Uniform Interpolation Part 3: Case Studies George Metcalfe - - PowerPoint PPT Presentation

uniform interpolation
SMART_READER_LITE
LIVE PREVIEW

Uniform Interpolation Part 3: Case Studies George Metcalfe - - PowerPoint PPT Presentation

Uniform Interpolation Part 3: Case Studies George Metcalfe Mathematical Institute University of Bern BLAST 2018, University of Denver, 6-10 August 2018 George Metcalfe (University of Bern) Uniform Interpolation August 2018 1 / 27 This Talk


slide-1
SLIDE 1

Uniform Interpolation

Part 3: Case Studies George Metcalfe

Mathematical Institute University of Bern BLAST 2018, University of Denver, 6-10 August 2018

George Metcalfe (University of Bern) Uniform Interpolation August 2018 1 / 27

slide-2
SLIDE 2

This Talk

  • Yesterday. . .

we described a general algebraic framework for (uniform) interpolation in varieties of algebras and connections with properties such as amalgamation, coherence, and existence of a model completion.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 2 / 27

slide-3
SLIDE 3

This Talk

  • Yesterday. . .

we described a general algebraic framework for (uniform) interpolation in varieties of algebras and connections with properties such as amalgamation, coherence, and existence of a model completion.

  • Today. . .

we consider case studies and general criteria for uniform interpolation, focussing first on varieties of algebras for modal logics.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 2 / 27

slide-4
SLIDE 4

Modal Logics

Modal logics are used to reason about modal notions such as necessity, knowledge, obligation, and proof; they correspond to expressive but computationally tractable fragments of first-order logic.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 3 / 27

slide-5
SLIDE 5

Modal Logics

Modal logics are used to reason about modal notions such as necessity, knowledge, obligation, and proof; they correspond to expressive but computationally tractable fragments of first-order logic. Description logics are multi-modal logics for reasoning about concept descriptions built from atomic concepts and roles such as Woman ⊓ ∀child.Woman “women having only daughters”

George Metcalfe (University of Bern) Uniform Interpolation August 2018 3 / 27

slide-6
SLIDE 6

Modal Logics

Modal logics are used to reason about modal notions such as necessity, knowledge, obligation, and proof; they correspond to expressive but computationally tractable fragments of first-order logic. Description logics are multi-modal logics for reasoning about concept descriptions built from atomic concepts and roles such as Woman ⊓ ∀child.Woman “women having only daughters” However, we consider here only the basic language of classical logic extended with a unary connective , defining also ♦α := ¬¬α.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 3 / 27

slide-7
SLIDE 7

Modal Logics

Modal logics are used to reason about modal notions such as necessity, knowledge, obligation, and proof; they correspond to expressive but computationally tractable fragments of first-order logic. Description logics are multi-modal logics for reasoning about concept descriptions built from atomic concepts and roles such as Woman ⊓ ∀child.Woman “women having only daughters” However, we consider here only the basic language of classical logic extended with a unary connective , defining also ♦α := ¬¬α. Modal logics may be presented syntactically via axiom systems, sequent calculi, etc., and semantically via Kripke models, modal algebras, etc.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 3 / 27

slide-8
SLIDE 8

Frames and Models

A Kripke frame W , R is an ordered pair consisting of a non-empty set

  • f worlds W and a binary accessibility relation R ⊆ W × W .

George Metcalfe (University of Bern) Uniform Interpolation August 2018 4 / 27

slide-9
SLIDE 9

Frames and Models

A Kripke frame W , R is an ordered pair consisting of a non-empty set

  • f worlds W and a binary accessibility relation R ⊆ W × W .

A Kripke model M = W , R, | = consists of a Kripke frame W , R together with a binary relation | = between worlds and formulas satisfying

George Metcalfe (University of Bern) Uniform Interpolation August 2018 4 / 27

slide-10
SLIDE 10

Frames and Models

A Kripke frame W , R is an ordered pair consisting of a non-empty set

  • f worlds W and a binary accessibility relation R ⊆ W × W .

A Kripke model M = W , R, | = consists of a Kripke frame W , R together with a binary relation | = between worlds and formulas satisfying w | = α ∧ β if and only if w | = α and w | = β

George Metcalfe (University of Bern) Uniform Interpolation August 2018 4 / 27

slide-11
SLIDE 11

Frames and Models

A Kripke frame W , R is an ordered pair consisting of a non-empty set

  • f worlds W and a binary accessibility relation R ⊆ W × W .

A Kripke model M = W , R, | = consists of a Kripke frame W , R together with a binary relation | = between worlds and formulas satisfying w | = α ∧ β if and only if w | = α and w | = β w | = α ∨ β if and only if w | = α or w | = β

George Metcalfe (University of Bern) Uniform Interpolation August 2018 4 / 27

slide-12
SLIDE 12

Frames and Models

A Kripke frame W , R is an ordered pair consisting of a non-empty set

  • f worlds W and a binary accessibility relation R ⊆ W × W .

A Kripke model M = W , R, | = consists of a Kripke frame W , R together with a binary relation | = between worlds and formulas satisfying w | = α ∧ β if and only if w | = α and w | = β w | = α ∨ β if and only if w | = α or w | = β w | = ¬α if and only if w | = α

George Metcalfe (University of Bern) Uniform Interpolation August 2018 4 / 27

slide-13
SLIDE 13

Frames and Models

A Kripke frame W , R is an ordered pair consisting of a non-empty set

  • f worlds W and a binary accessibility relation R ⊆ W × W .

A Kripke model M = W , R, | = consists of a Kripke frame W , R together with a binary relation | = between worlds and formulas satisfying w | = α ∧ β if and only if w | = α and w | = β w | = α ∨ β if and only if w | = α or w | = β w | = ¬α if and only if w | = α w | = α if and only if v | = α for all v ∈ W such that Rwv.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 4 / 27

slide-14
SLIDE 14

Frames and Models

A Kripke frame W , R is an ordered pair consisting of a non-empty set

  • f worlds W and a binary accessibility relation R ⊆ W × W .

A Kripke model M = W , R, | = consists of a Kripke frame W , R together with a binary relation | = between worlds and formulas satisfying w | = α ∧ β if and only if w | = α and w | = β w | = α ∨ β if and only if w | = α or w | = β w | = ¬α if and only if w | = α w | = α if and only if v | = α for all v ∈ W such that Rwv. A formula α is valid in M, written M | = α, if w | = α for all w ∈ W .

