : QBSML.support M (threeOrMore c).neg.enrich s) :

          Under negation the enrichment is inert, [¬(three ∨ more)]⁺ ⊨ ¬three ∧ ¬more ((61), Fact 10), so the simpler fewer than three blocks (61a).

          Obviation: the Fig. 14 countermodel #

          A single index at the world w_{PaQb} with the empty assignment; that world alone sees itself. The domain is the paper's two objects, and a world is the set of atomic facts true at it.

          Entity is the Fig. 14 domain of two objects, each its own individual constant.

          Instances For
            @[instance_reducible]
            Equations
            Equations
            • One or more equations did not get rendered due to their size.
            Instances For
              @[instance_reducible]
              Equations
              • One or more equations did not get rendered due to their size.

              In the Fig. 14 model only w_{PaQb} has an arrow, to itself.

              Equations
              • One or more equations did not get rendered due to their size.
              Instances For

                The accessibility is state-based on the Fig. 14 state, so obviation is not an artefact of dropping the frame condition behind ignorance.

                As Fig. 15 shows, the universal extension splits into the x/a index supporting [three(x)]⁺ and the x/b index supporting [more(x)]⁺.

                As Fig. 16 shows, at the x/b index the only accessible world is w_{PaQb}, where three holds of a alone, so ◇three(x) fails.

                The universal quantifier obviates ignorance, [∀x(three(x) ∨ more(x))]⁺ ⊭ ∀x(◇three(x) ∧ ◇more(x)) ((57), Fact 4).