Pragmatik 1

Embed Size (px)

Citation preview

  • 7/29/2019 Pragmatik 1

    1/405

  • 7/29/2019 Pragmatik 1

    2/405

  • 7/29/2019 Pragmatik 1

    3/405

    APRAGMATICLEGALEXPERTSYSTEM

  • 7/29/2019 Pragmatik 1

    4/405

    Tomyparents,KevinandMarli,fortheirunfailingloveandsupport;

    andtoPaula: forwhom,butforwhom. . .

  • 7/29/2019 Pragmatik 1

    5/405

    APragmaticLegalExpertSystem

    jamespopple

    VisitingFellowTheAustralianNationalUniversity

    DartmouthAldershot BrookfieldUSA Singapore Sydney

  • 7/29/2019 Pragmatik 1

    6/405

    James Popple 1996

    All rights reserved. No part of this publication may be reproduced, stored in aretrieval system, or transmitted in any form or by any means, electronic, mechani-cal, photocopying, recording, or otherwise without the prior permission of DartmouthPublishing Company Limited.

    Published byDartmouth Publishing Company LimitedGower HouseCroft RoadAldershotHants GU11 3HREngland

    Dartmouth Publishing CompanyOld Post RoadBrookfieldVermont 05036USA

    British Library Cataloguing in Publication DataPopple, James

    A pragmatic legal expert system. - (Applied legalphilosophy)1. Law 2. Expert systems (Computer science)I. Title340'.0285'533

    Library of Congress Cataloging-in-Publication DataPopple, James, 1964-

    A pragmatic legal expert system / James Popple.p. cm. (Applied legal philosophy)

    Includes bibliographical references and index.ISBN 1-85521-739-21. SHYSTER. 2. LawMethodologyData processing. 3. Law-

    -Interpretation and constructionData processing. 4. Expertsystems (Computer science)Australia. 5. LawAustralia--MethodologyData processing. I. Title. II. Series.

    K212.P664 1996340'.1dc20 96-7058

    CIP

    ISBN 1 85521 739 2

    Printed in Great Britain by the Ipswich Book Company, Suffolk

  • 7/29/2019 Pragmatik 1

    7/405

    ContentsFigures ixSeriespreface xiPreface xiiiChapter1: Legalexpertsystems 11.1 Introduction 21.2 Jurisprudence 51.3 Jurimetricsandthebehaviourists 191.4 Rule-basedsystems 241.5 Case-basedsystems 381.6 Hybridsystems 441.7 Conceptualmodels: deepandshallow 461.8 Conclusion 49

    Notes 50Chapter2: Apragmaticapproachtocase law 612.1 Introduction 622.2 Designcriteria 622.3 Amodeloflegalreasoning 662.4 Experts,users,andexpertusers 692.5 Knowledgerepresentation 702.6 Aspecificationlanguage 742.7 Weightingattributes 752.8 Detectingattributedependence 792.9 Calculatingdistances 812.10 Nearestcasesandnearestresults 892.11 Writingareport 922.12 Safeguards 1012.13 Testingandevaluation 1032.14 Conclusion 111

    Notes 118v

  • 7/29/2019 Pragmatik 1

    8/405

    vi Apragmatic legalexpertsystem

    Chapter3: ImplementingSHYSTER 1253.1 Introduction 1263.2 TheShystermodule 1283.3 TheStatutesmodule 1283.4 TheCasesmodule 1303.5 TheTokenizermodule 1313.6 TheParsermodule 1323.7 TheDumpermodule 1383.8 TheCheckermodule 1413.9 TheScales module 1463.10 TheAdjuster module 1473.11 TheConsultantmodule 1473.12 TheOdometermodule 1483.13 TheReporter module 1513.14 Conclusion 159

    Notes 159Chapter4: Casestudies 1634.1 Introduction 1644.2 AsimulationofFINDER 1644.3 Theauthorizationofcopyrightinfringement 1734.4 Employeesandindependentcontractors 1914.5 The implicationofnaturaljustice 2074.6 Conclusion 230

    Notes 234Chapter5: Conclusion 2435.1 Introduction 2445.2 EvaluatingSHYSTER 2445.3 Futureresearch 2505.4 ThecontributionoftheSHYSTERproject 252

    Notes 253

    AppendixA: Examplereports 255A.1 Introduction 256

    Notes 256A.2 ReportfileforParkerv.BA 257A.3 ReportfileforAPRA v.Jain 263A.4 ReportfileforRePorter 270A.5 ReportfilesforAinsworthv.CJC 278

  • 7/29/2019 Pragmatik 1

    9/405

    Contents vii

    AppendixB: Acompleteexample 289B.1 Introduction 290

    Notes 290B.2 LogfileforBWIU v.Odco 291B.3 DumpfilefortheEmployee specification 294B.4 ProbabilitiesfilefortheEmployee specification 322B.5 WeightsfilefortheEmployee specification 323B.6 DistancesfileforBWIU v.Odco 323B.7 ReportfileforBWIU v. Odco 331

    Bibliography 341Cases 367Statutes 373Index 375

  • 7/29/2019 Pragmatik 1

    10/405

  • 7/29/2019 Pragmatik 1

    11/405

    Figures2.1 Structureofanexamplespecification 712.2 Formsoffunctionaldependence 812.3 Two-wayassociationtable 852.4 Exampledistances 912.5 SHYSTERsalgorithmforchoosingcases 942.6 SHYSTERsalgorithmforusingcases: partA 962.7 SHYSTERsalgorithmforusingcases: partB 972.8 SHYSTERsalgorithmforusingcases: partC 983.1 SHYSTERscommandlineswitches 1293.2 KeywordsinSHYSTERsspecificationlanguage 1313.3 EBNFdefinitionofSHYSTERsspecificationlanguage 1333.4 AttributevaluesfortheEmployeearea 1393.5 Exampleattributevalues 1423.6 Probabilitiesforexampleattributes 1433.7 ExtractfromprobabilitiesfortheEmployeearea 1463.8

    Similarity

    measures

    149

    3.9 DistancesforBWIU v.Odco 1503.10 ExtractfromthelogfileforBWIU v.Odco 1574.1 ProbabilitiesfortheFinderarea 1704.2 WeightsfortheFinderarea 1714.3 DistancesforParkerv. BA 1744.4 ProbabilitiesfortheAuthorizationarea 1804.5 WeightsfortheAuthorizationarea 1814.6 DistancesforCBS v. Ames 1844.7 DistancesforNarichv. CPT 1984.8 Relationshipbetweenareas intheNaturalspecification 2104.9 WeightsfortheNaturalspecification 2154.10 ProbabilitiesfortheNatural specification 2164.11 DistancesforTwistv.Randwick intheNaturalarea 2194.12 Summaryoftestcases 2314.13 Summaryofgeneratedtesting 233

    ix

  • 7/29/2019 Pragmatik 1

    12/405

  • 7/29/2019 Pragmatik 1

    13/405

    SeriesprefaceThe objective of the Dartmouth Series in Applied Legal Philosophy is to publish work which adopts a theoretical approach tothestudyofparticularareasoraspectsoflawordealswithgeneraltheoriesoflawinawaywhichfocusesonissuesofpracticalmoralandpoliticalconcerninspecificlegalcontexts.

    Inrecentyearstherehasbeenanencouragingtendencyforlegalphilosophers to utilize detailed knowledge of the substance andpracticalities of law and a noteworthy development in the theoretical sophistication of much legal research. The series seeks toencouragethesetrendsandtomakeavailablestudiesinlawwhicharebothgenuinelyphilosophicalinapproachandatthesametimebasedonappropriatelegalknowledgeanddirectedtowardsissuesinthecriticismandreformofactuallawsandlegalsystems.

    The series will include studies of all the main areas of law,presented ina mannerwhichrelates totheconcerns ofspecialistlegal academics and practitioners. Each book makes an originalcontribution to an area of legal study while being comprehensibletothoseengaged inawidevarietyofdisciplines. Their legalcontent is principally Anglo-American, but a wide-ranging comparative approach is encouraged and authors are drawn from avarietyofjurisdictions.

    TomD.CampbellSeries

    Editor

    FacultyofLaw

    TheAustralianNationalUniversity

    xi

  • 7/29/2019 Pragmatik 1

    14/405

  • 7/29/2019 Pragmatik 1

    15/405

    PrefaceThe history of the development of legal expert systems has, forthemostpart,beencharacterizedbythedevelopmentandimplementationofcomplexmodelsoflegalreasoning.

    This book aims to show that a legal expert system need notbe based upon a complex model of legal reasoning in order toproduceusefuladvice. Itadvocatesapragmaticapproachtothedesignof legalexpertsystemsanapproachbasedontheway inwhich lawyersdealwiththelawonaday-to-daybasis.

    It argues that a system based upon a simple model of legalreasoningcanstillproducegoodadvice,wherethatadviceisevaluatedbyreferencetotheaccuracyofthesystemspredictionsandtothequalityofitsarguments. Furthermore,suchasystem,withitssimplerknowledgerepresentationstructure,makessimplertheprocessofknowledgeacquisition.

    Theseargumentsaremadetheoretically,andthenbyexample.Ihavedevelopeda legalexpertsystemwhich isbasedonapragmaticapproachtothe law. Thedevelopmentandtestingofthatsystem,calledSHYSTER,isdescribedinthisbook.

    SHYSTERisofageneraldesign,sothatitcanoperateindifferentlegaldomains. Itwasdesignedtoprovideadvice inareasofcaselaw that have been specified by a legal expert using a speciallydevelopedspecification language.

    SHYSTER isacase-based legalexpertsystem. Itsknowledgeofthe law isacquired,andrepresented,as informationaboutcases.Itproduces itsadvicebyexamining,andarguingabout,thesimilarities and differences between cases. By contrast, a rule-basedexpert system represents the law using rules. A hybrid systemuses both rule-based and case-based techniques. SHYSTER hasbeendesignedsothatitcanbelinkedwitharule-basedsystemtoformahybridlegalexpertsystem.

    xiii

  • 7/29/2019 Pragmatik 1

    16/405

    xiv Apragmatic legalexpertsystem

    AlthoughSHYSTERattemptstomodelthewayinwhichlawyersarguewithcases, itdoesnotattempttomodeltheway inwhichlawyers decide which cases to use in those arguments. It usesstatisticaltechniquestoquantifythesimilaritybetweencases,andchooses cases on the basis of thatsimilarity measure. ItsrepresentationstructurewasdesignedtobeassimpleaspossiblewhilecomplexenoughtoallowSHYSTER toproducegoodadvice.Thebodyofthisbookisdividedintofivechapters:

    Chapter1discussespreviousworkofrelevance tothedevelopment of legal expert systems. The value ofjurisprudencetothedevelopmentofsuchsystemsisalsodiscussed,andtheadoptionofapragmaticapproach(asopposedtoajurisprudentiallypureapproach)isrecommended.

