X -logics based multivalued reasoning for dialogical agents (ongoing - - PowerPoint PPT Presentation

x logics based multivalued reasoning for dialogical
SMART_READER_LITE
LIVE PREVIEW

X -logics based multivalued reasoning for dialogical agents (ongoing - - PowerPoint PPT Presentation

X -logics based multivalued reasoning for dialogical agents (ongoing work) Vincent Risch Aix-Marseille Univ., LSIS UMR CNRS 7296 Madeira Worshop on Belief Revision and Argumentation, 2015 V. Risch (LSIS) Argumentation, NMatrices, X


slide-1
SLIDE 1

X-logics based multivalued reasoning for dialogical agents (ongoing work)

Vincent Risch

Aix-Marseille Univ., LSIS – UMR CNRS 7296

Madeira Worshop on Belief Revision and Argumentation, 2015

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 1 / 29

slide-2
SLIDE 2

Motivations

A formal attempt around eristic argumentation: human disputes (either explicitely or implicitely) relies on rethoric techniques;

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 2 / 29

slide-3
SLIDE 3

Motivations

A formal attempt around eristic argumentation: human disputes (either explicitely or implicitely) relies on rethoric techniques;

  • ne goal of these techniques, beyond checking the correctness of the

reasoning of the speakers, is to help avoiding revision as long as possible;

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 2 / 29

slide-4
SLIDE 4

Motivations

A formal attempt around eristic argumentation: human disputes (either explicitely or implicitely) relies on rethoric techniques;

  • ne goal of these techniques, beyond checking the correctness of the

reasoning of the speakers, is to help avoiding revision as long as possible; e.g. greek sophists, Socrate, scholastic disputia, Thomas Aquina, Schopenhauer, modern political disputes...

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 2 / 29

slide-5
SLIDE 5

Motivations

A formal attempt around eristic argumentation: human disputes (either explicitely or implicitely) relies on rethoric techniques;

  • ne goal of these techniques, beyond checking the correctness of the

reasoning of the speakers, is to help avoiding revision as long as possible; e.g. greek sophists, Socrate, scholastic disputia, Thomas Aquina, Schopenhauer, modern political disputes... these techniques :

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 2 / 29

slide-6
SLIDE 6

Motivations

A formal attempt around eristic argumentation: human disputes (either explicitely or implicitely) relies on rethoric techniques;

  • ne goal of these techniques, beyond checking the correctness of the

reasoning of the speakers, is to help avoiding revision as long as possible; e.g. greek sophists, Socrate, scholastic disputia, Thomas Aquina, Schopenhauer, modern political disputes... these techniques :

assume dialogical agents;

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 2 / 29

slide-7
SLIDE 7

Motivations

A formal attempt around eristic argumentation: human disputes (either explicitely or implicitely) relies on rethoric techniques;

  • ne goal of these techniques, beyond checking the correctness of the

reasoning of the speakers, is to help avoiding revision as long as possible; e.g. greek sophists, Socrate, scholastic disputia, Thomas Aquina, Schopenhauer, modern political disputes... these techniques :

assume dialogical agents; aim to represent the mechanisms of a dispute;

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 2 / 29

slide-8
SLIDE 8

Motivations

A formal attempt around eristic argumentation: human disputes (either explicitely or implicitely) relies on rethoric techniques;

  • ne goal of these techniques, beyond checking the correctness of the

reasoning of the speakers, is to help avoiding revision as long as possible; e.g. greek sophists, Socrate, scholastic disputia, Thomas Aquina, Schopenhauer, modern political disputes... these techniques :

assume dialogical agents; aim to represent the mechanisms of a dispute; aim to achieve reasoning on arguments;

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 2 / 29

slide-9
SLIDE 9

Motivations

A formal attempt around eristic argumentation: human disputes (either explicitely or implicitely) relies on rethoric techniques;

  • ne goal of these techniques, beyond checking the correctness of the

reasoning of the speakers, is to help avoiding revision as long as possible; e.g. greek sophists, Socrate, scholastic disputia, Thomas Aquina, Schopenhauer, modern political disputes... these techniques :

assume dialogical agents; aim to represent the mechanisms of a dispute; aim to achieve reasoning on arguments;

One of our far(!) ideal(!!) goal : simulation of strategies for the ’game’ of argumentation

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 2 / 29

slide-10
SLIDE 10

Outline

1

X–logics

2

Their use in the context of a dialogical framework

3

Nmatrices, Nsequents

4

Transformation into classical sequents

5

LA, logic of attitudes

6

Links with MSPL (Avron et Al.)

