Table Of ContentCognitive Technologies
ManagingEditors: D.M.Gabbay J.Siekmann
Editorial Board: A.Bundy J.G.Carbonell
M.Pinkal H.Uszkoreit M.Veloso W.Wahlster
M.J.Wooldridge
AdvisoryBoard:
LuigiaCarlucciAiello AlanMackworth
FranzBaader MarkMaybury
WolfgangBibel TomMitchell
LeonardBolc JohannaD.Moore
CraigBoutilier StephenH.Muggleton
RonBrachman BernhardNebel
BruceG.Buchanan SharonOviatt
AnthonyCohn LuisPereira
Arturd’AvilaGarcez LuRuqian
LuisFarin˜asdelCerro StuartRussell
KoichiFurukawa ErikSandewall
GeorgGottlob LucSteels
PatrickJ.Hayes OlivieroStock
JamesA.Hendler PeterStone
AnthonyJameson GerhardStrube
NickJennings KatiaSycara
AravindK.Joshi MilindTambe
HansKamp HidehikoTanaka
MartinKay SebastianThrun
HiroakiKitano JunichiTsujii
RobertKowalski KurtVanLehn
SaritKraus AndreiVoronkov
MaurizioLenzerini TobyWalsh
HectorLevesque BonnieWebber
JohnLloyd
Forfurthervolumes:
http://www.springer.com/series/5216
· ·
Dov M. Gabbay Odinaldo T. Rodrigues
Alessandra Russo
Revision, Acceptability
and Context
Theoretical and Algorithmic Aspects
123
Prof.Dr.DovM.Gabbay Dr.OdinaldoT.Rodrigues
King’sCollegeLondon King’sCollegeLondon
Dept.ComputerScience Dept.ComputerScience
LondonWC2R2LS LondonWC2R2LS
UnitedKingdom UnitedKingdom
[email protected] [email protected]
Dr.AlessandraRusso
ImperialCollegeLondon
Dept.Computing
180Queen’sGate
LondonSW72BZ
UnitedKingdom
[email protected]
ManagingEditors
Prof.Dr.DovM.Gabbay Prof.Dr.Jo¨rgSiekmann
AugustusDeMorganProfessorofLogic ForschungsbereichDeduktions-und
King’sCollegeLondon Multiagentensysteme,DFKI
Dept.ComputerScience Stuhlsatzenweg3,Geb.43
LondonWC2R2LS 66123Saarbru¨cken,Germany
UnitedKingdom
CognitiveTechnologiesISSN1611-2482
ISBN978-3-642-14158-4 e-ISBN978-3-642-14159-1
DOI10.1007/978-3-642-14159-1
SpringerHeidelbergDordrechtLondonNewYork
LibraryofCongressControlNumber:2010932026
ACMComputingClassification(1998):I.2,F.4,H.2
(cid:2)c Springer-VerlagBerlinHeidelberg2010
Thisworkissubjecttocopyright.Allrightsarereserved,whetherthewholeorpartofthematerialisconcerned,
specificallytherightsoftranslation,reprinting,reuseofillustrations,recitation,broadcasting,reproductionon
microfilmorinanyotherway,andstorageindatabanks.Duplicationofthispublicationorpartsthereofis
permittedonlyundertheprovisionsoftheGermanCopyrightLawofSeptember9,1965,initscurrentversion,
andpermissionforusemustalwaysbeobtainedfromSpringer.Violationsareliabletoprosecutionunderthe
GermanCopyrightLaw.
Theuseofgeneraldescriptivenames,registerednames,trademarks,etc.inthispublicationdoesnotimply,
evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevantprotectivelawsand
regulationsandthereforefreeforgeneraluse.
Coverdesign:Ku¨nkelLopkaGmbH,Heidelberg
Printedonacid-freepaper
SpringerispartofSpringerScience+BusinessMedia(www.springer.com)
Preface
We are happy to present our new book to the community. Although this is a book
on revision theory, it uses general methodologies which are applied in other areas
oflogic.Wewouldliketohighlightthesemethodologiesinthispreface.