George Metcalfe (University of Bern) Uniform Interpolation August 2018 4 / 27

slide-15
SLIDE 15

Normal Modal Logics

The basic modal logic K can be defined by extending any axiomatization of classical propositional logic with the axiom schema (K) (α → β) → (α → β) and the necessitation rule: from α, infer α.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 5 / 27

slide-16
SLIDE 16

Normal Modal Logics

The basic modal logic K can be defined by extending any axiomatization of classical propositional logic with the axiom schema (K) (α → β) → (α → β) and the necessitation rule: from α, infer α. A normal modal logic is any axiomatic extension of K;

George Metcalfe (University of Bern) Uniform Interpolation August 2018 5 / 27

slide-17
SLIDE 17

Normal Modal Logics

The basic modal logic K can be defined by extending any axiomatization of classical propositional logic with the axiom schema (K) (α → β) → (α → β) and the necessitation rule: from α, infer α. A normal modal logic is any axiomatic extension of K; in particular, K4 = K + α → α

George Metcalfe (University of Bern) Uniform Interpolation August 2018 5 / 27

slide-18
SLIDE 18

Normal Modal Logics

The basic modal logic K can be defined by extending any axiomatization of classical propositional logic with the axiom schema (K) (α → β) → (α → β) and the necessitation rule: from α, infer α. A normal modal logic is any axiomatic extension of K; in particular, K4 = K + α → α KT = K + α → α

George Metcalfe (University of Bern) Uniform Interpolation August 2018 5 / 27

slide-19
SLIDE 19

Normal Modal Logics

The basic modal logic K can be defined by extending any axiomatization of classical propositional logic with the axiom schema (K) (α → β) → (α → β) and the necessitation rule: from α, infer α. A normal modal logic is any axiomatic extension of K; in particular, K4 = K + α → α KT = K + α → α S4 = K4 + α → α

George Metcalfe (University of Bern) Uniform Interpolation August 2018 5 / 27

slide-20
SLIDE 20

Normal Modal Logics

The basic modal logic K can be defined by extending any axiomatization of classical propositional logic with the axiom schema (K) (α → β) → (α → β) and the necessitation rule: from α, infer α. A normal modal logic is any axiomatic extension of K; in particular, K4 = K + α → α KT = K + α → α S4 = K4 + α → α GL = K4 + (α → α) → α

George Metcalfe (University of Bern) Uniform Interpolation August 2018 5 / 27

slide-21
SLIDE 21

Normal Modal Logics

The basic modal logic K can be defined by extending any axiomatization of classical propositional logic with the axiom schema (K) (α → β) → (α → β) and the necessitation rule: from α, infer α. A normal modal logic is any axiomatic extension of K; in particular, K4 = K + α → α KT = K + α → α S4 = K4 + α → α GL = K4 + (α → α) → α S5 = S4 + ♦α → ♦α.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 5 / 27

slide-22
SLIDE 22

Completeness

A normal modal logic L is said to be complete with respect to a class of frames C

George Metcalfe (University of Bern) Uniform Interpolation August 2018 6 / 27

slide-23
SLIDE 23

Completeness

A normal modal logic L is said to be complete with respect to a class of frames C if for any formula α, ⊢L α ⇐ ⇒ M | = α for every model M based on a frame in C.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 6 / 27

slide-24
SLIDE 24

Completeness

A normal modal logic L is said to be complete with respect to a class of frames C if for any formula α, ⊢L α ⇐ ⇒ M | = α for every model M based on a frame in C. The following normal modal logics are complete with respect to the given class of frames: Logic Frames K all frames

George Metcalfe (University of Bern) Uniform Interpolation August 2018 6 / 27

slide-25
SLIDE 25

Completeness

A normal modal logic L is said to be complete with respect to a class of frames C if for any formula α, ⊢L α ⇐ ⇒ M | = α for every model M based on a frame in C. The following normal modal logics are complete with respect to the given class of frames: Logic Frames K all frames K4 transitive frames

George Metcalfe (University of Bern) Uniform Interpolation August 2018 6 / 27

slide-26
SLIDE 26

Completeness

A normal modal logic L is said to be complete with respect to a class of frames C if for any formula α, ⊢L α ⇐ ⇒ M | = α for every model M based on a frame in C. The following normal modal logics are complete with respect to the given class of frames: Logic Frames K all frames K4 transitive frames KT reflexive frames

George Metcalfe (University of Bern) Uniform Interpolation August 2018 6 / 27

slide-27
SLIDE 27

Completeness

A normal modal logic L is said to be complete with respect to a class of frames C if for any formula α, ⊢L α ⇐ ⇒ M | = α for every model M based on a frame in C. The following normal modal logics are complete with respect to the given class of frames: Logic Frames K all frames K4 transitive frames KT reflexive frames S4 preorders

George Metcalfe (University of Bern) Uniform Interpolation August 2018 6 / 27

slide-28
SLIDE 28

Completeness

A normal modal logic L is said to be complete with respect to a class of frames C if for any formula α, ⊢L α ⇐ ⇒ M | = α for every model M based on a frame in C. The following normal modal logics are complete with respect to the given class of frames: Logic Frames K all frames K4 transitive frames KT reflexive frames S4 preorders GL transitive and conversely well-founded frames

George Metcalfe (University of Bern) Uniform Interpolation August 2018 6 / 27

slide-29
SLIDE 29

Completeness

A normal modal logic L is said to be complete with respect to a class of frames C if for any formula α, ⊢L α ⇐ ⇒ M | = α for every model M based on a frame in C. The following normal modal logics are complete with respect to the given class of frames: Logic Frames K all frames K4 transitive frames KT reflexive frames S4 preorders GL transitive and conversely well-founded frames S5 equivalence relations