    A pragmatic approach is proposed in chapter 2. It incorporatesasimplemodelof legalreasoning,andusesasimpleknowledge representation structure. Comparisons are madebetween

    this

    approach

    and

    others.

    This

    approach

    is

    adopted

    forthedevelopmentofSHYSTER. Specificdesigncriteria forSHYSTERaredetailed,andmethodsoftestingandevaluatingthesystemarediscussed.

    The implementation of SHYSTER is explained in chapter 3.The twelve modules that comprise SHYSTER are described,andtheiroperationisdemonstratedusingexamples.

    FourdifferentspecificationshavebeenwrittenforSHYSTER,and these are used as the basis of case studies in chapter 4.Eachspecificationrepresentsadifferentareaofcaselaw. Several different methods are employed to test SHYSTER andthesespecifications.

    Inchapter 5, conclusions are drawnabout SHYSTER and itsapproach to case law. Some enhancements to SHYSTER aresuggested, avenues of future research are identified, and thenature of the contribution made by the SHYSTER project isdiscussed.

    There are two appendices to this book. Appendix A providesexamplereportsrelatingtoeachofthefourspecifications. EachofthesereportsisSHYSTERsadviceononeofthetestcasesusedinchapter4. AcompleteexampleofSHYSTERsoutputforanothertestcaseisgiveninappendixB.

  • 7/29/2019 Pragmatik 1

    17/405

    Preface xv

    IhaveendeavouredtoexplainSHYSTER anditsapproachtocaselawasclearlyaspossible. Occasionallyespecially in chapters 2and3thisexplanationinvolvesrecoursetomathematicaldetail.Readers for whom the sight of a formula provokes either greatfearorprofoundindifference,andwhodontfeelaneedtounderstandtheintricaciesofSHYSTERsinnerworkings,areadvisedtodisregardanyoffendingmathematics.Forthosewhowouldliketoexperimentwiththesystem,SHYSTERandthecaselawspecificationsdiscussedinthisbookareavailableontheworldwidewebathttp://cs.anu.edu.au/software/shyster.ThisbookisbaseduponmydoctoralthesiswhichIwrotewhilearesearchstudentatTheAustralianNationalUniversity,andcompleted in April 1993. I thank the members of my supervisorypanelRobinStanton,RogerClarke,PeterDrahosandMalcolmNeweyfor theircounseland encouragement, andRobinCreykeandPhillipaWeeksforgivingsofreelyoftheirtimeandlegalexpertise. I have also been provided with valuable advice on thisworkbymythesisexaminersKevinAshley,BobMolesandAlanTyreeandbyRichardSusskind.

    Notwithstandingthisassistance, forwhich Iammostgrateful,responsibilityforthisbookremainsmine.

    JamesPoppleApril1995

    http://cs.anu.edu.au/software/shyster/http://cs.anu.edu.au/software/shyster/
  • 7/29/2019 Pragmatik 1

    18/405

  • 7/29/2019 Pragmatik 1

    19/405

    shy.ster \ shst (r)\ n -s [prob. after Scheusterfl1840 Am. attorney frequently rebuked in a New York court for pettifoggery] : one who is professionallyunscrupulousesp. inthepracticeof laworpolitics. . .

    WebstersThirdNewInternationalDictionary (1961)shyster( a st (r)). . . [Ofobscureorigin.

    Itmightbe f.shya.(sense7,disreputable)+-ster;butthissenseoftheadj. isapp.notcurrentintheU.S.]

    . . . Alawyerwhopractisesinanunprofessionalortrickymanner;especially,onewhohauntstheprisonsandlowercourtstopreyonpettycriminals;hence,any one who conducts his business in a tricky manner (Funks Stand. Dict.1895). Alsoattrib.oradj. Orig.andchieflyU.S.slang . . .

    TheOxfordEnglishDictionary (secondedition,1989)shyster. An unscrupulous lawyer (note that the definition presumes theexistenceofscrupulousones) . . .

    The term does not come fromas suggested in various dictionariesthesurnameScheuster,supposedlyalawyernotedforshyster-likepractices;fromthe name of the Shakespearean character, Shylock; . . . or from any of thevariousmeaningsofshy(e.g.,tobeshyofmoney). Rather. . . shyster evolvedfrom the underworld use ofshiser, a worthless fellow, which derived in turnfrom the German scheisse, excrement, via scheisser, an incompetent person(specifically,onewhocannotcontrolhisbodilyfunctions). . .

    HughRawson(1991)ADictionaryofInvective

    xvii

  • 7/29/2019 Pragmatik 1

    20/405

  • 7/29/2019 Pragmatik 1

    21/405

    1 LegalexpertsystemsIntheyahooworldofcomputerpublicrelations,itisoftentheloudestmouthswhichleadtheunsuspectingcomputeruser intothat lonelycanyonofemptypockets and broken promises. It now seems . . . that parts of the academicworldarefastapproachingthatdecibellevelsofarachievedonlybycomputersalesmen. Thus. . . itseemsthatspacecanonlybebookedontheband-wagonifone ispreparedtomakemoreoutrageousclaimsthanthenextman. AndsuchclaimsarebeingmadefortheapplicationofAItothe law.

    PhilipLeith(1986)1

    Thecomputerscientists,encouragedbythemodernpositivists,failtorecognize. . . thatlaw,positivemoralityandethicsareinseparablyconnectedpartsof a vast organic whole. Judgments are involved at every stage of the legalprocessandmachinescannotmakejudgments. Instatingthatlegalrulescanbe applied without furtherjudgment; that they apply in an all or nothingfashion; that legal decision making follows the form of the syllogism or thatit is a pattern-matching routine, the modern positivists,joined now by thecomputerscientiststakeusalongadangerousroad.

    RobertN.Moles(1987)DefinitionandRule inLegalTheory:

    AReassessmentofH.L.A.Hartand thePositivistTradition2

    Aussi, lorsquun homme se rend Thus when a man takes on absoluteplus absolu, songe-t-il dabord a power,hefirstthinksofsimplifyingthesimplifier les lois. On commence, law. In such a state one begins to be

    dans cet Etat, a etre plus frappe moreaffectedbytechnicalitiesthanbydes inconvenients particuliers, que thefreedomofthepeople,aboutwhichde la liberte des sujets dont on ne oneno longercaresatall.sesouciepointdutout.

    Montesquieu(1748)De lEspritdes lois3

    1

  • 7/29/2019 Pragmatik 1

    22/405

    2 Apragmatic legalexpertsystem 1.1

    1.1 IntroductionThis chapter reviews previous work of relevance to the designof legal expert systems. Jurisprudence and its value to the development of such systems is discussed (1.2) and the field ofjurimetricsandsomebehaviouristicresearchisexamined(1.3).Thedevelopmentoflegalexpertsystemsisdiscussedunderthreeheadings: rule-based,case-based,andhybridsystems.4

    Rule-based systems are examined first (1.4), and special attention is paid to the work of four projects. The problems ofknowledgeacquisitionandrepresentation,andfactrepresentationare also discussed (1.4.5 and 1.4.6). Rule-based systems havebeenusedtorepresentstatutesandcases. However,asexplainedin1.4.7,rule-basedsystemsarefundamentallyinadequateforrepresentingcaselaw.

    Case-based systems areexamined(1.5)andthreeprojectsareoutlined. Systemsthatuserulesinordertorepresentcaselawareconsidered to be rule-based systems; only systems which adoptcase-based reasoning methods are considered to be case-based.Applying this distinction is usually straightforward, though notalways.5 Attempts have been made to use semantic networks torepresentcaselawbut,asexplainedin1.5.4,theyarenotsuitedtothetask.

    Hybrid systems (1.6) employ both rule-based and case-basedmethods.

    Thesearchforconceptualmodels oflegalreasoningisdiscussed(1.7). Finally,conclusionsaredrawnfromtheliteratureastothebestapproachtothedesignoflegalexpertsystems(1.8).Butfirstitisnecessarytoexplainwhatismeantbyalegalexpertsystem,andtooutlinesomekeyconcepts.1.1.1 LegalexpertsystemsTherangeofcomputerapplicationsinthelawiswide. Itextendsfrom general applications, of use to lawyers, to applications designed specifically for the law. This book isconcerned onlywitha subset of those systems that make use of artificial intelligence(AI)techniquestosolvelegalproblems.

    Legal AI systems can usefully be divided into two categories:legalretrievalsystemsandlegalanalysissystems.6 Legalretrieval

  • 7/29/2019 Pragmatik 1

    23/405

    3 1.1 Legalexpertsystems

    systems allowlawyerstosearchthroughdatabasescontainingdetailsofstatutesanddecidedcases. AItechniquesmaybeemployedto simplify this task: e.g. by searching for keywords which havenotbeeninputbytheuserbutarededucedtobeequivalentto,orsufficientlyrelatedto,theinputkeywords.7 Legalanalysissystemstake informationaboutasetoffactsanddeterminetheramificationsofthosefactsinagivenareaoflaw.

    Mehl claims that there is no fundamental difference betweenthese two categoriesthat the difference is one of degree only.8ShannonandGolshanisuggestthatthedifferencebetweensystemsbasedonaconceptualmodeloflegalanalysisandtextretrievalsystems is that the latter do not understand any area of thelaw.9 Similarly, Susskind says that knowledge-based systems,as opposedto databasesystemsare capable of applyingtheirknowledgeofthelawtotheproblemdatapresentedtothem.10

    A better distinctionone which avoids the vexed question ofwhetheranyAI systemcanreallybesaidtoknoworunderstandanythingcanbemadebyreferencetotheoutputofthesystem.The

    output

    from

    alegal

    analysis

    system

    is

    such

    that,

    if

    it

    had

    been

    produced by a human, that human would be said to have legalexpertise. By contrast, the output from a legal retrieval systemcould be produced by a human possessed of no legal expertise;suchoutputisusedin,andisnottheproductof,legalanalysis.11

    Thisbookisconcernedonlywithlegalanalysissystems.Legal analysis systems can be divided into two categories: judgmentmachinesand legalexpertsystems.

