Documentation

Studies.Aloni2022

Free choice from neglect-zero in BSML #

Aloni derives free-choice inferences from a neglect-zero tendency. Her pragmatic enrichment [·]⁺ (BSML.enrich) conjoins NE to every subformula, so an enriched split disjunction needs two non-empty witnesses, and under a possibility modal each witness is a live option. The paper's free-choice facts are proved here for arbitrary NE-free α β, and its figures are checked by decide on its four worlds w_∅, w_a, w_b, w_ab, each the set of atoms true at it, whose running state is {w_a, w_b}.

Main results #

Implementation notes #

Facts 1, 2, 9, 10, 13 and the BSML* half of Fact 14 are in Logic/Team/BSML/Enrichment.lean, Facts 6–8 in Logic/Team/BSML/Negation.lean, and Fact 15 in Logic/Team/BSML/Classical.lean. The first-order extension of §6.2, which Aloni and van Ormondt develop, and the BSML◇ conjecture of §7 beyond its countermodel (63b) are out of scope.

References #

Free-choice facts #

theorem Aloni2022.witnesses_of_enrich_disj {W : Type u_1} [DecidableEq W] {A : Type u_2} {M : ModalLogic.KripkeModel W A} {α β : BSML.Formula A} {t : Finset W} (hα : α.NEFree) (hβ : β.NEFree) (h : BSML.support M (BSML.enrich (α.disj β)) t) :
(∃ s ⊆ t, s.Nonempty ∧ BSML.support M α s) ∧ ∃ s ⊆ t, s.Nonempty ∧ BSML.support M β s

An enriched split disjunction has a non-empty witness subteam for each disjunct, [α ∨ β]⁺ ⊨ (α ∧ NE) ∨ (β ∧ NE).

theorem Aloni2022.modalDisjunction {W : Type u_1} [DecidableEq W] {A : Type u_2} {M : ModalLogic.KripkeModel W A} {α β : BSML.Formula A} {t : Finset W} (hα : α.NEFree) (hβ : β.NEFree) (hSB : Team.IsStateBased M.access t) (h : BSML.support M (BSML.enrich (α.disj β)) t) :

Modal Disjunction (Fact 3) is [α ∨ β]⁺ ⊨ ◇α ∧ ◇β on a state-based R.

theorem Aloni2022.narrowScopeFC {W : Type u_1} [DecidableEq W] {A : Type u_2} {M : ModalLogic.KripkeModel W A} {α β : BSML.Formula A} {t : Finset W} (hα : α.NEFree) (hβ : β.NEFree) (h : BSML.support M (BSML.enrich (α.disj β).poss) t) :

Narrow Scope FC (Fact 4) is [◇(α ∨ β)]⁺ ⊨ ◇α ∧ ◇β.

theorem Aloni2022.narrowScopeFC_dependent {W : Type u_1} [DecidableEq W] {A : Type u_2} {M : ModalLogic.KripkeModel W A} {α β : BSML.Formula A} {t : Finset W} (hα : α.NEFree) (hβ : β.NEFree) (h : BSML.support M (BSML.enrich (α.disj (α.conj β)).poss) t) :
BSML.support M α.poss t ∧ BSML.support M (α.conj β).poss t

Free choice for logically dependent disjuncts is [◇(α ∨ (α ∧ β))]⁺ ⊨ ◇α ∧ ◇(α ∧ β).

theorem Aloni2022.wideScopeFC {W : Type u_1} [DecidableEq W] {A : Type u_2} {M : ModalLogic.KripkeModel W A} {α β : BSML.Formula A} {t : Finset W} (hα : α.NEFree) (hβ : β.NEFree) (hInd : Team.IsIndisputable M.access t) (h : BSML.support M (BSML.enrich (α.poss.disj β.poss)) t) :

Wide Scope FC (Fact 5) is [◇α ∨ ◇β]⁺ ⊨ ◇α ∧ ◇β on an indisputable R.

theorem Aloni2022.dualProhibition {W : Type u_1} [DecidableEq W] {A : Type u_2} {M : ModalLogic.KripkeModel W A} {α β : BSML.Formula A} {t : Finset W} (hα : α.NEFree) (hβ : β.NEFree) (h : BSML.support M (BSML.enrich (α.disj β).poss.neg) t) :

Dual Prohibition (Fact 11) is [¬◇(α ∨ β)]⁺ ⊨ ¬◇α ∧ ¬◇β.

theorem Aloni2022.doubleNegationFC {W : Type u_1} [DecidableEq W] {A : Type u_2} {M : ModalLogic.KripkeModel W A} {α β : BSML.Formula A} {t : Finset W} (hα : α.NEFree) (hβ : β.NEFree) (h : BSML.support M (BSML.enrich (α.disj β).poss.neg.neg) t) :

Double Negation (Fact 12) is [¬¬◇(α ∨ β)]⁺ ⊨ ◇α ∧ ◇β.

theorem Aloni2022.epistemicContradiction {W : Type u_1} [DecidableEq W] {A : Type u_2} {M : ModalLogic.KripkeModel W A} {φ : BSML.Formula A} {t : Finset W} (hSB : Team.IsStateBased M.access t) (h : BSML.support M (φ.poss.conj φ.neg) t) :
t = ∅

Epistemic contradiction (§4.1) says that on a state-based R, ◇φ ∧ ¬φ is supported only by ∅, the sole team supporting the weak contradiction ⊥.

