• No results found

Rules for exceptions [conditional defaults]

Definition B.10 LetAandLA0 be as earlier. The languageLA3 hasAas its non-logical vocabulary, and in its logical vocabulary one additional binary operator and one additional unary operator

presumable. A string of symbols φ is a sentence ofLA3 iff there are sentences ψ andχ of LA0 such thatφ=ψ, or φ=ψ χ, or φ= presumablyψ.

Definition B.11 (i) Let W be as before. A frame onW is a function π assigning to every subset dof W a pattern πd ond.

(ii) Let π be a frame on W and d, e ⊆W. The proposition e is a default in πd iff d∩e6= ∅ and πd·e=πd.

Definition B.12 Let π be a frame onW, and d⊆W.

(i) wis a normal world inπd iffw∈dand for every d0⊆dsuch thatw∈d0 it holds thatw≤πd0 v

for everyv∈d0;

(ii) nπd is the set of all normal worlds in πd;

(iii) π is coherent iff for every non-empty d⊆W, nπd6=∅.

Definition B.13 Let W be as before.

(i) σ is an information state iffσ =hπ, si and one of the following conditions is fulfilled: (a) π is a coherent frame onW, and sis a non-empty subset of W;

(b) π is the frame hι,∅i, where ιd={hw, wi|w∈d} for everyd⊆W. (ii) 0=hυ, Wi, where υd=d×d for everyd⊆W.

1=hι,∅i.

(iii) Let σ=hπ, si and σ0 =hπ0, s0i be states.

Let π00 be the frame such that for every d, π00d=πd∩π0d. Then σ+σ0 =hπ00, s∩s0i, if hπ00, s∩s0i is coherent;

σ+σ0 =1, otherwise.

Definition B.14

(i) Let π and π0 be frames, both based on W.

The frame π is a refinement of π0 iff πd⊆π0d for everyd⊆W. (ii) Let π be a frame and d, e⊆W. πd·e is the refinement of π given by

(a) if d0 =d, then πd·ed0=πd0;

(b) πd·ed=πd·e.

The frame πd·e is the result of refiningπd inπ withe.

Definition B.15 Let σ=hπ, si be an information state.

·σ[φ ψ] =1if kφk ∩ kψk=∅ or πkφk·kψk is incoherent.

Definition B.16 Let σ = hπ, si be a coherent information state and assume that e1, . . . , en are

defaults in πd1, . . . , πdn respectively.

(i) A world wcomplies with {e1, . . . , en} iff w∈ei for every isuch that w∈di (1≤i≤n).

(ii) The set of defaults {e1, . . . , en} applies withinsiff for every d⊇sthere is somew∈nπd such

thatw complies with {e1, . . . , en}. In this case, we also say thate1, . . . , en jointly apply withins.

Definition B.17 Let σ = hπ, si be a coherent information state and assume that e1, . . . , en are

defaults in πd1, . . . , πdn.

(i) Then {e1, . . . , en} is a maximal applicable set in σ iff e1, . . . , e2 jointly apply within s, and for

every en+1 and dn+1 such thaten+1 is a default in πdn+1, and e1, . . . , en+1 jointly apply within s

it holds that en+1 =ei anddn+1=di for some i≤n.

(ii) A world w is optimal inσ iff w∈sand w complies with a maximal applicable set of defaults. The set of optimal worlds is denoted by mσ.

(iii) σ[presumably ψ]is determined as follows:

· If mσ∩ kψk=mσ, then σ[presumably ψ] =σ.

C

Recap: The Dynamic-Epistemic Upgrade Logic

The Dynamic-Epistemic Upgrade Logic DEUL is introduced in the paper Dynamic Logic for Prefer- ence Upgrade[10]. We will use the version with world-eliminating announcement !ϕand use slightly different notation.

Outside this appendix we will also omit the subscripts. These are used in DEUL to deal with multi-agent situations, but for our purposes we can do fine effectively having only one of them.

Definition C.1 (Language) Take a set of propositional variables P and a set of agents I, with p ranging over P andi over I. The dynamic epistemic preference language is given by:

ϕ ::= ⊥|p|¬ϕ|ϕ∧ψ|iϕ|i ϕ|U ϕ|hπiϕ π ::= !ϕ|#ϕ

(Furthermore we will use ∨, ♦i, ♦≤i and E being the duals of∧, i,

i and U, respectively.)

Definition C.2 (Models) An epistemic preference model is a tuple M = (S,{∼i |i∈I},{≤i |i∈I}, V), withS a set of possible worlds,∼ithe usual equivalence relation of epistemic accessibility

for agent i, and V a valuation for proposition letters. Moreover, ≤i is a reflexive and transitive relation over the worlds.

Definition C.3 (Model Upgrades) Let M = (S,{∼i |i∈I},{≤i |i∈ I}, V). Then the upgraded

models M!ϕ andM#ϕ are defined as follows:

M!ϕ := (S0,{∼i |i∈I} ∩(S0×S0),{≤i |i∈I} ∩(S0×S0), V0), where

S0 = {w∈S|M, w|=ϕ}

V0(p) = V(p)∩S0

M#ϕ := (S,{∼i |i∈I},{≤∗i |i∈I}, V), where

≤∗i = ≤i −{(s, t)|M, s|=ϕand M, t|=¬ϕ}

Definition C.4 (Semantics) Given an epistemic preference model M = (S,{∼i |i∈ I},{≤i |i∈

ϕ:

M, s|=p iff s∈V(p) (1)

M, s|=¬ϕ iff notM, s|=ϕ (2)

M, s|=ϕ∧ψ iff M, s|=ϕandM, s|=ψ (3)

M, s|=iϕ iff for allt:s∼i timpliesM, t|=ϕ (4)

M, s|=≤i ϕ iff for allt:s≤i timpliesM, t|=ϕ (5)

M, s|=U ϕ iff for allt:M, t|=ϕ (6)

M, s|=hπiϕ iff Mπ, s|=ϕ (7)

(With convention that if sis eliminated by the updateπ then it is not the case thatMπ, s|=ϕ.)

