P ossible W orlds S emantics for C onditionals : T he C ase of C - - PowerPoint PPT Presentation

p ossible w orlds s emantics for c onditionals t he c ase
SMART_READER_LITE
LIVE PREVIEW

P ossible W orlds S emantics for C onditionals : T he C ase of C - - PowerPoint PPT Presentation

P ossible W orlds S emantics for C onditionals : T he C ase of C hellas -S egerberg S emantics Matthias Unterhuber Munic Center for Mathematical Philosophy matthias.unterhuber@lrz.uni-muenchen.de D agstuhl S eminar 15221 May 25-29, 2015 I


slide-1
SLIDE 1

Possible Worlds Semantics for Conditionals: The Case of Chellas-Segerberg Semantics

Matthias Unterhuber

Munic Center for Mathematical Philosophy matthias.unterhuber@lrz.uni-muenchen.de

Dagstuhl Seminar 15221

May 25-29, 2015

slide-2
SLIDE 2

Introduction

Chellas-Segerberg Semantics: A Possible-Semantics for Conditionals Goes back to Chellas (1975) and Segerberg (1989) Modal base logic for conditional logic A number of advantages Correspondence and completeness results for a lattice of systems based on thirty principles and frame conditions (Unterhuber, 2013; Unterhuber & Schurz, 2014) Can, for example, model the systems of Kraus, Lehmann, and Magidor (1990), Lewis (1973), and Adams (1966). Today: Outline of the correspondence and completeness result

slide-3
SLIDE 3

Advantage 1

Full language, in contrast to the probabilistic semantics of Adams (1966, 1977) without triviality à la Lewis (1976). Full language L:

1 contains the set of atomic variables AV = {p, p′, . . . } and 2 is closed under 1

truth-functional propositional operators ¬ (negation) and ∨ (disjunction) as well as

2

the two-place modal operator (conditional”) plus its dual .

slide-4
SLIDE 4

Semantics

Definition 1 FC = W, R is a Chellas frame iff (a) W is a non-empty set of indices and (b) R ⊆ W × W × Pow(W). Definition 2 Let F = W, R be a Chellas frame. Then, M = W, R, V is a Chellas model iff V is a valuation function from AV × W to {0, 1} and for all formulas α, β ∈ L and w ∈ W it holds: (i) w |=M ¬α iff w |=M α (ii) w |=M α ∨ β iff w |=M α or w |=M β (ii) w |=M α β iff for all w′: if wRαMw′ then w′ |=M β w |=M α: V(α, w) = 1 w |=M α: not w |=M α for w, w′ ∈ W and X ⊆ W wRXw′: w, w′, X ∈ R αM = {w | w |=M α}

slide-5
SLIDE 5

Proof Theory

LLE if α ↔ β then (α γ) → (β γ) RW if α → β then (γ α) → (γ β) AND (α β) ∧ (α γ) → (α β ∧ γ) LT α ⊤ More interesting: What it is missing

slide-6
SLIDE 6

Advantage 2: Correspondence

In CS Semantics structural (frame-based) conditions can be specified. Note: Such structural conditions cannot be specified when R ⊆ W × W × L. Examples: (Refl) α α (CM) (α γ) ∧ (α β) → (α ∧ β γ) (RM) (α γ) ∧ (α β) → (α ∧ β γ) (Or) (α γ) ∧ (β γ) → (α ∨ β γ) C∗

Refl

∀w∀w′(wRXw′ → w′ ∈ X) CCM ∀w(∀w′(wRXw′ → w′ ∈ Y) → ∀w′(wRX∩Yw′ → wRXw′)) CRM ∀w(∃w′(wRXw′ ∧ w′ ∈ Y) → ∀w′(wRX∩Yw′ → wRXw′)) COr ∀w∀w′(wRX∪Yw′ → wRXw′ ∨ wRYw′) A lattice of systems available as specified by thirty pairs of conditional logic principles and frame conditions (Unterhuber, 2013; Chellas, 1975; Segerberg, 1989).

slide-7
SLIDE 7

Advantage 2

For example, (α β) → β (T) corresponds to ∀w(wRXw) (C) in the following sense: (T) is valid on a Chellas frame W, R iff C holds for all X ⊆ W, where valid on frame W, R is being true at all worlds w ∈ W for all models W, R, V. For a given Chellas frame W, R, (T) is true at all worlds w ∈ W for all models W, R, V (short: valid on frame W, R) iff C holds for all X ⊆ W. This is not true for Chellas models. Consider (T) and consider the model M = W, R, V, where W = {w, w′}, where w and w′ can see each other with respect to ∅ and W and where V(α, w) = V(α, w′) for all α ∈ L.