    Ajudgmentmachine isamachinedesignedtoreplaceahumanjudge. Suchmachineswerefirstproposedfortyyearsago,thoughnosuchproposalshavebeenmade fornearlytwodecades. Writingsonjudgmentmachinesarediscussedbelowbecausetheyareofhistoricalinterest,andbecausetheideaofajudgmentmachineraisessome issues which arealso relevant to the secondcategoryoflegalanalysissystems: legalexpertsystems.

    A legal expert system, as the term is used in this book, is asystemcapableofperformingata levelexpected ofa lawyer. AIsystemswhichmerelyassistalawyerincomingtolegalconclusionsor preparing legal argument are not here considered to be legalexpert systems; a legal expert system must exhibit some legalexpertiseitself. Thisdefinitiondoesnotexcludesystemsthatcanonly be used by legal experts; several systemsincluding SHYSTERhavebeendevelopedforuseexclusivelybylawyers.12

  • 7/29/2019 Pragmatik 1

    24/405

    4 Apragmatic legalexpertsystem 1.1

    1.1.2 SourcesoflawThe

    law

    in

    so-called

    common

    law

    countriese.g.

    Australia,

    Britain, Canada and the USAis derived from legislation and from

    caselaw.Legislation, or statute law, consists of statutes and delegated

    legislation. InAustralia,statutesaremadebytheFederalParliament,oraStateParliament,andenactedwhengivenroyalassentby the Governor-General, or a State Governor. Delegated legislation(rules,regulations,ordinances,by-laws,etc.) ismadebyapersonorbodytowhom legislativepowerhasbeendelegatedbystatute.

    Caselaw,orthecommonlaw,isjudge-madelaw:judicialresolutionsofspecificdisputes. Thesourcesofcaselawarethepublishedreportsofthecasesasheardbeforevariouscourts.

    Legislation takes precedence over case law;13 parliaments canoverridejudge-madelawbylegislativeenactment. However,judgeshavethetaskofdeterminingthemeaningoflegislation.

    Any legal expert system which seeks todeal with the law in acommon lawcountrymustaccount forbothstatute lawandcaselaw.1.1.3 ThedoctrineofprecedentThedevelopmentofcase law isbasedupontheprincipleofstaredecisis whichholdsthatcourtsshouldapplythedoctrineofprecedent. Morrisetal.summarizethegeneralrulesofthedoctrineofprecedentasfollows:

    eachcourtisboundbydecisionsofcourtshigherinitshierarchy; adecisionofacourt inadifferenthierarchymaybeofconsid

    erableweight,butwillnotbebinding; onlytheratiodecidendi ofacaseisbinding; anyrelevantdecisions,althoughnotbinding,maybeconsidered

    andfollowed;andprecedentsarenotnecessarilyabrogatedbylapseoftime.14

    Courtshave a tendencyto follow similar cases, even when notstrictlyboundtodoso. AccordingtoCross:

    It is a basic principle of the administration ofjustice that likecases should be decided alike. This is enough to account for thefact that, in almost everyjurisdiction, ajudge tends to decide a

  • 7/29/2019 Pragmatik 1

    25/405

    1.2 Legalexpertsystems 5

    case in the same way as that in which a similar case has beendecidedbyanotherjudge.15Thereisconsiderableargumentabouttheextenttowhichjudges

    actually apply the doctrine of precedent.16 Some legal theoristscontend that the doctrine is simply part of the public discoursethatjudgesusetojustifytheir decisions. Stoneclaims thatthedegreeofcertaintyandstabilityinthelawsecuredbythedoctrineofstare decisis is far lessthan itappearstobe,andthatoneoftheimportantsocialfunctionsofthedoctrineistomaintainatamaximumthefeelingandappearanceofcertaintyandstability.17Yet,judgesofthehighestcourt inAustraliahavebeenknowntofollowpreviouscaseswhichtheybelievedtobewronglydecided.18

    What is clear is thatjudges use previously decided cases tojustify their decisionsif not to reach themand that lawyersusepreviouslydecidedcases in legalargument. Any legalexpertsystem which deals with case law assumes theapplication of thedoctrineofprecedent,atleasttothisextent.1.1.4 TerminologyFor consistency, the following three terms are used in this bookforavarietyoftermsthatareusedthroughouttheliterature. Anattribute is a legally important fact; some authors use thewordsdescriptor,dimension,factor,featureorvariable. Theinstantcase isthesituationaboutwhichtheexpertsystem is interrogated;thisissometimescalledtheactualcase,thecurrentfactsituation,thehypotheticalcaseorthenewcase. Leading cases make up the case base of a legal expert system; somecallthesethereferencesetorthetrainingset.1.2 Jurisprudence1.2.1 TheimportanceofjurisprudenceFor centuries, lawyers and philosophers have written on, and arguedabout,thenatureofhumanlaws. Onthefaceofit,itwouldseemsensibleforthedeveloperofalegalexpertsystemtohaveregardtothisjurisprudentialtheorybeforedesigningher/hissystem.However, asSusskindcomplained in 1987, most of thepublishedresearchonvariouslegalexpertsystemsmakesnouseofjurisprudentialresources.19

  • 7/29/2019 Pragmatik 1

    26/405

    6 Apragmatic legalexpertsystem 1.2

    Some expert system researchers have been doubtful as to thevalue ofjurisprudence to their research. Niblett, for example,claimsthatasuccessfulexpertsystemislikelytocontributemoretojurisprudencethantheotherwayround.20 Susskinddisagrees:

    It is beyond argument . . . that all expert systems must conformto somejurisprudential theory because all expert systems in lawnecessarily make assumptions about the nature of law and legalreasoning. To be more specific, all expert systems must embodytheories of legal knowledge, legal science, the structure of rules,the individuation of laws, legal systems and sub-systems, legalreasoning, and of logic and the law (as well perhaps as elementsof a semantic theory, a sociology, and a psychology of law), theories that must all themselves rest on more basic philosophicalfoundations. Ifthis isso, itwouldseemprudentthatthegeneraltheoryoflawimplicitinexpertsystemsshouldbeexplicitlyarticulated using (where appropriate) the relevant works of seasonedtheoreticians of law. Perhaps the reason that there is, as yet,no overwhelmingly successful system is that the vast corpus ofappositejurisprudential material has not yet been tapped in theconstructionprocess.21Althoughitistruethatalllegalexpertsystemsnecessarilymake

    assumptionsaboutthenatureof lawand legalreasoning, itdoesnotfollowthattheymustconformtosomejurisprudentialtheory.A lawyer must have a model of the law (maybe unarticulated)which includes assumptions about the nature of law and legalreasoning, but that model need not rest on basic philosophicalfoundations. It may be a pragmatic model, developed throughexperience within the legal system. Many lawyers perform theirworkwithlittleornojurisprudentialknowledge,22 andthereisnoevidence to suggest that they are worse, or better, at theirjobsthanlawyerswell-versed injurisprudence.

    Susskind concedes that it is possible to build a legal expertsystemwithoutjurisprudentialinsight,butsuggeststhatsuchasystemwouldbeofverypoorquality:

    Because successful legal knowledge engineering presupposes soprofound a familiarity with the nature of law and legal reasoning,itisscarcelyimaginablethatsuchamasterycouldbegainedotherthanthrough immersioninjurisprudence.23

    Harris,himselfajurisprudent,takesadifferentview:People acquire those technical skills of legal reasoning and legalargumentation which make up the concept of good lawyer by

  • 7/29/2019 Pragmatik 1

    27/405

    7 1.2 Legalexpertsystems

    immersingthemselvesinsubstantivelegalsubjects. Jurisprudencehastodo,notwiththelawyersroleasatechnician,butwithanyneed

    he

    may

    feel

    to

    give

    a

    good

    account

    of

    his

    lifes

    workeither

    tofellowcitizens,ortohimself,ortoanygodstherebe.24

    IfHarrisisrightandtheexistenceofgood(thoughjurisprudentially illiterate) lawyerssuggeststhathe isthenthe importanceofjurisprudenceto legalexpertsystemdesign isquestionable. Alegal expert system need only operate at the same level of abstraction as does a lawyer, rather than at the philosophical levelofajurisprudent. Thefactthatmany lawyershavemasteredtheprocessoflegalreasoning,withouthavingbeenimmersedinjurisprudence,suggeststhatitmayindeedbepossibletodeveloplegalexpertsystemsofgoodqualitywithoutjurisprudentialinsight.

    Thisdoesnotmeanthat legalexpertsystemsdesignersshouldcompletely ignorejurisprudential literature. However, as a legalexpert system need not conform to anyjurisprudential theory, apragmaticapproachtoexpertsystemdesignmaybepreferabletoa

    jurisprudentiallypureone. Susskind,asNiblettcomplains,givesthe role ofjurisprudence in the design of legal expert systems agreatersignificancethanitdeserves.25

    Itiswellbeyondthescopeofthisbooktocoverthevastcorpusof appositejurisprudential material,26 however several areas of

    jurisprudence of relevance to expert system design are discussedbelow.1.2.2 ScientificandmechanicaljurisprudenceThedictionarydefinesjurisprudence as:

    Thesciencewhichtreats ofhuman laws (written and unwritten)ingeneral;thephilosophyoflaw.27

    Poundreferstothescientificcharacterofthelaw:SirFrederickPollockgivesustheclewwhenhedefinesthereasonsthat compel law to take on this scientific character asthree: thedemandforfulljustice,thatisforsolutionsthatgototherootofcontroversies; the demand for equaljustice, that is a like adjustment of like relations under like conditions; and the demand forexactjustice,thatisforajusticewhoseoperations,withinreasonablelimits,maybepredictedinadvanceofaction. Inotherwords,themarksofascientificlaware,conformitytoreason,uniformity,andcertainty.28

  • 7/29/2019 Pragmatik 1

    28/405

    8 Apragmatic legalexpertsystem 1.2

    Thisapproachtothelawhasbeenmuchcriticized. Frankcontendsthatuncertaintyisinherentinthelegalprocess,andthatseekingcertaintyingenerallegalprinciplesissimplyanexpressionofinfantileemotionalattitudeswhichhavepersistedintoadulthood.29

    In his 1908 paper, Mechanical Jurisprudence, Pound distinguishesscientificjurisprudencefrommechanicaljurisprudence:

    Romanlawinitsdecadencefurnishesastrikingexample[ofmechanicaljurisprudence]. TheValentinianlawofcitationsmadeaselection ofjurisconsults of the past and allowed their writingsonly to be cited. It declared them, with the exception of Papinian,equalinauthority. Itconfinedthejudge,whenquestionsoflawwereinissue,tothepurelymechanicaltaskofcountingandofdeterminingthenumericalpreponderanceofauthority. Principleswere no longer resorted to in order to make rules fit cases. Theruleswereathandinafixedandfinalform,andcasesweretobefittedtotherules.30

    By

    contrast,

    Pound

    says,

    scientific

    law

    is

    a

    reasoned

    body

    of

    principlesfortheadministrationofjustice.31Loevinger also argues that there is a role for science in law,

    but sees little science injurisprudence. In 1949 he proposed analternativetojurisprudencewhichhetermedjurimetrics:

    Thenextstepforwardinthelongpathofmansprogressmustbefromjurisprudence(whichismerespeculationaboutlaw)tojurimetricswhichisthescientificinvestigationoflegalproblems. . . Theinescapablefactisthatjurisprudencebearsthesamerelationtoamodernscienceofjurimetricsasastrologydoestoastronomy,alchemy to chemistry, or phrenology to psychology. It is basedupon speculation, supposition and superstition; it is concernedwith meaningless questions; and, after more than two thousandyears,jurisprudence has not yet offered a useful answer to anyquestionoraworkabletechniqueforattackinganyproblem.32

    Thisproposalspawnedanewfieldofstudy,someworkfromwhichisdiscussedbelow.33

    Loevingercallsthosewhofearthedangersofmechanizedjurisprudencequixoticanduncomprehending34thoughhewasnotamechanicaljurisprudenthimself.35 Some,however, sawthedevelopmentofcomputersasanopportunitytodevelopajudgmentmachine: amachinethatcouldreplaceajudge.