The four-world illustrations #

A world of the paper's figures is the set of atoms true at it: w_a is "a world where only a is true, w_b only b, etc." (fn. 11, p. 5:14), so w_∅, w_a, w_b, w_ab are ∅, {a}, {b}, {a, b} and an atom holds at a world by membership. Each figNx fixes the accessibility arrows drawn in Figure N(x), worlds without arrows seeing ∅.

inductive Aloni2022.Atom :
Instances For
    @[instance_reducible]
    instance Aloni2022.instDecidableEqAtom :
    DecidableEq Atom
    Equations
    @[instance_reducible]
    Equations
    @[reducible, inline]
    abbrev Aloni2022.model (R : Finset Atom → Finset (Finset Atom)) :

    model R is the paper's Kripke model on the four worlds with accessibility R.

    Equations
    Instances For
      def Aloni2022.state :
      Finset (Finset Atom)

      state is the running state {w_a, w_b} of Figures 1, 2(a), 3 and 5.

      Equations
      Instances For
        @[reducible, inline]

        Figures 1–2 draw no arrows, so only atoms and disjunction are evaluated.

        Equations
        Instances For
          @[reducible, inline]

          Figure 3(a) has R[w_a] = R[w_b] = {w_ab, w_∅}.

          Equations
          Instances For
            @[reducible, inline]

            Figure 3(b) has R[w_a] = R[w_b] = {w_a, w_b}.

            Equations
            Instances For
              @[reducible, inline]

              Figure 3(c) has R[w_a] = {w_ab}, R[w_b] = {w_a, w_∅}.

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

                Figure 4(a) has R[w_ab] = {w_a}.

                Equations
                Instances For
                  @[reducible, inline]

                  Figure 4(b) has R[w_ab] = {w_a, w_b}.

                  Equations
                  Instances For
                    @[reducible, inline]

                    Figure 5(a) has R[w_a] = R[w_b] = {w_b}.

                    Equations
                    Instances For
                      @[reducible, inline]

                      Figure 5(b) has R[w_a] = {w_a}, R[w_b] = {w_b}.

                      Equations
                      Instances For

                        Negation and incompatibility #

                        theorem Aloni2022.not_forall_support_neg_of_forall_disjoint :
                        ¬∀ (φ : BSML.Formula Atom) (M : ModalLogic.KripkeModel (Finset Atom) Atom) (s : Finset (Finset Atom)), (∀ (t : Finset (Finset Atom)), BSML.support M φ t → Disjoint s t) → BSML.support M φ.neg s

                        Incompatibility does not define negation, so the converse of Fact 7 fails (p. 5:31). The state {w_b} is disjoint from every team supporting ¬((a ∧ NE) ∨ b) but does not support its negation, since a is no open possibility in {w_b}.

                        Negative free choice (Fact 14) #

                        BSML⁺ validates neither ◇¬(α ∧ β) ⊨ ◇¬α nor ¬□(α ∧ β) ⊨ ¬□α — the paper's (50), "Mary might not speak both Arabic and Bengali" ⇏ "she might not speak Arabic". The countermodel is Figure 5(b)'s frame at the state {w_a}: inside [¬(a ∧ b)]⁺ a zero witness anti-supports a, but no non-empty subteam of R[w_a] = {w_a} anti-supports a. BSML* validates both inferences (BSML.negativeFC_star_poss, BSML.negativeFC_star_nec).

                        The □ form follows from the ◇ form by the duality □φ := ¬◇¬φ.

                        BSML∅ against BSML⁺ (Table 5) #

                        Table 5 sets the NE-free fragment BSML∅, whose consequence is classical (BSML.consequence_iff_classicalConsequence, Fact 15), against BSML⁺: Positive FC holds only in BSML⁺, while Addition α ⊨ α ∨ β and Contraposition hold only in BSML∅. The BSML∅ failure of Positive FC is Figure 4(a); the BSML⁺ failures are refuted on Figure 5(a)'s frame and on the arrow-free model, the failing contrapositive being that of Positive FC itself.

                        theorem Aloni2022.positiveFC_plus {W : Type u_1} [DecidableEq W] {A : Type u_2} {α β : BSML.Formula A} :

                        Positive FC holds in BSML⁺, ◇(α ∨ β) ⊨⁺ ◇α ∧ ◇β.

                        Positive FC fails in BSML∅, since Figure 4(a) supports ◇(a ∨ b) but not ◇b.

                        theorem Aloni2022.addition {W : Type u_1} [DecidableEq W] {A : Type u_2} {α β : BSML.Formula A} (hα : α.NEFree) (hβ : β.NEFree) :

                        Addition holds in BSML∅, α ⊨ α ∨ β for NE-free α β, classically.

                        Addition fails in BSML⁺, [a]⁺ ⊭ [a ∨ b]⁺ at the zero-model {w_a}, where b has no non-empty witness.

                        theorem Aloni2022.contraposition {W : Type u_1} [DecidableEq W] {A : Type u_2} {α β : BSML.Formula A} (hα : α.NEFree) (hβ : β.NEFree) (h : BSML.Consequence α β) :

                        Contraposition holds in BSML∅, so for NE-free α β, α ⊨ β gives ¬β ⊨ ¬α.

                        Contraposition fails in BSML⁺. Positive FC holds, but its contrapositive ¬(◇a ∧ ◇b) ⊨⁺ ¬◇(a ∨ b) fails at {w_a} on Figure 5(a)'s frame, where R[w_a] = {w_b} anti-supports a but not a ∨ b.