7

Epilog Aim : attempt for defining a dialogical framework in which two ’agents’ can achieve ’some’ reasoning on their arguments.

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 3 / 29

slide-11
SLIDE 11

X–logics [Siegel, Forget, 96]

Definition Classical Inference: K ⊢ f iff K ∪ {f } = K

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 4 / 29

slide-12
SLIDE 12

X–logics [Siegel, Forget, 96]

Definition Classical Inference: K ⊢ f iff K ∪ {f } = K Generalisation : K ⊢X f iff K ∪ {f } ∩ X = K ∩ X

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 4 / 29

slide-13
SLIDE 13

X–logics [Siegel, Forget, 96]

Definition Classical Inference: K ⊢ f iff K ∪ {f } = K Generalisation : K ⊢X f iff K ∪ {f } ∩ X = K ∩ X Theorem K ⊢X f iff K ∪ {f } ∩ X ⊆ K i.e. (∀x ∈ X)(K ∧ f ⊢ x ⇒ K ⊢ x)

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 4 / 29

slide-14
SLIDE 14

X–logics [Siegel, Forget, 96]

Definition Classical Inference: K ⊢ f iff K ∪ {f } = K Generalisation : K ⊢X f iff K ∪ {f } ∩ X = K ∩ X Theorem K ⊢X f iff K ∪ {f } ∩ X ⊆ K i.e. (∀x ∈ X)(K ∧ f ⊢ x ⇒ K ⊢ x) Vocabulary f is compatible with K regarding X iff K ⊢X f , incompatible otherwise.

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 4 / 29

slide-15
SLIDE 15

X–logics

Properties

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 5 / 29

slide-16
SLIDE 16

X–logics

Properties (nonmonotonicity) K ⊢X f does not involve K ∪ K ′ ⊢X f

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 5 / 29

slide-17
SLIDE 17

X–logics

Properties (nonmonotonicity) K ⊢X f does not involve K ∪ K ′ ⊢X f (supraclassicity) K ⊢ f ⇒ K ⊢X f

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 5 / 29

slide-18
SLIDE 18

X–logics

Properties (nonmonotonicity) K ⊢X f does not involve K ∪ K ′ ⊢X f (supraclassicity) K ⊢ f ⇒ K ⊢X f (paraconsistancy) both a formula and its negation can be compatible with K regarding X (resp. incompatibles)

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 5 / 29

slide-19
SLIDE 19

X–logics

Properties (nonmonotonicity) K ⊢X f does not involve K ∪ K ′ ⊢X f (supraclassicity) K ⊢ f ⇒ K ⊢X f (paraconsistancy) both a formula and its negation can be compatible with K regarding X (resp. incompatibles) (cumulativity) If X c = X c then ⊢x is cumulative

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 5 / 29

slide-20
SLIDE 20

X–logics

Properties (nonmonotonicity) K ⊢X f does not involve K ∪ K ′ ⊢X f (supraclassicity) K ⊢ f ⇒ K ⊢X f (paraconsistancy) both a formula and its negation can be compatible with K regarding X (resp. incompatibles) (cumulativity) If X c = X c then ⊢x is cumulative (classical reasoning) If X = X then ⊢x is ⊢

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 5 / 29

slide-21
SLIDE 21

X–logics

Properties (nonmonotonicity) K ⊢X f does not involve K ∪ K ′ ⊢X f (supraclassicity) K ⊢ f ⇒ K ⊢X f (paraconsistancy) both a formula and its negation can be compatible with K regarding X (resp. incompatibles) (cumulativity) If X c = X c then ⊢x is cumulative (classical reasoning) If X = X then ⊢x is ⊢ Example

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 5 / 29

slide-22
SLIDE 22

X–logics

Properties (nonmonotonicity) K ⊢X f does not involve K ∪ K ′ ⊢X f (supraclassicity) K ⊢ f ⇒ K ⊢X f (paraconsistancy) both a formula and its negation can be compatible with K regarding X (resp. incompatibles) (cumulativity) If X c = X c then ⊢x is cumulative (classical reasoning) If X = X then ⊢x is ⊢ Example {a} ⊢{⊥} b and {a, ¬b} {⊥} b