  • 7/29/2019 Pragmatik 1

    29/405

    9 1.2 Legalexpertsystems

    1.2.3 JudgmentmachinesIn1955,Lasswellpredictedthat:

    When machines are more perfect [sic] a bench ofjudicial robots. . . can be constructed. The machine would apply a system ofweightstoallegationsoffactmadebypartiestoacontroversy,andalsotothejustificationsadvancedinsupportoftheclaimsputforwardbyparticipants. Litigationcanproceedbycounselfortheplaintiff and the defendant pressing buttons that translate theircasesintothephysicalsignsbuiltintothemachine. Manyresultswouldbenodecision. However,themachinecouldbedesignedtosettleacontroversyofthiskindwitharandomoperation(bylot).36

    Hewasnotaltogetherserious:Itisachallengingtaskforlegalhistorianstoassistinconstructingarobotwhoseweightswouldgivesubstantiallythesameresultasthoseproducedbythe[USSupreme]Courtatvariousperiods. Thetask would not be too difficult for somejustices on some issues.Butarobotfacsimileofthe lessrepetitivemembersoftheCourtwouldprovideagenuinechallengetotheengineers.37However,somehavetakentheideaofajudgmentmachinevery

    seriously indeed. In 1949, Frank wrote that a logic machinemight disclose all possible available alternative legal rules, although judges would still have to exercise the the sovereignprerogativeofchoicebetweentherulesonthebasisofthejudgesconscious or unconscious notions of policy.38 Similarly Mehlwrote, in 1959, that a machine could perform some of the functionsofajudge,butthataroleforhumanswouldremainbecausethesolutiontoa legalproblemmaydependuponextra-rationalfactors, involvingthewholeofhumanexperience.39

    Asrecentlyas1977,ithasbeenseriouslysuggestedthatajudgment machine could replace a humanjudge. DAmato proposedthatamachinecouldtaketherelevantfactsofacaseasinputandproduceanumber intherange1to+1(whereapositivenumberindicatesavictoryfortheplaintiff). Giventhemultiplicityoffactors,heclaims,aresultofzerowouldbeextremelyunlikely,40although it isnotclearwhythiswouldbeso. These factswouldbe determined by ajury, but the law would be decided by themachine. Somewhatgrudgingly,heallows foravestigeofhuman

  • 7/29/2019 Pragmatik 1

    30/405

    10 Apragmatic legalexpertsystem 1.2

    control: anappealcourtcouldreviewallofthemachinesdeterminations in a certain numerical range (e.g. 0.05 to 0.05) withinwhichthecaseswouldbesoclosethatare-examinationmightberequired. The review courts subsequent decision would then beincorporatedintothesystem.41

    Theideaofhumanjudgesbeingreplacedbymachineshasbeentrenchantlycriticized. AccordingtoWeizenbaum:

    The very asking of the question, What does ajudge . . . knowthatwecannottellacomputer? isamonstrousobscenity. Thatithastobeputintoprintatall,evenforthepurposeofexposingitsmorbidity,isasignofthemadnessofourtimes.

    Computerscanmakejudicialdecisions. . . Theycanflipcoinsinmuchmoresophisticatedwaysthancanthemostpatienthumanbeing. Thepointisthattheyought notbegivensuchtasks. Theymayevenbeabletoarriveatcorrectdecisionsinsomecasesbut always and necessarily on bases no human being should bewillingtoaccept.42

    ButDAmatoseesadvantages inreplacinghumanjudgesbymachines:Would we lose ajudges judgment, and how important wouldsuchalossbetoourlegalsystem? Surelycomputersdonotmakejudgments the way humans do, and so we would lose the humanaspectoflegaljudgments. Butwhatspecificallydowelosewhen we lose the humanness ofjudgments? Is humanjudgment

    justaeuphemismforarbitrariness,discretion,orbias?43Bycontrast,Stonestressestheimportanceoflegalchangespringing from deviance, tentativeness, and even indecision injudgment.

    It is these phenomena above all which promote ajudges sensitivitytonewideasortonewlyperceivedsocialsituationsandstirHamlet-like introspection. On these there rest some of the mainfoundations

    of

    the

    social

    good

    we

    call

    justice.

    44

    Proponentsoftheideaofautomatedjudgesclaimthatsuchsystemswouldreducethecostofthelegalsystem,findinconsistenciesinthe law,andprovidea levelofcertainty inthe lawwhichdoesnotexistatpresent.45 DAmatoclaimsthatajudgmentmachinewouldallowpeopletoliveundertheruleoflawandnotundertheruleofpersons.46

  • 7/29/2019 Pragmatik 1

    31/405

    1.2 Legalexpertsystems 11

    DAmatosargumentisbasedupontwoboldinitialpremises.Heassumesthat,fromajurisprudentialpointofview,thelawhasbeen made completely determinable. He also assumes that, if acomputer can beprogrammedto makejudicial decisions, humandiscretionwillhavebeencompletelyremoved.47

    Hisviewsonjudgmentmachinesareextreme,andhavenotbeenwidely accepted. For the most part, designers of legal analysissystemshavefocusedonbuildingsystemswhichprovideadvicetolawyersortolaypeople,ratherthansystemswhichusurptheroleofjudges. However,manyofthesedesignershavemadeDAmatosfirstassumptionthatthelawiscompletelydeterminableoftentacitly.481.2.4 PetrifactionofthelawTheideaofajudgmentmachineremovinguncertaintyandambiguityinthelawraisesthepossibilityofthepetrifactionofthelaw.Thispossibility isalsorelevanttolegalexpertsystemdesign.

    DAmatoclaims

    that,

    once

    programmed,

    the

    law

    would

    become

    settled. The computer would stop progress, although the legislature could always step in if anomalous results were beingproduced.49 Similarly, Kayton believes that the use of propositional logic could identify legalambiguity andprovideguidelinesforitsresolution. Heasks:

    Wouldnoteachconcretedemonstrationofambiguityanditsresolution tend to rigidify the law and eventually destroy the veryflexibilitywhichhasmadethecommonlawviable? . . . Despitethedangers of generalization, it is submitted that . . . [this question]shouldbeanswered inthenegative. Informationbroughttobearinarationalpursuitisalwaysbetterthanignoranceorconfusion.Theeliminationofambiguities,ratherthanossifythelaw,wouldproduceanoptimumconditionforthepurposeful,intelligent,andefficientdevelopmentofthelaw.50

    Schubert,anotherearlyresearcherinthisfield,isnotconvincedofthepossibility,ortheimportance,ofcertaintyinlaw. Hewrites:

    . . . theidealofcertainty inlawistolerableonly inthecontextofanempiricalworldinwhichforcesinducingchangearesomanifoldthattheattainmentofthegoalisneverpossible.51Tyreearguesthatadeterministiccomputerjudgecouldover

    rule himself if cases are added as they are decided.52 But as

  • 7/29/2019 Pragmatik 1

    32/405

    12 Apragmatic legalexpertsystem 1.2

    Poundpointedoutalmostninetyyearsago, theproblem ismorethanmereossificationofthelaw:

    The effect of all system is apt to be petrifaction of the subjectsystematized. Perfectionofscientificsystemandexpositiontendstocutoff individual initiative inthe future,tostifle independentconsiderationofnewproblemsandofnewphasesofoldproblems,andtoimposetheideasofonegenerationuponanother.53This problem is inherent in expert system designin the law

    andinotherdomains. Therearedangers,forexample,intheuseof a legal expert system byjudges. Stone considers the use ofpredictivetechniquesinthelawandwarnsthat:

    . . . reliance at thejudgment seat on new predictive techniqueswouldseriouslythreatenthejudgescentralconcernwithjustice.Foriftheresultsthuspredictedforhimdoaffirmativelyguidehiminpresentdecisions,eachjudgewilltendtovotesomewhatmoreconsistentlywithhispastrecord. Themarginofdeviationwouldaccordinglydisappearindeferencetopredictedpatterns,aidedbythehumantendencyto followthe lessagonizingbecausealreadytroddenpath. . . 54

    Judges,Stonesays,haveadutytostrivewholeheartedlyforjusticeatthemomentofjudgment.55 Althoughanexpertsystemmaybe of some use tojudges of lower courts, there is a critical borderbetweenamachineservingthelegalorderandthedangersofsubversionofthatorder.