George Metcalfe (University of Bern) Uniform Interpolation August 2018 6 / 27

slide-30
SLIDE 30

Completeness

A normal modal logic L is said to be complete with respect to a class of frames C if for any formula α, ⊢L α ⇐ ⇒ M | = α for every model M based on a frame in C. The following normal modal logics are complete with respect to the given class of frames: Logic Frames K all frames K4 transitive frames KT reflexive frames S4 preorders GL transitive and conversely well-founded frames S5 equivalence relations Moreover, all these normal modal logics have the finite model property.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 6 / 27

slide-31
SLIDE 31

Modal Algebras

A modal algebra consists of a Boolean algebra supplemented with a unary

  • peration satisfying

(x ∧ y) ≈ x ∧ y and ⊤ ≈ ⊤.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 7 / 27

slide-32
SLIDE 32

Modal Algebras

A modal algebra consists of a Boolean algebra supplemented with a unary

  • peration satisfying

(x ∧ y) ≈ x ∧ y and ⊤ ≈ ⊤. We let K denote the variety of all modal algebras.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 7 / 27

slide-33
SLIDE 33

Modal Algebras

A modal algebra consists of a Boolean algebra supplemented with a unary

  • peration satisfying

(x ∧ y) ≈ x ∧ y and ⊤ ≈ ⊤. We let K denote the variety of all modal algebras. In particular, each Kripke frame W , R yields a complex modal algebra P(W ), ∩, ∪, c, ∅, W , where A := {w ∈ W | Rwv for all v ∈ A}.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 7 / 27

slide-34
SLIDE 34

Equivalence

Theorem

For any normal modal logic L, let VL = {A ∈ K | ⊢L α = ⇒ A | = α ≈ ⊤}.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 8 / 27

slide-35
SLIDE 35

Equivalence

Theorem

For any normal modal logic L, let VL = {A ∈ K | ⊢L α = ⇒ A | = α ≈ ⊤}. Then VL is an equivalent algebraic semantics for L

George Metcalfe (University of Bern) Uniform Interpolation August 2018 8 / 27

slide-36
SLIDE 36

Equivalence

Theorem

For any normal modal logic L, let VL = {A ∈ K | ⊢L α = ⇒ A | = α ≈ ⊤}. Then VL is an equivalent algebraic semantics for L with transformers τ(α) = α ≈ ⊤ and ρ(α ≈ β) = (α → β) ∧ (β → α).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 8 / 27

slide-37
SLIDE 37

Equivalence

Theorem

For any normal modal logic L, let VL = {A ∈ K | ⊢L α = ⇒ A | = α ≈ ⊤}. Then VL is an equivalent algebraic semantics for L with transformers τ(α) = α ≈ ⊤ and ρ(α ≈ β) = (α → β) ∧ (β → α). That is, for any set of formulas T ∪ {α, β} and set of equations Σ, (a) T ⊢L α ⇐ ⇒ τ[T] | =VL τ(α); (b) Σ | =VL α ≈ β ⇐ ⇒ ρ[T] ⊢L ρ(α ≈ β); (c) α ⊢L ρ(τ(α)) and ρ(τ(α)) ⊢L α; (d) α ≈ β | =VL τ(ρ(α ≈ β)) and τ(ρ(α ≈ β)) | =VL α ≈ β.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 8 / 27

slide-38
SLIDE 38

Interpolation in Modal Logic

A normal modal logic L admits deductive interpolation

George Metcalfe (University of Bern) Uniform Interpolation August 2018 9 / 27

slide-39
SLIDE 39

Interpolation in Modal Logic

A normal modal logic L admits deductive interpolation α(x, y) ⊢L β(y, z) = ⇒ α ⊢L γ and γ ⊢L β for some γ(y)

George Metcalfe (University of Bern) Uniform Interpolation August 2018 9 / 27

slide-40
SLIDE 40

Interpolation in Modal Logic

A normal modal logic L admits deductive interpolation α(x, y) ⊢L β(y, z) = ⇒ α ⊢L γ and γ ⊢L β for some γ(y) if and only if VL admits the amalgamation property.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 9 / 27

slide-41
SLIDE 41

Interpolation in Modal Logic

A normal modal logic L admits deductive interpolation α(x, y) ⊢L β(y, z) = ⇒ α ⊢L γ and γ ⊢L β for some γ(y) if and only if VL admits the amalgamation property. For example, K, K4, S4, GL, and somewhere between 43 and 49 axiomatic extensions of S4 admit deductive interpolation, but not S5.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 9 / 27

slide-42
SLIDE 42

Interpolation in Modal Logic

A normal modal logic L admits deductive interpolation α(x, y) ⊢L β(y, z) = ⇒ α ⊢L γ and γ ⊢L β for some γ(y) if and only if VL admits the amalgamation property. For example, K, K4, S4, GL, and somewhere between 43 and 49 axiomatic extensions of S4 admit deductive interpolation, but not S5. Note, however, that L admits Craig interpolation ⊢L α(x, y) → β(y, z) = ⇒ ⊢L α → γ and ⊢L γ → β for some γ(y)

George Metcalfe (University of Bern) Uniform Interpolation August 2018 9 / 27

slide-43
SLIDE 43

Interpolation in Modal Logic

A normal modal logic L admits deductive interpolation α(x, y) ⊢L β(y, z) = ⇒ α ⊢L γ and γ ⊢L β for some γ(y) if and only if VL admits the amalgamation property. For example, K, K4, S4, GL, and somewhere between 43 and 49 axiomatic extensions of S4 admit deductive interpolation, but not S5. Note, however, that L admits Craig interpolation ⊢L α(x, y) → β(y, z) = ⇒ ⊢L α → γ and ⊢L γ → β for some γ(y) if and only if VL admits the super amalgamation property.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 9 / 27

slide-44
SLIDE 44

Uniform Interpolation in Modal Logic

Theorem (Ghilardi 1995, Visser 1996, Bílková 2007)

K has uniform interpolation.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 10 / 27

slide-45
SLIDE 45

Uniform Interpolation in Modal Logic

Theorem (Ghilardi 1995, Visser 1996, Bílková 2007)

K has uniform interpolation.