slide-8
SLIDE 8

Completeness with a Twist

Road map: Completeness proof with respect to classes of Chellas frames does not go through (as in Chellas, 1975). Problem: There are sets of possible worlds in Chellas models for which there is no formula α such that α is true in all worlds of that set. Completeness with respect to classes of Segerberg frames (to be defined), as in Segerberg (1989), is trivial: Any such logic system would be complete. Solution: Use a narrower class of Segerberg frames (standard Segerberg frames)

slide-9
SLIDE 9

Segerberg frames

Definition 3 FS = W, R, P is a Segerberg frame iff (a) W, R is a Chellas frame and (b) R ⊆ W × W × P for P ⊆ Pow(W) such that (i) ∅ ∈ P, (ii) if X ∈ P then −X ∈ P, (iii) if X, Y ∈ P then X ∪ Y ∈ P, and (iv) if X, Y ∈ P then X ∗Y ∈ P. X ∗Y = {w ∈ W | ∀w′ ∈ W(wRXw′ → w′ ∈ Y)} ∈ P for X, Y ⊆ W Segerberg frames correspond to general frames in normal modal logics. Definition 4 MS = W, R, P, V is a Segerberg model iff (a) W, R, P is a Segerberg frame and (b) W, R, V is a Chellas model such that for all α ∈ L: αM ∈ P.

slide-10
SLIDE 10

Completeness: Triviality and Solution

The source of the problem: Theorem 5 For each Chellas model M = W, R, V there exists a Segerberg frame MS = W, R, PV, where PV = {αM | α ∈ L}, so that all formulas valid on M are valid on MS, and vice versa. In particular, Segerberg frame completeness does not exclude non-standard models. Solution: Restrict the class of Segerberg frames which qualify for the extension of the base system (excluding non-standard Segerberg frames). Language of frame conditions: Multi-sorted language with two types of terms,

  • ne referring to single possible worlds and

the other referring to sets of possible worlds (set denoting)

slide-11
SLIDE 11

Completeness: Triviality and Solution

1 No trivial occurrence of subformulæ: No logical equivalent formula which

does not contain that subformula. Expressions such as wRXw′ ∨ ¬wRXw′ are excluded.

2 No trivial occurrence of set variables: No set denoting term is equivalent

in with an additional set variable occurs (Boolean equivalence + X). Expressions such as Y in X ∩ (Y ∪ ¬Y). Essential: No set variable occurs in a frame condition satisfying of (1) and (2) unless it occurs also in the third argument place of some occurrence of R. This gives us exactly what we wanted. For (T) ∀w∀w′((wRXw′ → w′ ∈ Y) → w′ ∈ Y) is excluded, whereas ∀w(wRXw) is not. Important: This requirement generalizes from Kripke model correspondence.

slide-12
SLIDE 12

Many thanks!

slide-13
SLIDE 13

References I

Adams, E. (1966). Probability and the Logic of Conditionals. In J. Hintikka & P . Suppes (Eds.), Aspects of Inductive Logic (pp. 265–316). Amsterdam: North-Holland Publishing Company. Adams, E. (1977). A Note on Comparing Probabilistic and Modal Logics of

  • Conditionals. Theoria, 18, 186–194.

Chellas, B. F. (1975). Basic Conditional Logic. Journal of Philosophical Logic, 4, 133–153. Kraus, S., Lehmann, D., & Magidor, M. (1990). Nonmonotonic Reasoning, Preferential Models and Cumulative Logics. Artificial Intelligence, 44, 167–207. Lewis, D. (1973). Counterfactuals. Blackwell. Lewis, D. (1976). Probabilities of Conditionals and Conditional Probabilities. The Philosophical Review, 85, 297–315. Segerberg, K. (1989). Notes on Conditional Logic. Studia Logica, 48, 157–168. Unterhuber, M. (2013). Possible Worlds Semantics for Indicative and Counterfactual Conditionals? A Formal Philosophical Inquiry into Chellas-Segerberg Semantics. Frankfurt am Main: Ontos Verlag (Logos Series).

slide-14
SLIDE 14

References II

Unterhuber, M., & Schurz, G. (2014). Completeness and correspondence in Chellas-Segerberg semantics. Studia Logica, 102.