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 5 / 29

slide-23
SLIDE 23

X–logics

Properties (nonmonotonicity) K ⊢X f does not involve K ∪ K ′ ⊢X f (supraclassicity) K ⊢ f ⇒ K ⊢X f (paraconsistancy) both a formula and its negation can be compatible with K regarding X (resp. incompatibles) (cumulativity) If X c = X c then ⊢x is cumulative (classical reasoning) If X = X then ⊢x is ⊢ Example {a} ⊢{⊥} b and {a, ¬b} {⊥} b {b} ⊢{⊥} a ∧ b and {b} ⊢{⊥} ¬(a ∧ b)

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 5 / 29

slide-24
SLIDE 24

X–logics

Properties (nonmonotonicity) K ⊢X f does not involve K ∪ K ′ ⊢X f (supraclassicity) K ⊢ f ⇒ K ⊢X f (paraconsistancy) both a formula and its negation can be compatible with K regarding X (resp. incompatibles) (cumulativity) If X c = X c then ⊢x is cumulative (classical reasoning) If X = X then ⊢x is ⊢ Example {a} ⊢{⊥} b and {a, ¬b} {⊥} b {b} ⊢{⊥} a ∧ b and {b} ⊢{⊥} ¬(a ∧ b) {b} {⊥,a,¬a} a ∧ b and {b} {⊥,a,¬a} ¬(a ∧ b)

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 5 / 29

slide-25
SLIDE 25

Why X-logics for argumentation?

they correspond to permissive inference relations (Bochman); as such, they caracterize a broad family of supra-classical relations; travel among relations via 2L; the inner structure of X allows to construct different kind of logics, hence different kinds of agents; provide parts of the underlying langage with a “logical” status; also(!): among the different Xs, try to get fragments of lower complexity...

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 6 / 29

slide-26
SLIDE 26

Agents and attitudes

Definition An agent is a couple [K, X], with K a consistant set of formulas, and X a set of formulas containing ⊥. The set of all agents is written A. a formula f is admissible by an agent [K, X] iff this formula is compatible with K regarding X. Attitudes

f is admissible f is non-admissible f is admissible f is non-admissible Puzzled Neutral For Against

X

f

X

f

X f X f

K K K K

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 7 / 29

slide-27
SLIDE 27

Answers and arguments

Definition An answer of the agent Φ = [K, X] to a set F is a set A composed both of formulas of K and of negated formulas of X, and such that F is incompatible with K regarding X. An argument α given by the agent [K, X] in the presence of a formula C is a couple S, ¬C such that S is an answer to C. S and ¬C are respectively the support and the conclusion of the argument. Definition Given α et β two arguments, and {s1, . . . , sn} ⊆ supp(β) : α attacks β iff concl(α) = ¬(s1 ∧ · · · ∧ sn) α defends β iff concl(α) = s1 ∧ · · · ∧ sn

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 8 / 29

slide-28
SLIDE 28

Example: the deafs dialogue

1: “You are rigid, be flexible” 2: “No. YOU are lax, be thorough”

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 9 / 29

slide-29
SLIDE 29

Example: the deafs dialogue

1: “You are rigid, be flexible” 2: “No. YOU are lax, be thorough” Consider [K1, X1] and [K2, X2] with K1 = {Flexible ⇒ ¬Lax, ¬Rigid ⇒ Flexible} X1 = {Rigid} K2 = {Thorough ⇒ ¬Rigid, ¬Lax ⇒ Thorough} X2 = {Lax}

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 9 / 29

slide-30
SLIDE 30

Example: the deafs dialogue

1: “You are rigid, be flexible” 2: “No. YOU are lax, be thorough” Consider [K1, X1] and [K2, X2] with K1 = {Flexible ⇒ ¬Lax, ¬Rigid ⇒ Flexible} X1 = {Rigid} K2 = {Thorough ⇒ ¬Rigid, ¬Lax ⇒ Thorough} X2 = {Lax} Answer of 2 to Rigid is A = {¬Lax} ∪ K2 (with K2 ⊢{⊥,Lax} Rigid)