Theorem (Kowalski and Metcalfe 2017)

K does not have uniform interpolation.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 10 / 27

slide-46
SLIDE 46

Uniform Interpolation in Modal Logic

Theorem (Ghilardi 1995, Visser 1996, Bílková 2007)

K has uniform Craig interpolation

Theorem (Kowalski and Metcalfe 2017)

K does not have uniform deductive interpolation.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 10 / 27

slide-47
SLIDE 47

Uniform Interpolation in Modal Logic

Theorem (Ghilardi 1995, Visser 1996, Bílková 2007)

K has uniform Craig interpolation; that is, for any formula α(x, y), there exist formulas αL(y) and αR(y) such that ⊢K α(x, y) → β(y, z) ⇐ ⇒ ⊢K αR(y) → β(y, z) ⊢K β(y, z) → α(x, y) ⇐ ⇒ ⊢K β(y, z) → αL(y).

Theorem (Kowalski and Metcalfe 2017)

K does not have uniform deductive interpolation.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 10 / 27

slide-48
SLIDE 48
  • Recall. . .

V has deductive interpolation if for any set of equations Σ(x, y), there exists a set of equations ∆(y) such that Σ(x, y) | =V ε(y, z) ⇐ ⇒ ∆(y) | =V ε(y, z).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 11 / 27

slide-49
SLIDE 49
  • Recall. . .

V has right uniform deductive interpolation if for any finite set of equations Σ(x, y), there exists a finite set of equations ∆(y) such that Σ(x, y) | =V ε(y, z) ⇐ ⇒ ∆(y) | =V ε(y, z).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 11 / 27

slide-50
SLIDE 50
  • Recall. . .

V has right uniform deductive interpolation if for any finite set of equations Σ(x, y), there exists a finite set of equations ∆(y) such that Σ(x, y) | =V ε(y, z) ⇐ ⇒ ∆(y) | =V ε(y, z). Equivalently, V has deductive interpolation and for any finite set of equations Σ(x, y), there exists a finite set of equations ∆(y) such that Σ(x, y) | =V ε(y) ⇐ ⇒ ∆(y) | =V ε(y).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 11 / 27

slide-51
SLIDE 51

Recall also. . .

Theorem (Kowalski and Metcalfe 2017)

The following are equivalent: (1) For any finite set of equations Σ(x, y), there exists a finite set of equations ∆(y) such that Σ(x, y) | =V ε(y) ⇐ ⇒ ∆(y) | =V ε(y).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 12 / 27

slide-52
SLIDE 52

Recall also. . .

Theorem (Kowalski and Metcalfe 2017)

The following are equivalent: (1) For any finite set of equations Σ(x, y), there exists a finite set of equations ∆(y) such that Σ(x, y) | =V ε(y) ⇐ ⇒ ∆(y) | =V ε(y). (2) For finite x, y, the compact lifting of F(y) ֒ → F(x, y) has a right adjoint; that is, Θ ∈ KCon F(x, y) = ⇒ Θ ∩ F(y)2 ∈ KCon F(y).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 12 / 27

slide-53
SLIDE 53

Recall also. . .

Theorem (Kowalski and Metcalfe 2017)

The following are equivalent: (1) For any finite set of equations Σ(x, y), there exists a finite set of equations ∆(y) such that Σ(x, y) | =V ε(y) ⇐ ⇒ ∆(y) | =V ε(y). (2) For finite x, y, the compact lifting of F(y) ֒ → F(x, y) has a right adjoint; that is, Θ ∈ KCon F(x, y) = ⇒ Θ ∩ F(y)2 ∈ KCon F(y). (3) V is coherent, i.e., all finitely generated subalgebras of finitely presented members of V are finitely presented.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 12 / 27

slide-54
SLIDE 54

A Failure of Coherence

Theorem (Kowalski and Metcalfe 2017)

The variety of modal algebras is not coherent; so it does not admit uniform deductive interpolation and its theory does not have a model completion.

  • T. Kowalski and G. Metcalfe.

Uniform interpolation and coherence. Submitted (2017).

  • T. Kowalski and G. Metcalfe.

Coherence in modal logic. Proceedings of AiML 2018, College Publications (2018).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 13 / 27

slide-55
SLIDE 55

Proof

Let ⊡α := α ∧ α,

George Metcalfe (University of Bern) Uniform Interpolation August 2018 14 / 27

slide-56
SLIDE 56

Proof

Let ⊡α := α ∧ α, and define Σ = {y ≤ x, x ≤ z, x ≈ ⊡x}

George Metcalfe (University of Bern) Uniform Interpolation August 2018 14 / 27

slide-57
SLIDE 57

Proof

Let ⊡α := α ∧ α, and define Σ = {y ≤ x, x ≤ z, x ≈ ⊡x} and ∆ = {y ≤ ⊡kz | k ∈ N}.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 14 / 27

slide-58
SLIDE 58

Proof

Let ⊡α := α ∧ α, and define Σ = {y ≤ x, x ≤ z, x ≈ ⊡x} and ∆ = {y ≤ ⊡kz | k ∈ N}.

  • Claim. Σ |

=K ε(y, z) ⇐ ⇒ ∆ | =K ε(y, z).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 14 / 27

slide-59
SLIDE 59

Proof

Let ⊡α := α ∧ α, and define Σ = {y ≤ x, x ≤ z, x ≈ ⊡x} and ∆ = {y ≤ ⊡kz | k ∈ N}.

  • Claim. Σ |

=K ε(y, z) ⇐ ⇒ ∆ | =K ε(y, z). It follows that if K were coherent, then ∆′ | =K ∆ for some finite ∆′ ⊆ ∆,

George Metcalfe (University of Bern) Uniform Interpolation August 2018 14 / 27

slide-60
SLIDE 60

Proof

Let ⊡α := α ∧ α, and define Σ = {y ≤ x, x ≤ z, x ≈ ⊡x} and ∆ = {y ≤ ⊡kz | k ∈ N}.

  • Claim. Σ |

=K ε(y, z) ⇐ ⇒ ∆ | =K ε(y, z). It follows that if K were coherent, then ∆′ | =K ∆ for some finite ∆′ ⊆ ∆, and from this that K | = ⊡nz ≈ ⊡n+1z for some n ∈ N, a contradiction.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 14 / 27

slide-61
SLIDE 61

Proof

Let ⊡α := α ∧ α, and define Σ = {y ≤ x, x ≤ z, x ≈ ⊡x} and ∆ = {y ≤ ⊡kz | k ∈ N}.