1. theideaoflogicbytranslation
GivenanewlogicalsystemL,whosepropertieswewanttoinvestigate,wetrans-
lateLintoawell-knownlogic,usuallyclassicallogicC,modallogicorintuition-
isticlogic.Westudythepropertieswewantinthetargetlogicandthentranslate
back into L. Our approach of revision by translation (see Chapter 7) is an in-
stanceofthatmethodology.Wecanalsodointerpolationbytranslation.Whatis
mostwellknowninthecommunityissemanticsbytranslationanddecidability
andundecidabilitybytranslation.Thisisacommonwayofobtainingsemantics
ordecidabilityresultsforlogicalsystems.
2. meta-level/object-levelmovement
Given a logic L,there may be some manipulations of L done in the meta-level.
Such notions can be provability or consequence, e.g., A(cid:2)B or deletion from a
database.Wemaywishtobringthesenotionstotheobjectlevel.Thiscanbea
complicatedtask.Weallknowabouttheproblemofbringingtheproofpredicate
intotheobjectlevel.Inthisbookwebringintheoperationofdeletion.Another
lesser known result is the bringing in to the object level of a non-monotonic
consequencerelation|∼,andwhatwegetwhenwedoitcorrectlyisaconditional
intheobjectlanguage.
3. context
The use of context can determine how we use our logic. The context is a label
inLabelledDeductiveSystemsanditisaveryrichlabelinsituationsemantics.
Inrevisiontheory,itaffectsthealgorithmsweusetorevise,andinmanycasesit
mighttellusnottoreviseatallandinsteadtotrytoobtainmoredataorsimply
livewiththeinconsistencyinacarefulway.
v
vi Preface
Acknowledgements We would like to acknowledge feedback and collaborations with Krysia
Broda, Fa´tima Cristina Canazaro Dargam, Artur d’Avila Garcez, Anthony Hunter, Lu´ıs Lamb,
Je´roˆmeLang,DavidMakinson,SanjayModgil,GabriellaPigozzi,KarlSchlechtaandJohnWoods.
ThismanuscriptwaswrittenusingLATEX.WeusedPaulTaylor’scommutativediagramspack-
ageforsomeofthediagrams.JaneSpurrkindlyhelpedinthesolutionofmanyformattingissues.
London, D.M.Gabbay
September2009 O.T.Rodrigues
A.Russo
Contents
1 BackgroundandOverview ...................................... 1
1.1 IntroductoryDiscussion ..................................... 1
1.2 Focusingon‘Acceptability’Ratherthan‘Inconsistency’ .......... 5
1.3 OverviewofThisBook...................................... 7
1.4 NotationUsedinThisBook.................................. 8
1.5 BasicMathematicalNotions ................................. 10
References..................................................... 11
2 IntroducingRevisionTheory .................................... 13
2.1 AGMBeliefRevision ....................................... 13
2.2 KatsunoandMendelzon’sCharacterisation ..................... 20
2.3 CounterfactualStatementsandtheRamseyTest ................. 21
2.4 Grove’sSystemsofSpheres.................................. 25
2.5 EpistemicEntrenchment..................................... 28
2.6 Discussion ................................................ 30
2.7 ActionUpdates ............................................ 30
2.8 GeneralisingtheConceptofRevision.......................... 33
2.9 IteratingtheRevisionOperation .............................. 37
2.9.1 DarwicheandPearl’sPostulatesforIteratedRevisions ..... 37
2.9.2 Lehmann’sApproach:BeliefRevision,Revised........... 40
2.10 CompromiseRevision....................................... 41
2.11 ControlledRevision ........................................ 45
2.12 RevisionbyTranslation ..................................... 47
2.13 AGeneralSettingforAlgorithmicRevision .................... 49
2.14 OutlineofThisBook ....................................... 52
References..................................................... 52
3 StepwiseRevisionOperations ................................... 55
3.1 Introduction ............................................... 55
3.2 QuantitativeMeasurementofChange.......................... 55
3.2.1 MinimalChangeandtheFunctiond .................... 56
vii
viii Contents
3.3 QualitativeMeasurementsofChange .......................... 63
3.4 RepresentationIssues ....................................... 65
3.5 RevisionofFormulae ....................................... 67
3.6 PropertiesoftheRevisionOperator◦ ......................... 70
r
3.7 OtherBeliefChangeOperators ............................... 77
3.7.1 BeliefContraction ................................... 77
3.7.2 ConsolidatingInformationontheBeliefBase ............ 82
3.8 ComparisonwithOtherBeliefChangeOperators ................ 85
3.9 OperatorsforReasoningAbouttheEffectsofActions............ 88
3.9.1 UpdatesofSentencesviaDistanced .................... 89
3.9.2 PropertiesoftheActionUpdateOperator ................ 93
3.9.3 ActionUpdatesofSentencesviaDistancediff ............ 96