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 9 / 29

slide-31
SLIDE 31

Example: the deafs dialogue

1: “You are rigid, be flexible” 2: “No. YOU are lax, be thorough” Consider [K1, X1] and [K2, X2] with K1 = {Flexible ⇒ ¬Lax, ¬Rigid ⇒ Flexible} X1 = {Rigid} K2 = {Thorough ⇒ ¬Rigid, ¬Lax ⇒ Thorough} X2 = {Lax} Answer of 2 to Rigid is A = {¬Lax} ∪ K2 (with K2 ⊢{⊥,Lax} Rigid) Counter-argument from 2 against Rigid: A, ¬Rigid, which deductively amounts to: A, Thorough

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 9 / 29

slide-32
SLIDE 32

Arguments and attitudes

Properties If [K, X] is against a subset of supp(α), it can construct at least one argument attacking α; for a subset of supp(α), it can construct at least one argument defending α; puzzled by a subset of supp(α), it can construct at least one argument both attacking and defending α; neutral about a subset of the support of supp(α), then it has no argument to give about α.

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 10 / 29

slide-33
SLIDE 33

Reasoning about attitudes...

Properties [K, X] is for f iff it is against ¬f , [K, X] is neutral about f iff it is neutral about ¬f , [K, X] is puzzled about f iff it is puzzled about ¬f , [K, X] is for the tautologies, [K, X] is against the contradictions. If (for instance) [K, X] is for f , and for g, which attitude will it be able to adopt regarding f and g, f or g. . . ?

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 11 / 29

slide-34
SLIDE 34

Example: combining attitudes

Consider [K, X], with K = {Inflation ⇒ ¬IncreasingPurchasingPower, IncreasingSalaries ⇒ IncreasingPurchasingPower , FixingBasicPrices ⇒ IncreasinPurchasingPower, IncreasingSalaries ∧ FixingBasicPrices ⇒ Inflation} X = {¬IncreasingSalaries, ¬FixingBasicPrices}

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 12 / 29

slide-35
SLIDE 35

Example: combining attitudes

Consider [K, X], with K = {Inflation ⇒ ¬IncreasingPurchasingPower, IncreasingSalaries ⇒ IncreasingPurchasingPower , FixingBasicPrices ⇒ IncreasinPurchasingPower, IncreasingSalaries ∧ FixingBasicPrices ⇒ Inflation} X = {¬IncreasingSalaries, ¬FixingBasicPrices} K is both for IncreasingSalaries and for FixingBasicPrices

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 12 / 29

slide-36
SLIDE 36

Example: combining attitudes

Consider [K, X], with K = {Inflation ⇒ ¬IncreasingPurchasingPower, IncreasingSalaries ⇒ IncreasingPurchasingPower , FixingBasicPrices ⇒ IncreasinPurchasingPower, IncreasingSalaries ∧ FixingBasicPrices ⇒ Inflation} X = {¬IncreasingSalaries, ¬FixingBasicPrices} K is both for IncreasingSalaries and for FixingBasicPrices K is against IncreasingSalaries ∧ FixingBasicPrices

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 12 / 29

slide-37
SLIDE 37

Valuation

Consider the valuation v ⊢X

4

: A × P − → FOUR such that ∀A ∈ A, ∀p ∈ P,

v ⊢X

4 (A, p) = 1

iff A is for p v ⊢X

4 (A, p) = ⊤

iff A is neutral regarding p v ⊢X

4 (A, p) = ⊥

iff A is puzzled regarding p v ⊢X

4 (A, p) = 0

iff A is against p

v ⊢X

4

is not functional... Two choices:

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 13 / 29

slide-38
SLIDE 38

Valuation

Consider the valuation v ⊢X

4

: A × P − → FOUR such that ∀A ∈ A, ∀p ∈ P,

v ⊢X

4 (A, p) = 1

iff A is for p v ⊢X

4 (A, p) = ⊤

iff A is neutral regarding p v ⊢X

4 (A, p) = ⊥

iff A is puzzled regarding p v ⊢X

4 (A, p) = 0

iff A is against p

v ⊢X

4

is not functional... Two choices:

1

adding constraints to X in order to determinize in an unique way the admissibility associated with the different logical combinations of two formulas;

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 13 / 29

slide-39
SLIDE 39

Valuation

Consider the valuation v ⊢X

4

: A × P − → FOUR such that ∀A ∈ A, ∀p ∈ P,

v ⊢X

4 (A, p) = 1

iff A is for p v ⊢X

4 (A, p) = ⊤

iff A is neutral regarding p v ⊢X

4 (A, p) = ⊥

iff A is puzzled regarding p v ⊢X

4 (A, p) = 0

iff A is against p

v ⊢X

4

is not functional... Two choices:

1

adding constraints to X in order to determinize in an unique way the admissibility associated with the different logical combinations of two formulas;

2

extending our valuation to sets of truth values → non-deterministic multi-valued logics

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 13 / 29

slide-40
SLIDE 40

Nmatrices [Avron et Al.]