  • Claim. Σ |

=K ε(y, z) ⇐ ⇒ ∆ | =K ε(y, z). It follows that if K were coherent, then ∆′ | =K ∆ for some finite ∆′ ⊆ ∆, and from this that K | = ⊡nz ≈ ⊡n+1z for some n ∈ N, a contradiction. Proof of claim. (⇐) Just observe that Σ | =K ∆.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 14 / 27

slide-62
SLIDE 62

Proof

Let ⊡α := α ∧ α, and define Σ = {y ≤ x, x ≤ z, x ≈ ⊡x} and ∆ = {y ≤ ⊡kz | k ∈ N}.

  • Claim. Σ |

=K ε(y, z) ⇐ ⇒ ∆ | =K ε(y, z). It follows that if K were coherent, then ∆′ | =K ∆ for some finite ∆′ ⊆ ∆, and from this that K | = ⊡nz ≈ ⊡n+1z for some n ∈ N, a contradiction. Proof of claim. (⇐) Just observe that Σ | =K ∆. (⇒) Suppose that ∆ | =K ε(y, z).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 14 / 27

slide-63
SLIDE 63

Proof

Let ⊡α := α ∧ α, and define Σ = {y ≤ x, x ≤ z, x ≈ ⊡x} and ∆ = {y ≤ ⊡kz | k ∈ N}.

  • Claim. Σ |

=K ε(y, z) ⇐ ⇒ ∆ | =K ε(y, z). It follows that if K were coherent, then ∆′ | =K ∆ for some finite ∆′ ⊆ ∆, and from this that K | = ⊡nz ≈ ⊡n+1z for some n ∈ N, a contradiction. Proof of claim. (⇐) Just observe that Σ | =K ∆. (⇒) Suppose that ∆ | =K ε(y, z). Then there is a complete modal algebra A and homomorphism e : Tm(y, z) → A such that ∆ ⊆ ker(e) and ε ∈ ker(e).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 14 / 27

slide-64
SLIDE 64

Proof

Let ⊡α := α ∧ α, and define Σ = {y ≤ x, x ≤ z, x ≈ ⊡x} and ∆ = {y ≤ ⊡kz | k ∈ N}.

  • Claim. Σ |

=K ε(y, z) ⇐ ⇒ ∆ | =K ε(y, z). It follows that if K were coherent, then ∆′ | =K ∆ for some finite ∆′ ⊆ ∆, and from this that K | = ⊡nz ≈ ⊡n+1z for some n ∈ N, a contradiction. Proof of claim. (⇐) Just observe that Σ | =K ∆. (⇒) Suppose that ∆ | =K ε(y, z). Then there is a complete modal algebra A and homomorphism e : Tm(y, z) → A such that ∆ ⊆ ker(e) and ε ∈ ker(e). Extend e with e(x) =

  • k∈N

⊡ke(z).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 14 / 27

slide-65
SLIDE 65

Proof

Let ⊡α := α ∧ α, and define Σ = {y ≤ x, x ≤ z, x ≈ ⊡x} and ∆ = {y ≤ ⊡kz | k ∈ N}.

  • Claim. Σ |

=K ε(y, z) ⇐ ⇒ ∆ | =K ε(y, z). It follows that if K were coherent, then ∆′ | =K ∆ for some finite ∆′ ⊆ ∆, and from this that K | = ⊡nz ≈ ⊡n+1z for some n ∈ N, a contradiction. Proof of claim. (⇐) Just observe that Σ | =K ∆. (⇒) Suppose that ∆ | =K ε(y, z). Then there is a complete modal algebra A and homomorphism e : Tm(y, z) → A such that ∆ ⊆ ker(e) and ε ∈ ker(e). Extend e with e(x) =

  • k∈N

⊡ke(z). Then also Σ ⊆ ker(e), and hence Σ | =K ε(y, z).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 14 / 27

slide-66
SLIDE 66

An Obvious Question

Can we generalize this proof to other varieties?

George Metcalfe (University of Bern) Uniform Interpolation August 2018 15 / 27

slide-67
SLIDE 67

A General Criterion

Theorem (Kowalski and Metcalfe 2017)

Let V be a coherent variety of algebras with a meet-semilattice reduct

George Metcalfe (University of Bern) Uniform Interpolation August 2018 16 / 27

slide-68
SLIDE 68

A General Criterion

Theorem (Kowalski and Metcalfe 2017)

Let V be a coherent variety of algebras with a meet-semilattice reduct and let t(x, ¯ u) be a term satisfying V | = t(x, ¯ u) ≤ x and V | = x ≤ y ⇒ t(x, ¯ u) ≤ t(y, ¯ u).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 16 / 27

slide-69
SLIDE 69

A General Criterion

Theorem (Kowalski and Metcalfe 2017)

Let V be a coherent variety of algebras with a meet-semilattice reduct and let t(x, ¯ u) be a term satisfying V | = t(x, ¯ u) ≤ x and V | = x ≤ y ⇒ t(x, ¯ u) ≤ t(y, ¯ u). Suppose also that for any finitely generated A ∈ V and a, ¯ b ∈ A, there exists B ∈ V containing A as a subalgebra

George Metcalfe (University of Bern) Uniform Interpolation August 2018 16 / 27

slide-70
SLIDE 70

A General Criterion

Theorem (Kowalski and Metcalfe 2017)

Let V be a coherent variety of algebras with a meet-semilattice reduct and let t(x, ¯ u) be a term satisfying V | = t(x, ¯ u) ≤ x and V | = x ≤ y ⇒ t(x, ¯ u) ≤ t(y, ¯ u). Suppose also that for any finitely generated A ∈ V and a, ¯ b ∈ A, there exists B ∈ V containing A as a subalgebra and satisfying

  • k∈N

tk(a, ¯ b) = t(

  • k∈N

tk(a, ¯ b), ¯ b).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 16 / 27

slide-71
SLIDE 71

A General Criterion

Theorem (Kowalski and Metcalfe 2017)

Let V be a coherent variety of algebras with a meet-semilattice reduct and let t(x, ¯ u) be a term satisfying V | = t(x, ¯ u) ≤ x and V | = x ≤ y ⇒ t(x, ¯ u) ≤ t(y, ¯ u). Suppose also that for any finitely generated A ∈ V and a, ¯ b ∈ A, there exists B ∈ V containing A as a subalgebra and satisfying