    For whenever an appellatejudge faces the duty to dojustice inthis sense, he must address himself to it with his own presentexperience and insights. It would be a corruption ofjustice forhim to shortcut this duty by resting on predictions of his futuredecisionsbasedonhisownpastbehaviorstillmoresoifthebasisistheaveragepastbehaviorofagroupofjudges.56

    So, as Stone points out, the pioneers and experts of the newtechniques must face the limits of the contributions they canmake.57

    1.2.5 ClearrulesandclearcasesH.L.A.Hartwasthemostfamousofthepositivistlegaltheorists.Hismajorwork,TheConceptofLaw,waspublishedin1961. Hartrealizes that because the law is expressed in natural language, it

  • 7/29/2019 Pragmatik 1

    33/405

    1.2 Legalexpertsystems 13

    issubjecttoconsiderablesemantic indeterminacy. Thishetermsthe open texture of law. However, he claims that it is possibleto use rules deductively to solve clear cases: that is those inwhich there is general agreement that they fall within the scopeofarule.58 Healsocontendsthat:

    . . . theresultoftheEnglishsystemofprecedenthasbeentoproduce,by itsuse,abodyofrulesofwhichavastnumber,ofbothmajorandminorimportance,areasdeterminateasanystatutoryrule. Theycannowonlybealteredbystatute,asthecourtsthemselvesoftendeclareincaseswherethemeritsseemtoruncountertotherequirementsoftheestablishedprecedents.59AsSusskindwrites,oneramificationofHartsanalysisisthat:. . . allexpertsystemsinlawwhoseinferenceproceduresaresolelydeductive will function exclusively in the clear case domain, andwillbeofnoaidinthesolvingofproblemsofthepenumbra.60

    But what constitutes a clear case is itself far from clear. AsHartconcedes:

    . . . it isamatterofsomedifficultytogiveanexhaustiveaccountofwhatmakesaclearcaseclearormakesageneralruleobviousanduniquelyapplicabletoaparticularcase.61Molesdisputestheveryexistenceofclearrulesandclearcases.

    He performsa detailed analysis oftheapplicationofan exampleofanostensiblyclearrule inaBritishstatute: aprovisionwhichprescribesthecircumstances inwhichan injunctionshouldbe issued in domestic violence cases.62 He summarizes the effect ofthreecasesasfollows:

    (1)BeforeB v. B63 weareconcernedwithastatutoryprovisionwhichonitsfaceappearstobeclearandcomprehensible(atleastto non-lawyers) and it would appear that an injunction shouldissueincasessuchasthosewehavelookedat.

    (2) After Cantliff,64 the provision has been considered twiceby the [English] Court of Appeal within two weeks. All of thesixjudges who considered the matter are in agreement that theinjunctionshouldnotissueandthattheruleisclear.

    (3)Thematter is furtherconsideredbyaspeciallyconstitutedCourt of Appeal of five,65 who decide that the rule is clear, butdifferentfromthatin(2),andthattheinjunctionshouldissue.66

  • 7/29/2019 Pragmatik 1

    34/405

    14 Apragmatic legalexpertsystem 1.2

    Moles also analyses thejudicial application of the rule of precedentwhich, with this statutory rule, must be regarded asamongsttheclearestavailabletous67todemonstratethatourexperience of the legal system bears little relationship to Hartsaccountofit.68

    Leithdevelopedalegalexpertsystemwhichoperatedonclearrulesinthelaw.69 Helaterrecanted,sayingthat:

    . . . thevery ideaofaclearruleis inherentlyconfusingand isnotobservableintherealoperationofthejudicialprocess. . . judicialcreativity is not an aberation [sic] of the legal process, but (asMoles. . . suggests),theveryheartofthelaw.70

    Asisexplainedbelow,71 manyexpertsystemdesignershaveadoptedaviewofthelawthatallowsforclearrulesandclearcaseswithout considering the possibility that the law may not be likethat at all. 1.2.6 LegalrealismandrulescepticismInthe1920sand1930s,amovementcalledlegalrealismdevelopedin America.72 Realists rejected the importance that mechanical

    jurisprudenceplacedonrules. However,therulescepticismoftheAmerican Realists is meek by comparison with the strong indeterminacythesis. DrahosandParkerexplainthat:

    According to Karl Llewellyn a characteristic feature of Realismwas the rejection of simple (by which he meant general) rulesandthesubstitutionofmoredetailedclassificatoryschemeswhichbettercapturedthespecificnatureofjudicialrule-making. . . OnLlewellyns account, rule scepticism emerges as a set of doubtsabouttheveracityoflegalactorsclaimstobefollowingthelegalrulestheysaytheyare. This is notrulescepticism inthestrongsense of denying the existence of rules, however. Legal actorsmaysimplybefollowingsomeotherrules. TherulescepticismofAmericanRealismcouldperhapsbemoreaccuratelydescribedasrulecynicism.73Kripke, by contrast, is a true rule sceptic. His argument is

    an example of the Wittgensteinian Paradox.74 Consider the twofunctionsplus(+)andquus(). The+functionisthemathematicalfunction,addition. Thefunctionisdefinedasfollows:

    x+y, ifx, y

  • 7/29/2019 Pragmatik 1

    35/405

    1.2 Legalexpertsystems 15

    SupposethatKripkehasused+inthepast,butalwayswithvaluesofxandysmallerthan57. Heperformsthecomputation68+57andgetsaresultof125. Yetitcouldbesaidthatwhenhethoughthewasusingtheplusfunctioninthepast,hewasinfactusingquus. AsKripkeexplains:

    . . . inthisnewinstance,IshouldapplytheverysamefunctionorrulethatIappliedsomanytimes inthepast. Butwho istosaywhat function this was? In the past I gave myself only a finitenumberofexamplesinstantiatingthisfunction. Soperhapsinthepast I used plus and + to denote . . . quus . . . Who is to saythat[]isnotthefunctionIpreviouslymeantby+?75

    There is nojustification for Kripke answering 125 rather than 5.Thereisnowayofdetermining(fromhispastbehaviourevenhispastthoughts)whetherbyplusKripkemeant+or. Rulesarederivedfromafinitenumberofexamples;foranygivenrule,thereisalwaysanalternativerulewhichalsoexplainsthoseexamples.

    Mathematicians could counter that the meaning of + is welldefined. ButKripkearguesthat:

    . . . scepticismaboutarithmeticshouldnotbetakentobeinquestion: wemayassume,ifwewish,that68+57is 125. . . Icannotdoubtcoherentlythatplus,asInowuseit,denotesplus! PerhapsIcannot. . . doubtthisaboutmypresent usage. ButIcandoubtthatmypast usageofplusdenotedplus. . .

    . . . Thereisnoobjectivefactthatweallmeanadditionby+,oreventhatagivenindividualdoesthatexplainsouragreementin particular cases. Rather our license to say of each other thatwemeanadditionby+ispartofalanguagegamethatsustainsitselfonlybecauseofthebrutefactthatwegenerallyagree.76DrahosandParkerwritethat:Theconsequencesofthisargument,ifvalid,areshattering. Rulesturnouttobenomorethanleapsinthedarkandthewholenotionofrulefollowingseems illusory.77The implications of Kripkes argument have been realized by

    legaltheorists.78 AsDrahosandParkerpointout:Saying that there are no rules to follow, only social practices,means that propositions about the law are potentially open towildfluctuations. Theargumentalsoputspaidtoanypossibilityofacorrespondencetheoryoftruth inlaw.79

  • 7/29/2019 Pragmatik 1

    36/405

    16 Apragmatic legalexpertsystem 1.2

    Theyproposeasolutiontothisproblem:TheKripkeanargumentdoesnotpreventpeoplefromsaying thatthey are following rule X. Rather it stops them from being abletojustifytheexistenceofthatrulebyreferencetosomeobjectivemeaning. Thisstillleavesthepractice,asopposedtothejustification,ofsuccessfulrulefollowingtobeaccountedfor.80

    Their solution is to view the law as a set of rules and conventions: atypeofruleusedtofixorinterpretthemeaningofotherrules.

    81Theyclaimthereisadistinctionbetweenruleknowledgeand rule understanding, the latter being a matter of absorbing

    conventionsrelatingtoruleuse.82 Alawyerrequiresknowledgeoftherules,andunderstandingastohowtoapplythoserulesproperly. DrahosandParkerarguethattheproblemsfacingdesignersof legal expert systems flow from the difficulties of representingrule understanding rather than rule knowledge.83 Rules can beused to represent rule knowledge; the problem is how to represent with tolerable accuracy . . . conventions which confer ruleunderstanding.84

    Buttheyconcedethat,aswithotherrules,conventionshavetorunthegauntletofscepticism.85

    1.2.7 Ajurisprudentialconsensus?The only major examination of the role ofjurisprudence in thedevelopment of legal expertsystems isSusskinds 1987 bookExpert Systems in Law: A Jurisprudential Inquiry.86 As discussedabove,87 hearguesthatallexpertsystemsmustconformtosome

    jurisprudentialtheory.Healsosetsouttofindconsensusinjurisprudentialtheory. As

    thereisfartoomuchliteratureinthefieldforonepersontocoverit all, he limits his choice of material: the vast majority ofsources that he uses are British analyticaljurisprudential writingssincethemid-1950sandearly1960s,theimpetusforwhich,he concedes, was derived very largely from the publications ofH.L.A.Hart.88

    Susskindconcludesthattherearenotheoreticalobstacles,fromthe point of view ofjurisprudence, to the development of rule-based expert systems in law of limited scope.89 He also claims

  • 7/29/2019 Pragmatik 1

    37/405

    1.2 Legalexpertsystems 17

    that the divergence of views withinjurisprudence has been overstated because legal theorists tend to focus on the differences.He claims that there is a jurisprudential consensus, albeit ofmundaneandlimitedapplication.90

    Susskind provided the theoreticaljustification for not only hisowndevelopmentwork,91 butformuchofthesubsequentworkofother developers of legal expert systems. He could also be saidtohaveprovided,retrospectively,legaltheoreticaljustificationformostoftheworkonrule-basedexpertsystemsthatprecededhisbook.

    However,Susskindsapproachtojurisprudenceisfundamentallyflawed. AsMolespointsout:[Susskind] said that he would carry out a survey of thejurisprudential literature. Heacknowledged, ofcourse, that itwouldnotbe possible to survey the whole of thejurisprudential literature.Infact,hedeterminedthatlawwasasystemofrulesbysurveyingonlythosewhoseavowedpositionwasbasedonthefactthatthe lawwas asystemofrules. . .

    I would venture to suggest that this is in fact a misuse of thesurveytechnique. . . SusskindwasperfectlyfamiliarwiththeworkofHartandhisfollowers,andwaswellable,therefore,tofindanynumberofbooksandarticleswhichsupportedthelawasrulesview. Hethendevelopedhisposition. . . onthebasisofwhatthispurportedconsensus withinjurisprudencehadtosay.

    . . . Of course, Susskind was telling certain sections of the AIand Law community what they wanted to hear, and hence theirenthusiasmfor it.92

    Moles also notes Susskinds admission that the most rigorousof these writings constituted the source materials with greatestpotential given the overall purpose of the project.93 Becausethe first objective of Susskinds work was to design, develop andimplement an expert system in law,94 Moles cites this admissionasevidencethatSusskindprejudgedhissurvey.

