Timeline EarlyWork(<1980) FoundationsofAnAI&LawCommunity(1987) CaseBasedReasoningandHYPO(1987) RuleBasedSystemsandTheMAKEProject(1990) DeonticLogic(1991) CABARET,aCBRandRuleBasedHybrid(1991) Argumentation(1991) LegalDocumentProcessingandFLEXICON(1993) ArgumentationFrameworks(1995) OntologyandONLINE(1995) E-GovernmentandZENO(1997)
LegalTextSummarisationandSALOMON(1998) METAlex,TheE-PowerProjectandLegalXML(2003) MachineLearningandDataMining(2005) TheBurdenofProofandCARNEADES(2007) OpinionMining,SocialMediaandAI&Law(2007) LegalTextInformationRetrievalandTREC(2010) MultiAgent-basedSystems,PADUAandPISA(2012) AI&LawandtheeraofBigData(2013) WhereNext---TwoExamples(2017)
AI&LawAIandLaw(LegalInformatics):TheapplicationandutilisationoftechniquesfromthecomputersciencedomainofArtificialIntelligencetothelegalapplicationdomain.
+
EarlyWork PublicationsonwhatmightbenowbeconsideredtobeAIandLawcanbetracedbacktotheearly1950s---LaymanAllen(1956),"Symboliclogic:Arazor-edgedtoolfordraftingandinterpretinglegaldocuments".
ThephraseAI&Lawstartedtobeusedinthe1970s---BruceBuchananandThomasHeadrick(1970),“Somespeculationaboutartificialintelligenceandlegalreasoning”.
OneearlysystemofnoteistheTAXMANsystem,afoundationformuchfuturework(TAXMANIIin1981)---ThorneMcCarty(1977),ReflectionsonTaxman:AnExperimentinArtificialIntelligenceandLegalReasoning.
ThebirthofanAI&LawcommunitycanbearguedtobethefoundationoftheInternationalConferenceonAIandLAW(ICAIL)whichwasfirstheldinBostoninMay1987(the16thICAILwasheldthisyear).
InEuropethefirstconferenceofthe“StichtingJuridischeKennissystemen”(JURIX),whichtranslatestothe”FoundationforLegalKnowledgeSystems”,washeldin1988.
FoundationsofAnAI&LawCommunity
FoundationsofAnAI&LawCommunitycont.
InternationalAssociationforAIandLaw(IAAIL)wasinauguratedin1991. TheJournalofArtificialIntelligenceandLawwaslaunchedin1992. JURISINseriesofworkshopsstartedin2007inJapan.
CaseBasedReasoningandHYPO
Workinthe1980swasdirectedat:(i)Informationextractionandinformationretrievalofvariouskinds,and(ii)theconstructionofexpertsystemsofvariouskinds.
Therewasmuchdiscussionastowhetherrulebasedorcasebasedwasthebestwaytogo.
InthecontextofCBRtheHYPOreasoningwithcasesand“hypotheticals”systemisofnote---EdwinaRisslandandKevinAshley(1987),“Acase-basedsystemfortradesecretslaw”.
TheSignificanceofHYPOarethat:§ Itisthefirstsystemthatthereasonedwithcasesinthelegaldomain,and
§ Itwasthecatalystformanysubsequentsystems(CABARET,CATO,IBP).
RuleBasedSystemsandTheMAKEProject
Rulebasedapproachestookafairlystandard(forthetime)expertsystemsapproach.
TheToulminmethodofinformalreasoningwaspopular.
Earlyexample---MarekSergotetal.(1986),TheBritishNationalityActasalogicprogram.
Laterexample,theMAKEproject---Bench-Capon,T.J.M.andCoenen,F.P.(1994),TheMaintenanceofLegalKnowledgeBasedSystems. CollaborativeprojectbetweenICL,BritishCoalandtheUniversityofLiverpool. Focuson:
§ Industrialinjuryclaims,and§ Themaintenanceoflegalexpertsystems.
DeonticLogic(s) Anotherfocusofthelate1980sandearly1990swasworkonvariouslogicalformalisms. InparticularDeonticlogics,whichwaspopularisedwithintheAIandLawcommunityinthe1980s---HenningHerrestad(1991),Normsandformalization. Usedthethenotation:
§ OA-itoughttobethecasethatA § PA-itispermissiblethatA
Addedthefollowingaxiomstostandardpropositionallogic:§ O(A->B) -> (OA->OB) § PA -> (not O not A)
Arguedtobesuitedtotherepresentationofdomainswherewehavenormsprovidingcompensationsforviolations,suchasinthecaseofcontactlaw.
CABARET,aCBRandRuleBasedHybrid
FollowingonfromHYPO,CABARETAdoptedaCBRandRulebasehybridapproach---EdwinaRisslandandDavidSkalak(1989),Combiningcase-basedandrule-basedreasoning:aheuristicapproach. Keyidea:Toplevela(argument)treederivedfromtherules,casesusedtoresolvetheleafnodes. Essentiallyashellthatintegratedreasoningwithrulesandcaseswithanemphasisontheapplicationofrulescontainingill-definedterms. Operatedusingacollectionofcontrolheuristicsdirectedatinterleavingcase-basedmethodsandrule-basedmethodstoconstructanargumenttosupportaparticularinterpretation.