The Nmatrice MLA associated to FOUR is a triple M = V, D, O with

1

V = {O, ⊥, ⊤, 1}, set of truth values;

2

D = {1, ⊤}, set of designated valued,

3

Non-designated values: N = V \ D ;

4

O = {¬, ∨, ∧}, set of operators whose behaviour is described by the corresponding truth tables:

α ¬α 1 {0} ⊤ {⊤} ⊥ {⊥} {1} α ∧ β 1 ⊤ ⊥ 1 {1, ⊤, ⊥, 0} {⊤, 0} {⊥, 0} {0} ⊤ {⊤, 0} {⊤, 0} {0} {0} ⊥ {⊥, 0} {0} {⊥, 0} {0} {0} {0} {0} {0} α ∨ β 1 ⊤ ⊥ 1 {1} {1} {1} {1} ⊤ {1} {1, ⊤} {1} {1, ⊤} ⊥ {1} {1} {1, ⊥} {⊥, 1} {1} {1, ⊤} {1, ⊥} {1, ⊤, ⊥, 0}

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 14 / 29

slide-41
SLIDE 41

Nsequent

Multivalued Sequent Sequent in a matrix M = set of signed formulas The classical sequent Γ ⇒ ∆ is interpreted by {0 : Γ} ∪ {1 : ∆} o` u V = {0, 1} et D = {1} Conventions : V = {t0, . . . , tn−1} (with n ≥ 2) and D = {td, . . . , tn−1} (with d ≥ 1) Definition A n–sequent on a language L is an expression Σ of the form Γ0|Γ1| . . . |Γn−1 where for every 0 ≤ i ≤ n − 1, Γi is a finite set of formulas

  • n L.

Notation Replace | by ⇒: Γi1| . . . Γir ⇒ Γj1| . . . |Γjs where i1, . . . , ir ∈ N and j1, . . . , js ∈ D

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 15 / 29

slide-42
SLIDE 42

Multivalued sequents for LA

Axioms: any set of signed formulas of the form {a : ϕ | a ∈ V, ϕ ∈ F} Structural rules: weakening Logical rules (after simplification), e.g.: Conjonction:

Ω,⊥:ϕ,1:ϕ Ω,⊥:ψ,1:ψ Ω,⊥:ϕ,ψ Ω,⊥:ϕ∧ψ,0:ϕ∧ψ Ω,1:ϕ,⊤:ϕ Ω,1:ψ,⊤:ψ Ω,⊤:ϕ,ψ Ω,0:ϕ∧ψ,⊤:ϕ∧ψ Ω,⊥:ϕ,ψ,0:ϕ,ψ Ω,0:ϕ,ψ,⊤:ϕ,ψ Ω,0:ϕ∧ψ

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 16 / 29

slide-43
SLIDE 43

Transformation into Nsequents

Example: Conjonction:

Γ⊥,ϕ|Γ0⇒Γ1,ϕ|Γ⊤ Γ⊥,ψ|Γ0⇒Γ1,ψ|Γ⊤ Γ⊥,ϕ,ψ|Γ0⇒Γ1|Γ⊤ Γ⊥,ϕ∧ψ|Γ0,ϕ∧ψ⇒Γ1|Γ⊤ Γ⊥|Γ0⇒Γ1,ϕ|Γ⊤,ϕ Γ⊥|Γ0⇒Γ1,ψ|Γ⊤,ψ Γ⊥|Γ0⇒Γ1|Γ⊤,ϕ,ψ Γ⊥|Γ0,ϕ∧ψ⇒Γ1|Γ⊤,ϕ∧ψ Γ⊥,ϕ,ψ|Γ0,ϕ,ψ⇒Γ1|Γ⊤ Γ⊥|Γ0,ϕ,ψ⇒Γ1|Γ⊤,ϕ,ψ Γ⊥|Γ0,ϕ∧ψ⇒Γ1|Γ⊤