    Inthe lightofthiscriticism, it isironictonoteSusskindsowncautionthat:. . . alittlejurisprudentialknowledgecanbeadangerousthing! Itistemptingforthejurisprudentialneophytetobecomeanardentdevoteeofaparticularschoolofthoughtwithin legaltheoryandto go on from there to implement all and only the teachings ofthatschool. Thiscourseofactionshouldbeavoidedatallcosts.

  • 7/29/2019 Pragmatik 1

    38/405

    18 Apragmatic legalexpertsystem 1.2

    Familiarity with a wide range of works should be achieved priortocommitmenttoanyparticularjurisprudentialposture.95Clark,anothercriticofSusskind,saysthat:Theshortcomingsof[his]bookcoincidewiththelimitsofpositivistlegaltheory,andeventhensomedoubtmustremainastowhetherexpert systems in law mark a revival of the kind of mechanical

    jurisprudencewhichHartopposedsovigorouslyfromwithintheconfinesofpositivistlegalthought.96

    ButthestrongestcriticismofSusskindcomesfromLeith:. . . [Susskind]believesthathecancometosomesortofcompromise with the various theoretical positions taken by the renownedthinkersofthefieldandproduceageneraltheory(anindication,Imightsuggest,of[his]poortheoreticalconceptions). . .

    . . . Susskindseesjurisprudenceasprovidingavarietyoftheoreticalmodelswhichcanbemodelledmathematicallyandtranslatedintocomputerprograms.

    . . . [O]f necessity formal specification requires a formalisationoflaw;therecanbenoinformalmodelswhicharemysteriouslyformalisedintoacomputermodel. . .

    . . . Therefore, in order to use formal specifications, Susskindmust provide a formal specification of law which can then be incorporated intotherule-formatofacomputerprogram . . . And,ofcourse,formalspecificationsof lawarerenownedfortheirtheoretical and practical inadequacy. Legal formalism can, surely,hardlybethecompromisehewishesmightarisefromtheconflictingpositionsofKelsen,Hart,Dworkinetal.97

    . . . [I]fhereallydoesbelievethathisinformaltheoreticalmodelscanbetransformedintoformaltheoreticalmodelswithoutlossoftheir informal attributes, then I must suggest that he has reallylittleunderstandingofthedisciplineofcomputerscience.98Susskindsclaim to have found a consensus injurisprudence

    even one of mundane and limited applicationis absurd. Asshown above,99 the views of just two jurisprudents (Hart andMoles) are completely irreconcilable. Similarly, Harts views aretotallyatoddswiththoseofKripke.100 Susskindstateshistheory,anddevelopsanexpertsystemwhichconformstothattheory. Hismistakeistoclaimthathistheoryisdefinitive,inthatitrepresentsajurisprudentialconsensus.

  • 7/29/2019 Pragmatik 1

    39/405

    1.3 Legalexpertsystems 19

    1.3 JurimetricsandthebehaviouristsAsmentionedabove,

    101thetermjurimetricswascoinedbyLoevinger. He defines jurimetrics as the study of law and legal

    problems by scientific methods and concepts, the employment ofscienceinlawtotheextentthatitisapplicableoradaptable.102Loevingerisneitheramechanicaljurisprudentnorapositivist:

    Tobeginwithwe must beclearthat science offers us neitherultimate nor certain answers to legal problems. The dream thatsciencemightsomedaytelluswhichofseveralcompetinginterestsis the more important is a vain one. Science essays no such answersinanyfield. Sciencedoesnotassignsocialorethicalvalues.Science may, indeed, provide data from which social or ethical

    judgmentsmaybemade;butthejudgmentswillremainwithman. . .

    Thereisnoprospectofanyprocessthatwillprecludeconsiderationofsocialdesirabilityorwisdom. Theopportunitywillalwaysbe available to argue that precedent should not be followed, andthat considerations of policy, or expediency, require a differentruleoraspecialresult. . . . . . Sciencedoesnotandwillnotofferusanylawmachinesthatgiveautomaticanswerstospecificquestionsputtothem,whetheras to particular cases or as to ultimate legal issues such as therelative importance of interests that may be in conflict. By thesame token, science will provide us with no formulae or calculusthatwillgiveuscertaintyeitherofprediction,analysisoranswerstoultimatequestionssuchaswhich interest istobepreferredorwhichdesirehasgreatersocialvalue.103

    Althoughthesecommentsareeminentlyreasonable,someofLoevingersturnsofphraseareexasperating. In1949heclaimedthatputtingthelawonarationalbasiswastheindispensableconditionofthesurvivalofthehumanrace.104 Inthelightofsucharidiculousstatement,andafterhistrenchantcriticismofthevalueofjurisprudence (quoted above105), it is not surprising that, asGardnersays:

    The attitudes Loevinger and his colleagues expressed were neveradopted by the legal profession generally . . . As a movementwithin the legal profession,jurimetrics has not been much heardfromsincetheearly1970s.106

    Nevertheless,someoftheworkofhiscolleagues,applyingvariousstatistical and logical techniques to legal analysis, is worthy ofcommenthere.107

  • 7/29/2019 Pragmatik 1

    40/405

    20 Apragmatic legalexpertsystem 1.3

    1.3.1 KortKort

    uses

    mathematical

    expressions

    to

    represent

    American

    judicial

    decisions.108 Heidentifiesattributeswhichareofimportanceandrepresentsdecidedcasesinthefollowingform:

    n

    Aijwi =Vji=1

    where n is the number of attributes, Aij is the value of the ithattribute for thejth case (attribute values being 0 or 1), wi isthe weight of the ith attribute, andVj is the number of votes of

    judgesfavourabletothepartyseekingredressinthejthcase. Thisapproachisjustifiedonthebasisthatthedecisionsofsplitcourtsdonotnecessarilyconstitutetwooppositeextremes,butrepresentcertaindegreesofsupportforoneparty.109

    By solving these simultaneous equations, weights are obtainedfor the attributes. The sum of the weights of those attributespresent in the instant case is, according to Kort, the number oflikelyjudicialvotesinfavourofthepartyseekingredress.1.3.2 LawlorLawlor uses logical expressions to represent previously decidedcases.110 He analyses right-to-counsel cases heard before the USSupremeCourtover30years, andfindsthat inallofthesecaseseachjustice behaved consistently with his own personal staredecisis. He identifies legally significant attributes and builds alogical expressionwhichrepresentsthebehaviourofeachjustice.Aprogramusestheselogicalexpressionstopredictthelikelyoutcome,givenacompositionoftheCourtspecifiedbytheuser.

    LawlorssystemsuccessfullypredictedtheUS SupremeCourtsoverrulingofBettsv.Brady111 inGideonv.Wainwright.112 Butitpredicteda5:4majority;theSupremeCourtsdecisionwasunanimous.

    Applyingtraditionalstaredecisis(i.e.consideringthecourtasawhole,ratherthanthedecisionsofindividualjustices)Lawlorssystem did not predict the Supreme Courts change. Given thesamecases,Kortsapproachdidnotpredictthechangeeither.

    Thisfailuredrawscriticismfromthosewhodisagreewithsuchalogarithmicapproachtojustice.113 Wienercomplainsthatadvocates ofthecomputerresttheirargumentsonanassumption

  • 7/29/2019 Pragmatik 1

    41/405

    1.3 Legalexpertsystems 21

    thatcourtswilladheretothedoctrineofstaredecisis,whichdoesnotalwayshold.114

    Indefenceofthebehaviourists,KaytonreferstotheirfailuretopredictthedecisioninGideonv.Wainwright andwrites:Shouldwehaveexpectedotherwise? Ofcoursenot! Areversal isbydefinition a logical inconsistency. That which by stare decisishadbeencalledblackisnowcalledwhite.115

    Thedevelopersofthesepredictionsystemsacceptthattheyrelyonthedoctrineofprecedent. Hence,areversalofpreviousauthoritywillalwaysbebeyondthepredictivecapacityofthesesystemsasitisbeyondthepredictivecapacityofmostlawyers.1.3.3 NagelandSchubertNagel and Schubert116 examine the personal attitudes of US Supreme Courtjustices towards various political and economic relations, and claim that these off-the-bench attitudes (as Nagelcalls

    them)

    affect

    the

    justices

    decision-making.

    Correlationswere. . . madebetweenresponsestospecificitemsandvariousdecisionalpropensities. Forexample,therewasahighandstatistically significant correlation between disagreeing with [onequestionnaireitem](Ourtreatmentofcriminalsistooharsh;weshould try to cure, not to punish them) and being above theaverageofonescourtwithregardtotheproportionoftimesonevotedfortheprosecutionincriminalcases. Off-the-benchjudicialattitudes

    thus

    do

    seem

    to

    correlate

    in

    a

    meaningful

    way

    with

    on-the-benchjudicialdecisions.117

    1.3.4 Haar,SawyerandCummingsIn the mid-1970s Haar, Sawyer and Cummings made use of regression analysis to build a predictive model for zoning amendment cases in Connecticut.118 Regression analysis is a statisticaltechnique for analyzing the relationship of a set of independentvariablestoadependentvariable. ForHaaretal.thedependentvariableistheoutcomeofthecases,theindependentvariablesareattributes.

    Theyidentify167attributeswhichappearedtobeimportantto the courts, 40 of which were deemed significant using a 2(chi-square)test forassociation. This istoomanytouse inaregressionanalysis,119 soHaaretal.employtwodifferentmethods

  • 7/29/2019 Pragmatik 1

    42/405

    22 Apragmatic legalexpertsystem 1.3

    toreducethisnumberfurther: groupingattributesonthebasisofexperience,knowledge,andintuition,120 andfactoranalysis.1.3.5 TheroleofpredictionThebehaviouristspredictiveresearchhasbeencriticizedbecausethey do not attempt to model legal reasoning.121 But, in notso doing, the behaviouristsjust reflect the influence upon themof the American Realists. They adopt Holmess analysis thatfor any individual the law is simply a prediction of the way inwhich

    the

    public

    force

    possessed

    by

    the

    government

    will

    act

    upon

    him.122 Loevingerclaimsthatsomemethodoflegalpredictionisindispensable;123 Lawlor says that the ultimate goal of all scientificmethodsisreliablepredictionoffutureeventsand[r]eliableprediction is also one of the ultimate goals of law.124 Susskind(certainlynobehaviourist)alsostressestheimportanceofthepredictionofjudicialdecisions.125

    Tapperwritesthat:Although the statistical techniques employed by some of theseworkershavebeencriticized,therecanbenorealdoubtthatthisworkprovidesasuccessfulapproachtotheanalysisofthedecisionswhichhavebeenreached inthepast,andat leastassatisfactorya method of predicting future decisions as can be arrived at bynativewitandunaidedintuition.126

    However,hecriticizestheconclusionsthathavebeendrawnusingthesetechniques:

    Either the behaviourist is to content himself with observing theobjectivephenomena,inwhichcasehecanconcludenothingastomotivation, or he is to ascribe motivation to the phenomena, inwhich case he ceases to be objective. It is precisely at the pointat which decisions of the [US] Supreme Court, for example, arecharacterizedaspro-civillibertiesoranti-labourthatdoubtsariseastotherealobjectivityofthestudies.