One of the interesting things aboutDEU Lis the fact it hasreduction axioms eliminating the model- upgrade modalities, making it self-embed into regular modal logic. For more on the importance of these and what to do with them, consult [10].

Theorem C.5 (Soundness) The following formulas are valid:

h!ϕip ↔ ϕ∧p h!ϕi¬ψ ↔ ϕ∧ ¬h!ϕiψ h!ϕi(ψ∧φ) ↔ h!ϕiψ∧ h!ϕiφ h!ϕi♦iψ ↔ ϕ∧♦ih!ϕiψ h!ϕii ψ ↔ ϕ∧i h!ϕiψ h!ϕiEψ ↔ ϕ∧Eh!ϕiψ h#ϕip ↔ p h#ϕi¬ψ ↔ ¬h#ϕiψ h#ϕi(ψ∧φ) ↔ h#ϕiψ∧ h#ϕiφ h!ϕi♦iψ ↔ ♦ih#ϕiψ h#ϕii ψ ↔ (¬ϕ∧i h#ϕiψ)∨(♦i≤(ϕ∧ h!ϕiψ)) h!ϕiEψ ↔ Eh!ϕiψ

D

Recap: Circumscription

This definition taken directly from Lifschitz ([5]), the simplest form of circumscription is as follows: First off, for any predicates of the same arity,

P =Q stands for ∀x(P(x)≡Q(x))

P ≤Q stands for ∀x(P(x)⊇Q(x))

P < Q stands for (P ≤Q)∧ ¬(P =Q)

Now if A(P) is a sentence containing the predicate constant P,the circumscription of P in A(P) is simply the following second-order sentence:

A(P)∧ ¬∃p[A(p)∧p < P]

This added clause in this sentence can be seen as meaning that you cannot reduce the extension of

P while retaining the truth ofA(P). However, it is problematic in only talking about reducing the extension ofP while leaving all other predicates the same.

Hence there is also such a thing asthe circumscription ofP in A(P) with varied Z1, . . . , Zn, which is the following second-order sentence:

A(P, Z1, . . . , Zn)∧ ¬∃p, z1, . . . , zn[A(p, z1, . . . , zn)∧p < P]

A circumscription can be further circumscribed in, and varying a predicate that was circumscribed before can be used as a prioritizing tool.

Examples involving default rules are typically handled using abnormality predicates which are circumscribed against while varying the other predicates.

References

[1] N. Asher and M. Morreau. Commonsense entailment: A modal theory of nonmonotonic reasoning. In J. van Eijck, editor,Logics in AI: Proc. of the European Workshop JELIA’90, pages 1–30. Springer, Berlin, Heidelberg, 1991.

[2] M. L. Ginsberg, editor. Readings in Nonmonotonic Reasoning. Morgan Kaufmann, Los Altos, CA, 1987.

[3] J. Groenendijk and M. Stokhof. Two theories of dynamic semantics. In J. van Eijck, edi- tor,Logics in AI: Proc. of the European Workshop JELIA’90, pages 55–64. Springer, Berlin, Heidelberg, 1991.

[4] V. Lifschitz. Benchmark problems for formal nonmonotonic reasoning, version 2.00. In M. Re- infrank, J. de Kleer, M. L. Ginsberg, and E. Sandewall, editors, Non-Monotonic Reasoning: 2nd International Workshop, Grassau, Germany, pages 202–219. Springer, Berlin, Heidelberg, 1989.

[5] Vladimir Lifschitz. Circumscription. In Dov Gabbay, Christopher J. Hogger, and J. A. Robin- son, editors, Handbook of Logic in Artificial Intelligence and Logic Programming, Volume 3: Nonmonotonic Reasoning and Uncertain Reasoning, pages 298–352. Oxford University Press, 1994.

[6] Aarati Martino. Formalizing Elaboration Tolerance. PhD thesis, Stanford University, August 2003.

[7] J. McCarthy. Applications of circumscription to formalizing common-sense knowledge. In M. L. Ginsberg, editor,Readings in Nonmonotonic Reasoning, pages 153–166. Kaufmann, Los Altos, CA, 1987.

[8] J. McCarthy. Circumscription: A form of non-monotonic reasoning. In M. L. Ginsberg, editor,

Readings in Nonmonotonic Reasoning, pages 145–151. Kaufmann, Los Altos, CA, 1987. [9] Johan van Benthem. Situation calculus meets modal logic.

[10] Johan van Benthem and Fenrong Liu. Dynamic logic of preference upgrade.Journal of Applied Non-Classical Logic, 17(2):157–182, 2007. forthcoming.

[11] Frank Veltman. Defaults in update semantics. Journal of Philosophical Logic, 25(3):221–261, 1996.

[12] Gerard Vreeswijk. Interpolation of benchmark problems in defeasible reasoning. InWOCFAI, pages 453–468, 1995.

Related documents