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 17 / 29

slide-44
SLIDE 44

Expressiveness condition

Condition(Avron, Ben–naim, Konikowska, 07): a n–sequent calculus can be translated into a two–sided sequent calculus only if the underlying langage is sufficiently expressive for the semantics induced by the Nmatrix M Intuition: for any valuation of an initial formula, by introducing new formulas compounded only from the initial formula with any connector, one can still adress any subsequent valuation of these new formulas either in N or D This ensures (in a strong combinatoric way) the partition of any multi-valued sequent into a two-valued sequent

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 18 / 29

slide-45
SLIDE 45

partition sequence

Σ = Γ1 | Γ2 | . . . | Γn a n–sequent de L. Definition (Avron, Ben–naim, Konikowska, 07) A partition sequence for Σ is a tuple π = π1, . . . , πn such that for 1 ≤ i ≤ n, πi is a partition of Γi of the form πi = {Γ

ij | 1 ≤ j ≤ li} ∪ {Γ

′′

ik | 1 ≤ k ≤ mi}

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 19 / 29

slide-46
SLIDE 46

Nsequents and classical sequents

For a partition sequence π and for all 1 ≤ i ≤ n, define: ∆

i =

  • {Ai

j(Γ

ij) | 1 ≤ j ≤ li}

′′

i =

  • {Bi

k(Γ

′′

ik) | 1 ≤ k ≤ mi}

Σπ = ∆

1, ∆

2, . . . , ∆

n ⇒ ∆

′′

1, ∆

′′

2, . . . , ∆

′′

n

where Ai

j(Γ

ij) = {Ai jϕ | ϕ ∈ Γ

ij} and Bi k(Γ

′′

ik) is defined in the same way.

Let Π be the set of all partition sequences of Σ, the set of two-sided sequents generated by TWO(Σ) = {Σπ | π ∈ Π} Theorem (Avron, Ben–naim, Konikowska, 07) If L is sufficiently expressive langage for every n-sequent Σ = Γ1 | Γ2 | . . . | Γn, and any valuation v of formulas of L, v | = Σ iff v | = Σ′ for every Σ′ ∈ TWO(Σ).

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 20 / 29

slide-47
SLIDE 47

Preliminary results

Theorem For every agent A and every formula α:

v ⊢X

4 (A, α) = 1 iff v ⊢X 4 (A, α) ∈ D and v ⊢X 4 (A, ¬α) ∈ N

v ⊢X

4 (A, α) = ⊤ iff v ⊢X 4 (A, α) ∈ D and v ⊢X 4 (A, ¬α) ∈ D

v ⊢X

4 (A, α) = ⊥ iff v ⊢X 4 (A, α) ∈ N and v ⊢X 4 (A, ¬α) ∈ N

v ⊢X

4 (A, α) = 0 iff v ⊢X 4 (A, α) ∈ N and v ⊢X 4 (A, ¬α) ∈ D

→ ensures the two-sided partition of every sequent Σ = Γ1 | Γ2 | . . . | Γn

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 21 / 29

slide-48
SLIDE 48

Calculus SLA

Axioms: ϕ ⇒ ϕ for every formula ϕ Rules:

Γ⇒∆,ϕ,ψ Γ⇒∆,ϕ∨ψ Γ,ϕ,ψ⇒∆ Γ,ϕ∧ψ⇒∆ Γ,¬ϕ,¬ψ⇒∆ Γ,¬(ϕ∨ψ)⇒∆ Γ⇒∆,¬ϕ,¬ψ Γ⇒∆,¬(ϕ∧ψ) Γ⇒∆,ϕ Γ⇒∆,¬¬ϕ Γ,ϕ⇒∆ Γ,¬¬ϕ⇒∆

SLA gets only one of the two rules for disjonction and only one of the two rules for conjonction None of the classical rules for negation The calculus is symetrical

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 22 / 29

slide-49
SLIDE 49

MSPL (Avron, Ben–naim, Konikowska, 07)

Given a set S of sources of information and a processor P each source s ∈ S can tell if a formula φ is true, if it is false or if it has no information about φ The processor P collects the formulas and combines them from the informations given by the sources :

it has information that φ is true but no information that φ is false it has information that φ is false but no information that φ is true it has both informations that φ is true and that φ is false it has no information on φ at all

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 23 / 29

slide-50
SLIDE 50

MSPL (Avron, Ben–naim, Konikowska, 07)

The Nmatrice MSPL associated to FOUR is M = V, D, O with

1

V = {O, ⊥, ⊤, 1}, set of truth values;