Argumentation Studyofhowconclusionsarereachedthroughlogicalreasoninginthecontextofdebate,dialogue,persuasionandnegotiation. InthecontextofAI&lawtherelevanceiswithrespectto:(i)theexplanationoflegaldecisions,(ii)thestructureofargumentsinlegalcasesand(iii)therolethatargumentscanplayincase-basedreasoning. Earlywork(1990s)featuredworkonstrategiesforgeneratingargumentsinthecontextoftheCABARETsystem---DavidSkalakandEdwinaRissland(1991),Argumentmovesinarule-guideddomain. Thiswasafoundationforlaterworkfocusedonthecharacterisationandconstructionoflegalargumentdialoguegamesforlegalreasoning.
LegalDocumentProcessingandFLEXICON
The1990ssawthestartofwork(ongoing)onintelligentlegaldocumentmanagementtools. OnewelldocumentedexampleisFLEXICON---DaphneGelbartandJoeSmith(1993),FLEXICON:anevaluationofastatisticalrankingmodeladaptedtointelligentlegaltextmanagement. Usedtosummariselegaltextssoastofacilitaterapidsearch.
FLEXICONfeaturesthreeelements:(i)originaldocuments,(ii)multiplelexiconsand(iii)machinegeneratedformallexicalrepresentationsofeachdocument(FlexNotes). ThequalityoftheFlexNotesdependsonthelexiconsandtheadoptedinformationextractionprocesses.
ArgumentationFrameworks ArgumentationframeworkswerepopularisedwithintheAIcommunityfromthelate1990sononwards. OfnoteisDung’sframework---PhanMinhDung(1995),Ontheacceptabilityofargumentsanditsfundamentalroleinnonmonotonicreasoning. BroadlyanargumentationframeworkF=<AR>comprises:
§ AsetofabstractargumentsA,and§ AsetofbinaryattackrelationsR.
ArangeofargumentationframeworkshavebeenproposedwithintheAI&Lawcommunity:(i)logicbased,(ii)valuebased,(ii)assumptionbased,(iv)….. Muchsubsequentworkonargumentationframeworks,e.g.Zeno,ASPICandCarneades.
LegalTextSummarisationandSALOMON
TheAI&Lawcommunityhasbeeninterestedinsummarisinglegaltextssinceitsfoundation. OfnoteistheSALOMONsystem(project)---Marie-FrancisMoens(1997),Abstractingoflegalcases:theSALOMONexperience. SummarisesBelgiancriminalcasesfromthecasetext. Casecategoryfirstidentifiedbasedonthecasestructureandirrelevanttextunitswereidentifiedbasedonaknowledgebase. Informativetextunitsoftheallegedoffencesandoftheopinionofthecourtextractedusingstatisticaltechniques. Catalystformuchfurtherincludingargumentationmining.
OntologyandONLINE “Anexplicitformalspecificationsofthetermsinthedomainandtherelationshipsthatexistamongthem” TheconceptofontologieswasintroducedtoComputerScienceintheearly1990swiththeadventoftheSemanticWeb. ThefirstlegalontologycanbearguedtobethatincorporatedintoONLINE---AndreValenteandJoostBreuker(1995),ON-LINE:anarchitectureformodellinglegalinformation. Onlinewasalegalinformationmanagementsystemwithalegalreasoningfacilityprovidedbyalegalontology,the“FunctionalOntologyofLaw”. Interestinglytheideaoflegalontologiesshiftslegalreasoningfromarules-orientedproblemtoaknowledgerepresentationproblem. Seealso---TrevorBench-Capon(2004),Ontologiesinlegalinformationsystems;theneedforexplicitspecificationsofdomainconceptualisations.
E-GovernmentandZENO Theconceptofe-Government(e-Democracy)wasbeginningtotherecognisedinthemid1990s. Directedattheusageoftheinternettoprovidefordigitalinteractionbetweengovernmentsandcitizens.Forexamplee-Petitions. Zenowasoneofthefirstattemptstodefineaformalmodelofargumentationtosupportonlineconsultationsystemsfore-government---TomGordonandNikosKaracapilidis(1997),TheZenoargumentationframework. TheZenoframeworkprovidedthefoundationforanumberofsubsequentprojectssuchastheDelphiMediationOnlineSystem(DEMOS),Parmenedis---AnnMacintosh(2009),Providingargumentsupportfore-Participation.
METAlex,TheE-PowerProjectandLegalXML
ThenotionoftheSemanticWebbegantobetakenseriouslywithintheAI&Lawcommunityintheearly2000swithattemptsatadoptingaSemanticWebapproachinthecontextoflegalresources,usingjurisdictionindependentXMLstandardstostructuredocuments. EarlyexampleisMETAlex,anopenXMLstandardforthemarkupoflegaldocuments---RadboudWinkelsetal.(2003),Metalex:AnXMLstandardforlegaldocuments. DevelopedaspartoftheEuropeane-Powerproject,meantasaninterchangeformatforlegaldocuments. LedtoLegalXML,isanon-profitorganizationdevelopingopenstandardsforlegaldocumentsandrelatedapplications.
