17.01.2014 Views

Propositional Argumentation Systems and Symbolic Evidence Theory

Propositional Argumentation Systems and Symbolic Evidence Theory

Propositional Argumentation Systems and Symbolic Evidence Theory

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

40 4 ASSUMPTION–BASED SYSTEMS<br />

Formulae R 1 to R 6 can easily be transformed into a set Σ = {ξ 1 , . . . , ξ 10 } of clauses.<br />

This forms the entire knowledge base that can be used to solve the diagnostics<br />

problem (see Example 4.4 of Section 4.3).<br />

ξ 1 : ∼ok 1 ∨ ∼in ∨ ∼v ξ 6 : ∼ok 2 ∨ ∼v ∨ ∼out<br />

ξ 2 : ∼ok 1 ∨ in ∨ v ξ 7 : ∼ok 2 ∨ v ∨ out<br />

ξ 3 : ok 1 ∨ ∼in ∨ v ξ 8 : ok 2 ∨ ∼v ∨ out<br />

ξ 4 : ok 1 ∨ in ∨ ∼v ξ 9 : ok 2 ∨ v ∨ ∼out<br />

ξ 5 : ∼in ξ 10 : out<br />

4.3 Hypotheses <strong>and</strong> Arguments in Assumption–Based <strong>Systems</strong><br />

Given an assumption–based system A = (P, A, Σ), we may be interested in<br />

certain hypotheses h about the knowledge contained in A. Such hypotheses<br />

can be expressed as logical formulae in L N . What can we learn from Σ about<br />

the possible truth of h? If some assumptions are considered to be either true<br />

or false, then h can possibly be deduced from Σ. Such combinations of true<br />

<strong>and</strong> false assumptions or, more generally, logical formulae which are only<br />

composed by assumptions can be regarded as arguments in favour of h.<br />

This is the link to the general evidence theory of Chapter 2: the possible<br />

hypotheses are logical formulae in L N , <strong>and</strong> the possible arguments are logical<br />

formulae in L A . More precisely, the finite Boolean algebras H of hypotheses<br />

<strong>and</strong> A of arguments correspond to the sets [L N ] <strong>and</strong> [L A ] of equivalence<br />

classes of formulae in L N <strong>and</strong> L A respectively.<br />

In view of these remarks let us consider logical formulae a ∈ L A , such that<br />

a ∧ ξ |= h, where ξ is the conjunction of the clauses contained in Σ, <strong>and</strong><br />

h ∈ L N represents a hypothesis. Such formulae a allow to deduce h from<br />

Σ. They are called supporting arguments for h relative to Σ. If a is the<br />

coarsest (least precise) supporting argument of h, i.e. if there is no other<br />

supporting argument a ′ of h with a ′ |= a <strong>and</strong> a ′ ≠ a, then the equivalence<br />

class [a] of a is called quasi–support of h relative to Σ, denoted by qs[h, Σ]<br />

(instead of [qs(h, Σ)]). It can be shown (Kohlas & Monney, 1995) that<br />

quasi–support satisfies the conditions (Q1) to (Q4) from Chapter 2,<br />

(Q1) qs[⊤, Σ] = [⊤],<br />

(Q2) qs[h 1 ∧ h 2 , Σ] = qs[h 1 , Σ] ∧ qs[h 2 , Σ],<br />

(Q3) if h 1 |= h 2 , then qs[h 1 , Σ] |= qs[h 2 , Σ],

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!