2

D = {1, ⊤}, the designated values,

3

N = V \ D ;

4

O = {¬, ∨, ∧} the set of operators described by the following truth-tables:

α ¬α 1 {0} ⊤ {⊤} ⊥ {⊥} {1} α ∧ β 1 ⊤ ⊥ 1 {1, ⊤} {⊤} {⊥, 0} {0} ⊤ {⊤} {⊤} {0} {0} ⊥ {⊥, 0} {0} {⊥, 0} {0} {0} {0} {0} {0} α ∨ β 1 ⊤ ⊥ 1 {1} {1} {1} {1} ⊤ {1} {⊤} {1} {⊤} ⊥ {1} {1} {1, ⊥} {⊥, 1} {1} {⊤} {1, ⊥} {⊤, 0}

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 24 / 29

slide-51
SLIDE 51

LA vs. MSPL

LA

Γ⇒∆,ϕ,ψ Γ⇒∆,ϕ∨ψ Γ,ϕ,ψ⇒∆ Γ,ϕ∧ψ⇒∆ Γ,¬ϕ,¬ψ⇒∆ Γ,¬(ϕ∨ψ)⇒∆ Γ⇒∆,¬ϕ,¬ψ Γ⇒∆,¬(ϕ∧ψ) Γ⇒∆,ϕ Γ⇒∆,¬¬ϕ Γ,ϕ⇒∆ Γ,¬¬ϕ⇒∆

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 25 / 29

slide-52
SLIDE 52

LA vs. MSPL

LA

Γ⇒∆,ϕ,ψ Γ⇒∆,ϕ∨ψ Γ,ϕ,ψ⇒∆ Γ,ϕ∧ψ⇒∆ Γ,¬ϕ,¬ψ⇒∆ Γ,¬(ϕ∨ψ)⇒∆ Γ⇒∆,¬ϕ,¬ψ Γ⇒∆,¬(ϕ∧ψ) Γ⇒∆,ϕ Γ⇒∆,¬¬ϕ Γ,ϕ⇒∆ Γ,¬¬ϕ⇒∆

MSPL = LA +

Γ⇒∆¬ϕ Γ⇒∆¬ψ Γ⇒∆,¬(ϕ∨ψ) Γ⇒∆ϕ Γ⇒∆ψ Γ⇒∆,ϕ∧ψ

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 25 / 29

slide-53
SLIDE 53

Back to cumulativity

Consider (roughly) X ⇌ X ′ iff (⊢X⇔⊢X ′) and ˆ X the least representative of the corresponding equivalence class Define S∧ as the (And)-closure of S If ( ˆ X c)∧ = ˆ X c then

Conjunctive Cautions Monotony holds; LA turns into MSPL.

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 26 / 29

slide-54
SLIDE 54

Epilog

A non-deterministic multivalued calculus with four truth values Describes how an agent evaluates a compound formula from its elementary attitudes A generic ’classical’ calculus SLA Describes how an agent admits a compound formula from the admissibility of its subformulas Since SLA relies on the only distinction between designated and non-designated values, it amounts to the common behaviour of all the agents horizon: investigate more accurately the role of MSPL in argumentation Far horizon: relating the reasoning of an agent with strategies of construction of new arguments Complementary direction: how additional constraints on X can determinize the connectors

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 27 / 29

slide-55
SLIDE 55

Acknowledgements

We thank Geoffroy Aubry for some previous work on attitudes, and Michel Klein for his collaboration on the LA calculus.

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 28 / 29

slide-56
SLIDE 56

References

[1] Aubry G., Risch V., 2005, Toward a Logical Tool for Generating New Arguments in an Argumentation-Based Framework. ICTAI 2005, p. 599–603. [2] Aubry G., Risch V., 2006, Managing Deceitful Arguments with X-logics. ICTAI 2006, p. 216–219. [3] Avron A., Ben-Naim J., Konikowska B., 2007, Cut-Free Ordinary Sequent Calculi for Logics Having Generalized Finite-Valued

  • Semantics. Logica Universalis 1 (2007), p. 41–70.

[4] Forget L., Risch V., Siegel P., 2001, Preferential Logics Are X-logics. Journal of Logic and Computation, Vol. 11, No. 1, february 2001, p. 71–83.

  • V. Risch (LSIS)

Argumentation, NMatrices, X–logics BRA’15 29 / 29