    The same general line of argument may be advanced againstthefact-orientedapproach . . . Herethepointatwhichthestudyloses itsobjectivity is inthecharacterizationofthe factspresentinthecase.127

    (Althoughthismaybeavalidcriticismofabehaviouristicstudy,itisnotanargumentagainsttheuseofpredictivetechniqueswherethe characterization of facts has been performedsubjectively,admittedlybyalegalexpert,asisdoneforSHYSTER.)

  • 7/29/2019 Pragmatik 1

    43/405

    1.3 Legalexpertsystems 23

    Gardnerisalsocriticalofthejurimetricprograms:. . .

    programs,

    done

    from

    a

    political

    science

    viewpoint

    rather

    than

    a legal one, in which the data concern legally irrelevant matterssuchastheideologyandsocialbackgroundofindividualjudges.128

    By legally irrelevant matters Gardner means those matters towhichjudges do notand/or should notexplicitly have regardwhencomingtotheirdecisions. Butifasystemisdesignedtopredictdecisionsthenanyattributewhichassistsinpredictionshouldbeincludedregardlessofwhetherregardought tobehadtothatattribute. These legally irrelevant matters are not necessarilyirrelevanttothepredictionofjudicialdecisions.

    Stonedismissesconcernaboutjurimetricprediction: thebehavioralists [sic] may have no conscious designs on the integrityof the decisional process. Thejudgment ofjustice may be of noconcerntothem.129

    Hedefinesthejudgmentofjusticebyreferencetoasituationwhereajudgemakesadecisionwhich:

    . . . does not merely declare the existing law but decides whatjustice requires that the law should be. (. . . [W]e include tacitlyhere the reversal of earlier decisions, that is, creative decisionswhich unsettle and resettle law: this is afortiori creative.) Itis this kind of creative judgment which we have here termed a

    judgmentofjustice.130Thebehaviouristsare,Stonesays:

    . . . onlyobserverslookingatwhathasalreadybeen done injudgment. And when in the course of prediction they turn theirattention to futurejudgments, it is to ask not what thejudgeshoulddotofurtherjusticebutonlywhatkindofdecisionhewillgive if he acts consistently with values attributed to him on thebasisofhispastdecisions.131

    Inotherwords,thebehaviouristsworkattemptedtobepredictivenotnormative.

    When a lawyer gives advice, she/he is expected to make somepredictionastothelikelyresult. Thispredictionneednotbeemphatic: it may be no more than a tentative statement about thestrengthofapersons legalposition. Hence, legalexpertsystemsshould have some degree of predictive capacity. Like the behaviourists research, a legal expert systems predictive power is a

  • 7/29/2019 Pragmatik 1

    44/405

    24 Apragmatic legalexpertsystem 1.4

    projectionbasedonpastcases. There isandcanbenoallowance for changing social mores. This is not a serious limitationupon such a systems utility: predicting ajudge-made change inthelaw isbeyondallbuttheverybestlawyers.

    However, as discussed below,132 thejustification that a legalexpertsystemprovidesforitspredictionisalsoimportant.1.4 Rule-basedsystems

    Thereare

    many

    examples

    of

    rule-based

    legal

    expert

    systems,

    the

    most importantofwhich are discussed in detailbelow: theworkof McCarty; Bench-Capon, Kowalski and Sergot; Gardner; andSusskind.

    Thefirstproposalforarule-basedlegalexpertsystemwasmadebyBuchananandHeadrick in1970. Theycomplainedthat interdisciplinary work between lawyers and computer scientists hadfloundered on the misconceptions that each has of the othersdiscipline,133 andsuggestedthatthecomputermodellingoflegalreasoningwouldbeafruitfulareaforresearch.134

    Their proposal asserts that [i]n the absence of any reason tospeculate on how they carry on their work, [lawyers] now apply complex sets of rules without being aware ofthe rules themselves.135 Theymakenoreferencetojurisprudentialwriting,butthey do make it clear that their approach is based upon two assumptionsabouthumanproblem-solvingingeneral: (1)problemscanbebrokendownintoasetofsubproblems,and(2)thesolutiontoanysubproblemrequiresaseriesofdecisionsthataregovernedbydecisionrules.136

    Theliteratureisrepletewithexamplesofprojectsinwhichtheassumption that lawyers work with rules is unstated and/or unsupported. MaggsanddeBessonet,forexample,tacitlymakethisassumption.137 The law, they claim, can be expressed as rulesusingpropositionalcalculus. Aprogramwhich implementstheserulescouldanswerquestions fromauser, andallowthecheckingof statute law for redundancy and contradiction. (It is not clearwhatrolecaselawplaysintheirsystem,orindeedintheirmodelofthelaw.)

    Popp and Schlinks JUDITH system138 uses rules to representparts of the German Civil Code. The principles behind JUDITHarestrikinglysimilartothoseoftheMYCIN system(anexpertsystem dealing with bacterial infections)139so similar that, its

  • 7/29/2019 Pragmatik 1

    45/405

    1.4 Legalexpertsystems 25

    developers claim, it would be possible to create a legal knowledgebaseforMYCINandamedicalknowledgebaseforJUDITH.140Michaelsen and Michies TAXADVISOR system141 is implementedusing MYCIN.142 TAXADVISOR is designed to assist lawyers toadviseclientsontaxationandestateplanning.143

    Meldmanssystemusestwodifferentkindsofrules: generalruleswhich definetheelements ofthe claim, andspecificrulesextracted fromcases.144 Things andrelations areusedtorepresenttheeveryday world of human affairs,145 and are classified hierarchically intocategories. Afact comprisestwothingsandarelationbetween them; facts are assembled into situations. These situations are compared with the situation of the instant case, andthe system determines the extent to which the instant case fallswithin or near the law of intentional torts (e.g. assault and battery).

    WatermanandPetersonsLDSsystem146 isarule-basedsystemforthefieldofproductliability. Itdoesnotdeterminewhetherliability exists, but is designed to assist legal experts in settlingproduct liability cases. LDS is a typical example of a systemwhichdevelopedwithoutanylegaltheoreticaljustification. Afteradmitting that there is no deep model of the legal process,147Watermanandhiscolleaguesproceedontheunstatedassumptionthatanysuchmodelmustinvolverules:

    Onemightexpectthatthelargebodyoflegalrulesandregulationsthat have been accumulated and formalized in the legal domainwould make expert system development easier. Unfortunately,this is not the case. Instead, this characteristic of the domain,having rules that already exist, has led to trouble . . . First, theformal rules thatdefineand regulate legal activity are oftenambiguous, contradictory and incomplete. Andsecond, there existsa body of informal rules or procedures about how to access, interpret and use the formal rules. Without these informal rulesthe formal rules can not be used in any efficient or cost-effectiveway.148

    Thisbodyofrules,theywrite,needstobemappedintocode149beforealegalexpertsystemcanbebuilt.

    BingsSARA system150 isdesignedtoanalysediscretionarydecisions. It uses rules to represent legal norms: some strict andsome discretionary. These discretionary norms are weighted usingcorrelationtechniques. SARA allowsa lawyertobackuphisqualitativelegalreasoningbyquantitative indications.151

  • 7/29/2019 Pragmatik 1

    46/405

    26 Apragmatic legalexpertsystem 1.4

    StampersLEGOL language152 alsousesrulestorepresent legalnorms. PattisonandCiesielskiusearule-basedsystemtoreviewcontracts.

    153SoftLaws STATUTE, acommercially successful system,usesrulestorepresentstatutes,regulationsanddepartmental

    guidelinesintaxation,socialsecurityandveteransaffairslaw.1541.4.1 McCartyMcCartyhasbeendescribedasthefatherofAIandlaw.155 HisTAXMAN project156 isconcernedwiththedevelopmentofacomputationaltheoryoflegalreasoning,usingcorporatetaxlawasanexperimentalproblemdomain.

    Heclaimsthatacomputer-basedlegalconsultationsystemmustbe able to represent the facts, at some comfortable level ofabstraction, and the law, which would consist of a system ofconcepts and rules. These concepts and rules are relativelyabstract(i.e.theysubsumelargeclassesof lower-level factualdescriptions),andtheyhavenormativeimplications(i.e.theyspecifywhich

    actions

    are

    permitted

    and

    which

    are

    obligatory).

    Legalanalysis,initssimplestform,wouldthenbeaprocessofapplyingthelawtothefacts. Putthisway,theparadigmseemsto be an ideal candidate for an artificial intelligence approach:the facts would be represented in a lower-level semantic network, perhaps; the law would be represented in a higher-levelsemantic description; and the process of legal analysis would berepresentedbyapattern-matchingroutine.157

    However, McCarty concedes, the representation of facts in sucha system is more difficult than in other expert systems, becausethe facts of a legal case typically involve all the complexities ofdailylife: humanactions,beliefs,intentions,motivations,etc.158Evenifthefactscanberepresented,hewrites,theruleswilloftenbeproblematic:

    Some rules, usually those embodied in statutes, have a preciselogical structure, and this makes them amenable to the existingartificial intelligencetechniques. But it isacommonplaceamonglawyersthatthemostimportantlegalrulesdonothavethisformat all: instead they are said to have an open texture; theirboundaries are not fixed, but are constructed and modifiedas they are applied to particular factual situations. A sophisticated legal consultation system would not be able to ignore thesecomplexities,butwouldhavetoaddressthemdirectly.159

  • 7/29/2019 Pragmatik 1

    47/405

    1.4 Legalexpertsystems 27

    McCarty also makes the startling claim that the simplestproblemsforfirst-yearlawstudentsarethehardestforanAIsystembecausethestudentdrawsuponordinaryhumanexperience:

    Paradoxically, the cases that are most tractable for an artificialintelligence system are those cases, usually involving commercialandcorporatematters,whichalawyerfindsmostcomplex. Thereis a simple reason why this is so. A mature legal system in anindustrializeddemocracy is composed of many levelsof legal abstractions. . . Becauseoftheirtechnicalcomplexity,thelegalrulesatthetoplevelsofthisconceptualhierarchyaredifficultformostlawyerstocomprehend,butthiswouldbenoobstacleforanartificialintelligencesystem.160McCartychosetheareaoftaxlawforhissystembecausecom

    mercial abstractions, in fact, are artificial and formal systemsthemselves,drainedofmuchofthecontentoftheordinaryworldand, by legal standards, well structured.161 The field of corporate tax law, he says, is very near the apex of the hierarchy ofcommercial abstractions162whatever that means, commentsMoles.163InTAXMANI,McCartysfirstprototypesystem,thebasicfactsofacorporatecasearecapturedinarelativelystraightforwardrepresentation(e.g.acorporation issuessecurities). Belowthis levelis an expanded representation of the meaning of various entities(e.g. a security interest) in terms of their component rights andobligations.