  • k∈N

tk(a, ¯ b) = t(

  • k∈N

tk(a, ¯ b), ¯ b). Then V | = tn(x, ¯ u) ≈ tn+1(x, ¯ u) for some n ∈ N.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 16 / 27

slide-72
SLIDE 72

Strong Kripke Completeness

A normal modal logic L is called strongly Kripke complete

George Metcalfe (University of Bern) Uniform Interpolation August 2018 17 / 27

slide-73
SLIDE 73

Strong Kripke Completeness

A normal modal logic L is called strongly Kripke complete if for any set

  • f formulas T ∪ {α},

T ⊢L α ⇐ ⇒ for any Kripke model M based on a frame for L, M | = β for all β ∈ T = ⇒ M | = α.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 17 / 27

slide-74
SLIDE 74

Strong Kripke Completeness

A normal modal logic L is called strongly Kripke complete if for any set

  • f formulas T ∪ {α},

T ⊢L α ⇐ ⇒ for any Kripke model M based on a frame for L, M | = β for all β ∈ T = ⇒ M | = α. E.g., K, KT, K4, S4, and S5 are strongly Kripke complete, but not GL.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 17 / 27

slide-75
SLIDE 75

Coherence and Weak Transitivity

Applying our general criterion with t(x) = ⊡x, using strong Kripke completeness to establish the fixpoint condition, we obtain:

George Metcalfe (University of Bern) Uniform Interpolation August 2018 18 / 27

slide-76
SLIDE 76

Coherence and Weak Transitivity

Applying our general criterion with t(x) = ⊡x, using strong Kripke completeness to establish the fixpoint condition, we obtain:

Theorem

Any coherent strongly Kripke-complete variety of modal algebras is weakly transitive:

George Metcalfe (University of Bern) Uniform Interpolation August 2018 18 / 27

slide-77
SLIDE 77

Coherence and Weak Transitivity

Applying our general criterion with t(x) = ⊡x, using strong Kripke completeness to establish the fixpoint condition, we obtain:

Theorem

Any coherent strongly Kripke-complete variety of modal algebras is weakly transitive: that is, it satisfies ⊡n+1x ≈ ⊡nx for some n ∈ N

George Metcalfe (University of Bern) Uniform Interpolation August 2018 18 / 27

slide-78
SLIDE 78

Coherence and Weak Transitivity

Applying our general criterion with t(x) = ⊡x, using strong Kripke completeness to establish the fixpoint condition, we obtain:

Theorem

Any coherent strongly Kripke-complete variety of modal algebras is weakly transitive: that is, it satisfies ⊡n+1x ≈ ⊡nx for some n ∈ N (equivalently, it admits equationally definable principal congruences).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 18 / 27

slide-79
SLIDE 79

Coherence and Weak Transitivity

Applying our general criterion with t(x) = ⊡x, using strong Kripke completeness to establish the fixpoint condition, we obtain:

Theorem

Any coherent strongly Kripke-complete variety of modal algebras is weakly transitive: that is, it satisfies ⊡n+1x ≈ ⊡nx for some n ∈ N (equivalently, it admits equationally definable principal congruences). Hence a large family of varieties of modal algebras for non-weakly-transitive modal logics, including K and KT, are not coherent, do not admit uniform deductive interpolation, and their theories do not have a model completion.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 18 / 27

slide-80
SLIDE 80

Weakly Transitive Modal Logics (1)

We can also show that weakly transitive logics such as K4 and S4 are not coherent

George Metcalfe (University of Bern) Uniform Interpolation August 2018 19 / 27

slide-81
SLIDE 81

Weakly Transitive Modal Logics (1)

We can also show that weakly transitive logics such as K4 and S4 are not coherent using the ternary term t(x, y, z) = ♦(y ∧ ♦(z ∧ x)) ∧ x.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 19 / 27

slide-82
SLIDE 82

Weakly Transitive Modal Logics (1)

We can also show that weakly transitive logics such as K4 and S4 are not coherent using the ternary term t(x, y, z) = ♦(y ∧ ♦(z ∧ x)) ∧ x. For any normal modal logic L, VL | = t(x, y, z) ≤ x and VL | = u ≤ v ⇒ t(u, y, z) ≤ t(v, y, z).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 19 / 27

slide-83
SLIDE 83

Weakly Transitive Modal Logics (1)

We can also show that weakly transitive logics such as K4 and S4 are not coherent using the ternary term t(x, y, z) = ♦(y ∧ ♦(z ∧ x)) ∧ x. For any normal modal logic L, VL | = t(x, y, z) ≤ x and VL | = u ≤ v ⇒ t(u, y, z) ≤ t(v, y, z).

Lemma

Suppose that L admits finite chains:

George Metcalfe (University of Bern) Uniform Interpolation August 2018 19 / 27

slide-84
SLIDE 84

Weakly Transitive Modal Logics (1)

We can also show that weakly transitive logics such as K4 and S4 are not coherent using the ternary term t(x, y, z) = ♦(y ∧ ♦(z ∧ x)) ∧ x. For any normal modal logic L, VL | = t(x, y, z) ≤ x and VL | = u ≤ v ⇒ t(u, y, z) ≤ t(v, y, z).

Lemma

Suppose that L admits finite chains: that is, for each n ∈ N there exists a frame W , R for L such that |W | = n and the reflexive closure of R is a total order.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 19 / 27

slide-85
SLIDE 85

Weakly Transitive Modal Logics (1)

We can also show that weakly transitive logics such as K4 and S4 are not coherent using the ternary term t(x, y, z) = ♦(y ∧ ♦(z ∧ x)) ∧ x. For any normal modal logic L, VL | = t(x, y, z) ≤ x and VL | = u ≤ v ⇒ t(u, y, z) ≤ t(v, y, z).