TheBurdenofProofandCARNEADES
TheCarneadessystem,developedfromZeno. Softwaresystemdevelopedforconstructing,evaluatingandvisualisingargumentsfromformalmodelsoflegalconcepts,rules,andcases----TomGordonetal.(2007),TheCarneadesmodelofargumentandburdenofproof. Oneofthefirstapproachestoacknowledge,andtakeaccountofthe“burdenofproof”. Ledtotheideaof“argumentativestorytelling”wherewehavetwocompeting/conflictingstories---FlorisBex(2011),Arguments,storiesandcriminalevidence:Aformalhybridtheory.
MachineLearningandDataMining
Machinelearninganddatamininghavealonghistory.
TheAIandLawcommunitywereslowtotakeuptheconceptswithfirstpapersappearinginthemid2000s.
Firstexamplesweredirectedatlegaldocumentsummarisationutilisingtextminingtechniques---BenHacheyandClaireGrover(2005),Automaticlegaltextsummarisation:experimentswithsummarystructuring.
Furtherworklookedat:
§ Theclassificationoflegaldocuments,
§ Opinionminingfromlawblogs…
OpinionMining,SocialMediaandAI&Law
Theideaofopinionmininginthecontextofrecommendersystemswasasubjectofmuchresearchinthemachinelearningcommunityinthemid2000s.
FirstexampleinthelegaldomainconsideredopinionmininginthecontextofLawblogs(“blawgs”thatexistinthe‘‘blawgosphere’’!)---JackConradandFrankSchilder(2007),Opinionmininginlegalblogs.
ThesignificanceofConradandFrankisthatitwasthecatalystformuchsubsequentworkrelatingsocialmedialtothedomainofAI&Law.
LegalTextInformationRetrievalandTREC
ThefirstTextREtrievalConference(TREC)washeldin1992. EveryconferencefeaturesanumberofTracks(Workshops)eachdirectedatparticularcurrentretrievaldomain. Eachtrackhasachallengecenteredongivendatasets. From2006to2011TRECfeaturedaLegalTrack. Anideathatresultedfromthis,thathashassignificantinfluencewithintheAIandLawise-Discoveryfoundedonrandomsampling---DouglasOardetal.(2010),EvaluationofinformationretrievalforE-discovery.
MultiAgent-basedSystems,PADUAandPISA
InlinewiththeAI&Lawcommunity’spracticeofadoptingtechnologieswhenandwheretheyfittheaspirationsofthecommunitythemid2000ssawtheadoptionofMAS. Theideaofagentsarguingfororagainstacase. ExamplesincludethePADUAandPISAsystems---MayaWardehetal.(2012),PISA:AFrameworkforMultiagentClassificationUsingArgumentation. BroughttogetherdataminingandargumentationundertheMASumbrella. Examplesfoundedonhousingbenefitscenarios.
AI&LawandtheeraofBigData
TheBigData(DataAnalytics)phenomenahittheheadlinesintheearly2010s. Inthelegalcontextthiswasaccompaniedbytherealisationthatevenroutinecivillitigationcasesnowrequiredsearchingofextensivedigitalrepositories. Bigdatatechnologyalsoallows:§ Largescalelegaloutcomepredictionin
themoldofmachinelearningbutusingdeeplearningtechniques(RNNs,etc.).
§ Provisionofevidenceextractedfrompublicdatasetsusingtechniquesfromdataanalytics.
§ Andmore
WhereNext,TwoExamples AdefiningfeatureoftheBigDataeraisthefactthatithasseenextensiveinterestsanduptakefromindustry.ThishasbeenaccompaniedbyaresurgenceininterestinAI.Thelegalindustryisnoexception. RiverviewLaw:Reasoningconcerningcontractrenewalstrategiesinthecontextofverylargenumbersofcontracts(tensofthousands)---MatiasGarcia-Constantinoetal.(2016),CLIEL:Context-BasedInformationExtractionfromCommercialLawDocuments. Fletchers:Decisionmakingregardingmedicalnegligencecasesfoundedonmedicalrecordsandtranscripts.
TimelineRevisited EarlyWork(<1980) FoundationsofAnAI&LawCommunity(1987) CaseBasedReasoningandHYPO(1987) RuleBasedSystemsandTheMAKEProject(1990) DeonticLogic(1991) CABARET,aCBRandRuleBasedHybrid(1991) Argumentation(1991) LegalDocumentProcessingandFLEXICON(1993) ArgumentationFrameworks(1995) OntologyandONLINE(1995) E-GovernmentandZENO(1997)
LegalTextSummarisationandSALOMON(1998) METAlex,TheE-PowerProjectandLegalXML(2003) MachineLearningandDataMining(2005) TheBurdenofProofandCARNEADES(2007) OpinionMining,SocialMediaandAI&Law(2007) LegalTextInformationRetrievalandTREC(2010) MultiAgent-basedSystems,PADUAandPISA(2012) AI&LawandtheeraofBigData(2013) WhereNext---TwoExamples(2017)