3.9.4 AmbiguousActionUpdates ........................... 98
3.9.5 TakingCausalityintoAccount .........................100
References.....................................................102
4 IteratingRevision ..............................................105
4.1 Introduction ...............................................105
4.2 MotivatingStructureforBeliefRevision .......................107
4.3 IterationoftheRevisionOperation ............................108
4.4 PrioritisedDatabases........................................111
4.4.1 PropertiesoftheRevisionsofPDBs ....................115
4.4.2 DiscussionaboutIterationofRevision...................119
4.5 StructuredDatabases........................................124
4.6 ApplicationsandExamples ..................................126
4.7 RelatedWork..............................................128
4.7.1 PrioritisedBaseRevision..............................128
4.7.2 OrderedTheoryPresentations..........................131
4.8 UsingAdditionalInformationforActionUpdates ...............132
References.....................................................136
5 StructuredRevision ............................................139
5.1 IdentifyingInconsistency ....................................139
5.2 ReasoningwithPartialPriorities ..............................148
5.2.1 DegreeofConfidence/ReliabilityoftheSource ...........149
5.2.2 Linearisations .......................................165
5.3 Clustering.................................................166
5.4 ApplicationsinSoftwareEngineering .........................168
5.4.1 RequirementsSpecification............................168
5.4.2 AnExamplewiththeLightControlSystem ..............169
References.....................................................175
Contents ix
6 AlgorithmicContextRevision ...................................177
6.1 Introduction ...............................................177
6.2 AbductiveRevision.........................................177
6.2.1 IntroducingLDSfor⇒ ...............................180
6.2.2 Goal-DirectedAlgorithmfor⇒ ........................183
6.2.3 DiscussionontheAbductionProcedure .................187
6.2.4 AbductionAlgorithmfor⇒ ...........................192
6.2.5 AbductionforIntuitionisticLogic ......................194
6.3 CompromiseRevision ......................................197
6.3.1 IntroducingCompromiseRevisionfor⇒ ................197
6.3.2 ComparisonwithAGMRevision .......................210
6.4 ControlledRevision ........................................214
6.4.1 ProofTheory........................................216
6.4.2 PoliciesforInconsistency .............................217
6.4.3 Conclusions.........................................221
References.....................................................222
7 RevisionbyTranslation.........................................223
7.1 Introduction ...............................................223
7.2 BeliefRevisionforModalLogic..............................225
7.2.1 AnOverviewofPropositionalModalLogics .............225
7.2.2 HilbertSystemsforModalLogics ......................230
7.2.3 TranslationoftheModalLogicK intoClassicalLogic .....233
7.3 RevisinginŁukasiewicz’FinitelyMany-ValuedLogicŁ .........235
n
7.3.1 Łukasiewicz’FinitelyMany-ValuedLogicŁ ............235
n
7.3.2 TranslatingŁukasiewicz’Many-ValuedLogicŁ into
n
ClassicalLogic ......................................241
7.3.3 RevisioninŁukasiewicz’Many-ValuedLogic(Ł )........244
n
7.4 RevisinginAlgebraicLogics.................................246
7.4.1 TranslatingAlgebraicLogicintoClassicalLogic..........246
7.4.2 RevisioninAlgebraicLogics ..........................248
7.5 IntroducingBelnap’sFour-ValuedLogic .......................249
7.5.1 BeliefRevisioninBelnap’sFour-ValuedLogic ...........249
7.6 Belnap’sFour-ValuedLogic..................................251
7.6.1 AxiomatisingBelnap’sFour-ValuedLogic ...............251
7.6.2 EntailmentinBelnap’sFour-ValuedLogic ...............253
7.6.3 GeneralisingBelnap’sNotionofEntailmenttoInfinite
Theories............................................255
7.6.4 TranslatingBelnap’sLogicintoClassicalLogic...........259
7.6.5 RevisinginBelnap’sFour-ValuedLogic .................265
7.7 ConclusionsandDiscussions.................................268
References.....................................................269
Description:An important aspect in the formalisation of common-sense reasoning is the construction of a model of what an agent believes the world to be like to help in her reasoning process. This model is often incomplete or inaccurate, but new information can be used to refine it. The study of techniques that