Lemma

Suppose that L admits finite chains: that is, for each n ∈ N there exists a frame W , R for L such that |W | = n and the reflexive closure of R is a total order. Then VL | = tn(x, y, z) ≈ tn+1(x, y, z) for all n ∈ N.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 19 / 27

slide-86
SLIDE 86

Weakly Transitive Modal Logics (2)

Theorem

Let L be a normal modal logic admitting finite chains

George Metcalfe (University of Bern) Uniform Interpolation August 2018 20 / 27

slide-87
SLIDE 87

Weakly Transitive Modal Logics (2)

Theorem

Let L be a normal modal logic admitting finite chains such that VL is canonical: that is, closed under taking canonical extensions.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 20 / 27

slide-88
SLIDE 88

Weakly Transitive Modal Logics (2)

Theorem

Let L be a normal modal logic admitting finite chains such that VL is canonical: that is, closed under taking canonical extensions. Then (a) VL is not coherent; (b) VL does not admit uniform deductive interpolation; (c) the first-order theory of VL does not have a model completion.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 20 / 27

slide-89
SLIDE 89

Weakly Transitive Modal Logics (2)

Theorem

Let L be a normal modal logic admitting finite chains such that VL is canonical: that is, closed under taking canonical extensions. Then (a) VL is not coherent; (b) VL does not admit uniform deductive interpolation; (c) the first-order theory of VL does not have a model completion. In particular, VK4 and VS4 are not coherent and do not admit uniform deductive interpolation, and their theories do not have a model completion.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 20 / 27

slide-90
SLIDE 90

Remarks

Note that GL admits finite chains but is not canonical.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 21 / 27

slide-91
SLIDE 91

Remarks

Note that GL admits finite chains but is not canonical. In fact, VGL is coherent and admits uniform deductive interpolation (Shavrukov 1993);

George Metcalfe (University of Bern) Uniform Interpolation August 2018 21 / 27

slide-92
SLIDE 92

Remarks

Note that GL admits finite chains but is not canonical. In fact, VGL is coherent and admits uniform deductive interpolation (Shavrukov 1993); also, its theory has a model completion (Ghilardi and Zawadowski 2002).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 21 / 27

slide-93
SLIDE 93

Remarks

Note that GL admits finite chains but is not canonical. In fact, VGL is coherent and admits uniform deductive interpolation (Shavrukov 1993); also, its theory has a model completion (Ghilardi and Zawadowski 2002). Ghilardi and Zawadowski (2002) have also proved that no logic extending K4 that has the finite model property and admits all finite reflexive chains and the two-element cluster is coherent.

  • S. Ghilardi and M. Zawadowski.

Sheaves, Games and Model Completions, Kluwer (2002).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 21 / 27

slide-94
SLIDE 94

Remarks

Note that GL admits finite chains but is not canonical. In fact, VGL is coherent and admits uniform deductive interpolation (Shavrukov 1993); also, its theory has a model completion (Ghilardi and Zawadowski 2002). Ghilardi and Zawadowski (2002) have also proved that no logic extending K4 that has the finite model property and admits all finite reflexive chains and the two-element cluster is coherent.

  • S. Ghilardi and M. Zawadowski.

Sheaves, Games and Model Completions, Kluwer (2002).

However, their methods are more complicated and less general than ours;

George Metcalfe (University of Bern) Uniform Interpolation August 2018 21 / 27

slide-95
SLIDE 95

Remarks

Note that GL admits finite chains but is not canonical. In fact, VGL is coherent and admits uniform deductive interpolation (Shavrukov 1993); also, its theory has a model completion (Ghilardi and Zawadowski 2002). Ghilardi and Zawadowski (2002) have also proved that no logic extending K4 that has the finite model property and admits all finite reflexive chains and the two-element cluster is coherent.

  • S. Ghilardi and M. Zawadowski.

Sheaves, Games and Model Completions, Kluwer (2002).

However, their methods are more complicated and less general than ours; they also yield similar but incomparable results.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 21 / 27

slide-96
SLIDE 96

Coherence in Algebra

Any locally finite variety (e.g., Boolean algebras, Sugihara monoids, etc.) is coherent

George Metcalfe (University of Bern) Uniform Interpolation August 2018 22 / 27

slide-97
SLIDE 97

Coherence in Algebra

Any locally finite variety (e.g., Boolean algebras, Sugihara monoids, etc.) is coherent — also the varieties of Heyting algebras, abelian groups, abelian ℓ-groups, and MV-algebras.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 22 / 27

slide-98
SLIDE 98

Coherence in Algebra

Any locally finite variety (e.g., Boolean algebras, Sugihara monoids, etc.) is coherent — also the varieties of Heyting algebras, abelian groups, abelian ℓ-groups, and MV-algebras. The variety of groups is not coherent, however, since every finitely generated recursively presented group embeds into some finitely presented group (Higman 1961).

George Metcalfe (University of Bern) Uniform Interpolation August 2018 22 / 27

slide-99
SLIDE 99

Lattices

Theorem (Schmidt 1981)

The variety LAT of lattices is not coherent, does not admit uniform deductive interpolation, and its theory does not have a model completion.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 23 / 27

slide-100
SLIDE 100

Lattices

Theorem (Schmidt 1981)

The variety LAT of lattices is not coherent, does not admit uniform deductive interpolation, and its theory does not have a model completion. We obtain an easy proof of this result using our criterion with the term t(x, u1, u2, u3) = (u1 ∧ (u2 ∨ (u3 ∧ x))) ∧ x.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 23 / 27

slide-101
SLIDE 101

Lattices

Theorem (Schmidt 1981)

The variety LAT of lattices is not coherent, does not admit uniform deductive interpolation, and its theory does not have a model completion. We obtain an easy proof of this result using our criterion with the term t(x, u1, u2, u3) = (u1 ∧ (u2 ∨ (u3 ∧ x))) ∧ x. Just note that LAT is closed under taking canonical completions;

George Metcalfe (University of Bern) Uniform Interpolation August 2018 23 / 27

slide-102
SLIDE 102

Lattices

Theorem (Schmidt 1981)

The variety LAT of lattices is not coherent, does not admit uniform deductive interpolation, and its theory does not have a model completion. We obtain an easy proof of this result using our criterion with the term t(x, u1, u2, u3) = (u1 ∧ (u2 ∨ (u3 ∧ x))) ∧ x. Just note that LAT is closed under taking canonical completions; LAT | = x ≤ t(x, ¯ u) and LAT | = x ≤ y ⇒ t(x, ¯ u) ≤ t(y, ¯ u);

George Metcalfe (University of Bern) Uniform Interpolation August 2018 23 / 27

slide-103
SLIDE 103

Lattices

Theorem (Schmidt 1981)

The variety LAT of lattices is not coherent, does not admit uniform deductive interpolation, and its theory does not have a model completion. We obtain an easy proof of this result using our criterion with the term t(x, u1, u2, u3) = (u1 ∧ (u2 ∨ (u3 ∧ x))) ∧ x. Just note that LAT is closed under taking canonical completions; LAT | = x ≤ t(x, ¯ u) and LAT | = x ≤ y ⇒ t(x, ¯ u) ≤ t(y, ¯ u); LAT | = tn(x, ¯ u) ≈ tn+1(x, ¯ u) for each n ∈ N.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 23 / 27

slide-104
SLIDE 104

Residuated Lattices

A residuated lattice is an algebraic structure A, ∧, ∨, ·, \, /, e such that A, ∧, ∨ is a lattice, A, ·, e is a monoid, and for all a, b, c ∈ A, b ≤ a\c ⇐ ⇒ a · b ≤ c ⇐ ⇒ a ≤ c/b.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 24 / 27

slide-105
SLIDE 105

Residuated Lattices

A residuated lattice is an algebraic structure A, ∧, ∨, ·, \, /, e such that A, ∧, ∨ is a lattice, A, ·, e is a monoid, and for all a, b, c ∈ A, b ≤ a\c ⇐ ⇒ a · b ≤ c ⇐ ⇒ a ≤ c/b. Applying our criterion with the term t(x) = (x ∧ e)2, we obtain

Theorem

Any coherent variety of residuated lattices that is closed under canonical extensions satisfies (x ∧ e)n+1 ≈ (x ∧ e)n for some n ∈ N.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 24 / 27

slide-106
SLIDE 106

Residuated Lattices

A residuated lattice is an algebraic structure A, ∧, ∨, ·, \, /, e such that A, ∧, ∨ is a lattice, A, ·, e is a monoid, and for all a, b, c ∈ A, b ≤ a\c ⇐ ⇒ a · b ≤ c ⇐ ⇒ a ≤ c/b. Applying our criterion with the term t(x) = (x ∧ e)2, we obtain

Theorem

Any coherent variety of residuated lattices that is closed under canonical extensions satisfies (x ∧ e)n+1 ≈ (x ∧ e)n for some n ∈ N. It follows that varieties of residuated lattices corresponding to the most commonly studied substructural logics are not coherent, and do not admit uniform deductive interpolation.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 24 / 27

slide-107
SLIDE 107

Challenge 1: Understanding Fixpoints

Our general criterion shows that in a coherent variety with a semilattice reduct, terms satisfying certain conditions admit fixpoints.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 25 / 27

slide-108
SLIDE 108

Challenge 1: Understanding Fixpoints

Our general criterion shows that in a coherent variety with a semilattice reduct, terms satisfying certain conditions admit fixpoints. Might it be the case that, conversely, admitting such fixpoints guarantees the coherence of the variety?

George Metcalfe (University of Bern) Uniform Interpolation August 2018 25 / 27

slide-109
SLIDE 109

Challenge 1: Understanding Fixpoints

Our general criterion shows that in a coherent variety with a semilattice reduct, terms satisfying certain conditions admit fixpoints. Might it be the case that, conversely, admitting such fixpoints guarantees the coherence of the variety? Indeed for certain fixpoint modal logics, the fixpoint operators have been used to construct uniform interpolants.

  • G. D’Agostino. Uniform interpolation, bisimulation quantifiers, and fixed points.

Proceedings of TbiLLC’05, pages 96–116, 2005.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 25 / 27

slide-110
SLIDE 110

Challenge 2: Dealing with Failure

We have seen that the most well-studied modal and substructural logics, and many important varieties from algebra, are not coherent.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 26 / 27

slide-111
SLIDE 111

Challenge 2: Dealing with Failure

We have seen that the most well-studied modal and substructural logics, and many important varieties from algebra, are not coherent. In such cases, can we determine instead which terms do admit uniform interpolants?

George Metcalfe (University of Bern) Uniform Interpolation August 2018 26 / 27

slide-112
SLIDE 112

Challenge 2: Dealing with Failure

We have seen that the most well-studied modal and substructural logics, and many important varieties from algebra, are not coherent. In such cases, can we determine instead which terms do admit uniform interpolants? This problem has been considered for certain description logics, using bisimulations to calculate uniform interpolants when they exist.

  • C. Lutz and F. Wolter

Foundations for uniform interpolation and forgetting in expressive description logics. Proceedings of IJCAI 2011, AAAI Press, pages 989–996, 2011.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 26 / 27

slide-113
SLIDE 113

Challenge 2: Dealing with Failure

We have seen that the most well-studied modal and substructural logics, and many important varieties from algebra, are not coherent. In such cases, can we determine instead which terms do admit uniform interpolants? This problem has been considered for certain description logics, using bisimulations to calculate uniform interpolants when they exist.

  • C. Lutz and F. Wolter

Foundations for uniform interpolation and forgetting in expressive description logics. Proceedings of IJCAI 2011, AAAI Press, pages 989–996, 2011.

Can we develop similar methods for constructing uniform interpolants for modal logics, lattices, residuated lattices, etc.?

George Metcalfe (University of Bern) Uniform Interpolation August 2018 26 / 27

slide-114
SLIDE 114

Challenge 3: Understanding Model Completions

Can we understand the conservative congruence extension property appearing in Wheeler’s theorem as a property of consequence?

Theorem (Wheeler 1976)

The theory of a variety has a model completion if and only if it is coherent, admits the amalgamation property, and has the conservative congruence extension property for its finitely presented members.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 27 / 27

slide-115
SLIDE 115

Challenge 3: Understanding Model Completions

Can we understand the conservative congruence extension property appearing in Wheeler’s theorem as a property of consequence?

Theorem (Wheeler 1976)

The theory of a variety has a model completion if and only if it is coherent, admits the amalgamation property, and has the conservative congruence extension property for its finitely presented members. Can we extend Ghilardi and Zawadowski’s theorem to quasivarieties or (positive) universal classes?

Theorem (Ghilardi and Zawadowski 2002, van Gool et al. 2017)

If the a variety admits left and right uniform interpolation and the join-semilattice of compact congruences of any finitely generated free algebra is dually Brouwerian, then its theory has a model completion.

George Metcalfe (University of Bern) Uniform Interpolation August 2018 27 / 27