    Above

    this

    levelpresumably

    above

    both

    levels,

    althoughthisisnotmadeclearisthelaw(statutoryruleswhich

    classifytransactionsastaxableornon-taxableetc.).164McCartyfoundthat,althoughtherulesarecomplex,theunder

    lyingrepresentationsaremanageable. Heconcludesfromhisearlywork that the construction of an expert consultation system inthisareaofthelawisafeasibleproposition.165

    McCarty sees the development of legal expert systems as anopportunity

    to

    contribute

    to

    jurisprudence.

    Although

    the

    jurisprudential literature includes many illuminating examples and

    manyvaluableinsightsaboutthestructureanddynamicsoflegalconcepts,hecomplainsthattakenasawholeitisnotoriouslyimprecise.166

    TheTAXMANsystemaddsastrongdoseofprecisionandrigortothesediscussionsoflinguisticandconceptualproblems. Itscritical

  • 7/29/2019 Pragmatik 1

    48/405

    28 Apragmatic legalexpertsystem 1.4

    task is to clarify the concepts of corporate reorganization law insuchawaythattheycanberepresentedincomputerprograms.167Molesisverycriticalofrule-basedexpertsystemsdesigners,and

    McCartyinparticular. McCartytakesapositivistapproachtothelaw,andMolescomplainsthat:

    Thecomputerscientistshavetakenthede-humanizingaspectsofmodernpositivismtotheirextreme. . . .

    Thecomputerscientists,encouragedbythemodernpositivists,fail to recognize the point which Austin168 correctly emphasizedthroughout his workthat law, positive morality and ethics areinseparablyconnectedpartsofavastorganicwhole.169

    MostdamningisMolessstatementthat:Thesadthingisthat[McCarty]hasnotshowntheslightestawarenessofthenatureofthelegalenterprise. Farfromhavingemphasizedanydifficulties,heshowsthathesimplydoesnotunderstandwhattheyare.170McCarty

    does

    identify

    two

    major

    limitations

    to

    the

    approach

    takeninTAXMANI. Firstly,heconcedesthatthefactualdescriptions, although manageable in the corporate domain, would betoo complex in (for example) the average contract or tort problem. Secondly,thehigher-levelconceptualrepresentationsarenotadequate for all domains becausejudicially created concepts areincurably open-textured, have a dynamic structure with the capacity to evolve and adapt to new situations, and the evolutionof these concepts is governed by a sense of purpose.171 Havingrecognized these problems, McCarty set out to solve them withTAXMANII.TAXMANII172 usesprototypesanddeformationstorepresentalegalconceptbyspecifyingtheprerequisiteconditionsforthatconcept,asetofcases(realandhypothetical) inwhichthatconceptdoes or does not apply, and a set of transformations for gettingfrom one case to another. If a given case (representing a legalconcept)canbetransformedintotheinstantcase,whilestillsatisfyingtheprerequisiteconditions fortheconcept,then itcanbeusedinargumentasanexampleofthatconcept.

    AshleyseesMcCartysworkasasignificantadvance,butpointstoseveralshortcomingsofTAXMANII:

    Thework is largelyanexerciseinknowledgerepresentation. Mc-Cartydoesnotset forthacontrolorprocessmodelthat clarifies

  • 7/29/2019 Pragmatik 1

    49/405

    1.4 Legalexpertsystems 29

    howaprogramwouldactuallygeneratealegalargument. . . Thereportedresearch involvesahandsimulationofthearguments inone

    US

    Supreme

    Court

    case

    173and

    work

    done

    on

    hand

    simulations

    ofseveralsubsequentcases.

    . . . [TAXMAN II] has no mechanisms for comparing cases intermsofhowonpointtheyare,fordistinguishingcases,orselectingthebestprecedents. . .

    McCartys model assumes a much neater domain than existsin law. He assumes that in reality, legal cases are consistentlyallocated as positive and negative exemplars of concepts. Theyarenot. Heassumesthatthereisanearmatchbetweenconceptsandthefeaturesofacasethatarerelevanttotheconcept. Thereisnot.174

    McCartys definition of legal primitives has also been criticized.Moleswrites:

    McCartyappearsnottoappreciatethatcorporations,securities,property, dividends and so on are not subsumed beneath thelaw, but are each the products of complex legal analysis. Thequestionofwhethercertaintransactionsaretaxableornot is intimatelytiedintothatlegalanalysis.175

    Ashleymakesasimilarpoint:The problem with such primitives, if they are taken seriously asa means for definingconcepts, is thattheyassume what isto beshown. Farfrombeingaprimitive,thatsomeonehasarightoradutyinagivenfactsituationisanarguablelegalconclusionthatmustbejustifiedbycitingauthorities.176

    1.4.2 Bench-Capon,KowalskiandSergotBench-Capon,Kowalski,Sergotandtheircolleaguesuseprogramswritten inthePROLOG languagetomodelstatutes.177 Theirapproach to legislation is the most extreme of all expert systemdevelopers, due to their attitude towards knowledge acquisition.Theywrite:

    The formalisation of legislation by means of rules has almost allthecharacteristicsofanexpertsystem. Itdiffers,however,inoneimportantrespect. Inaclassicalexpertsystem,beforeknowledgecan be formalized, it has to be elicited from the subconscious ofan expert. Eliciting this knowledge is generally regarded as themain bottle-neck in the construction of expert systems. It is entirely absent, however, in the case of legislation which is already

  • 7/29/2019 Pragmatik 1

    50/405

    30 Apragmatic legalexpertsystem 1.4

    formulatedandwrittendown. Thustheuseofexpertsystemtechniquesforrepresentinglegislationhasvirtuallyalltheadvantagesof

    expert

    systems

    without

    the

    attendant

    disadvantages

    of

    eliciting

    theknowledge.178

    This statement is nothing less than astounding. Even if it is acceptedthatstatutelawcanberepresentedusingPROLOGclauses,itisaboldclaim indeedtoassertthatconstructingthoseclausesrequires no expertise. As Moles says, [c]learly these researchersdo not distinguish between the writing (which is the legislation)andthemeaning ofthatwriting.179

    Bench-Capon et al. have worked with several statutes. MostfamousistheirworkwiththeBritishNationalityAct1981(UK).180Consistent with their approach to knowledge acquisition, theirPROLOG representation of the British Nationality Act was implementedintwomonthsbyastudent,withoutanyexpertlegalassistance.181

    They also represented the Supplementary Benefits Act 1976(UK)anditsregulationsusingasimilarmethod,andmadeasimilarclaimabouttheimportanceoflegalexpertise:

    Forourproject,theaccuracyoftherepresentationwasnotacritical consideration at this [early] stage. Our formalisation couldtherefore be undertaken with no expert legal assistance . . . Ingeneral, accuracy of the formalisation is, of course, critical, particularly if one were constructing a representation to be used inpractice.182

    Predictablyand correctlythis approach has been stronglycriticized. Molespointsoutthat:

    Thisistoassume. . . thattheproblemwithregardtoaccuracyis merely a matter of changing the detail of content. It fails toappreciatethatanexpertmayhaveagreatmanyusefulthingstosayabouthowonegoesabouttheprocessof interpretation. Theexpertadvicewillthereforehaveimplicationsforthemethodbeingemployedandthewayinwhichtheknowledgeisstructured.183

    KowalskiandSergotmakethisassumptionexplicit:Access to an expert adviser might well have changed the exactformoftherules inourprogram, but itwouldnot havechangedthemethodweusedtoformulateandcomputewiththerules.184

    Buthowcantheypossiblybesure?

  • 7/29/2019 Pragmatik 1

    51/405

    1.4 Legalexpertsystems 31

    Ontheneedforlegalexpertiseinthedevelopmentoflegalexpertsystems,SusskindquotesHayes-Roth,WatermanandLenat:

    Itisveryeasytobedeludedintothinkingoneknowsagreatdealaboutthedomain. Remember: theexpertbecameoneonlyafteryearsoftrainingandexperience.185TheworkofBench-Caponetal.hasbeenmostvehementlycri

    ticizedbyLeith. Hequotesthem:The formalisation of the British Nationality Act is an axiomatictheory similar, for example, to an axiomatisation of EuclideanGeometry. In principle, any logical consequence of the axiomatisationcanbegeneratedandtestedmechanically. . . 186

    thencomments:Doesthissound likeawildclaim? Isuspectso. No lawyer,withwhom I have so far discussed Kowalskis legal work, has beenimpressed in the least: the common reply is, Surely he doesntbelieve that? And this is the rub. In AI research, funders areprepared to provide research funding for the most inane of ideaswithout appealing to those with some expertise in the area ofintelligencebeingresearched.187

    1.4.3 GardnerTheaimofGardnersresearch188 isnottodevelopaprogramthatsolveslegalproblems:

    Instead, the objective is to enable the program to recognize theissues a problem raises and to distinguish between those it hasenoughinformationtoresolveandthoseonwhichcompetenthumanjudgmentsmightdiffer. Towardthisendaheuristicdistinctionbetweenhardandeasyquestionsisproposed.189Her chosen domain is an aspect of the law of contract that is

    dominated by case law: offer and acceptance. She claims that,although an area based on statute might seem easier for an AIprogram to handle, the reverse is true. Case law must be takeninto account in statutory areas too, and statutory interpretationraises itsownproblems. Hence, sheconcludes, [b]eginning fromstatutes therefore seems likelier to add a layer of complicationthantoremoveone.190 Susskindagrees: Gardner. . . isrightinrecognisingthatitisunrealistictofocusonstatuteattheexpenseofprecedent. Thereisalessonhereforallworkersinthefield.191

  • 7/29/2019 Pragmatik 1

    52/405

    32 Apragmatic legalexpertsystem 1.4

    Gardners system has four different levels, the last of which isnot implemented. The network level is an augmented transitionnetworkwhichrepresentslegalstatesandevents. Therulelevel isasetofruleswhichoperateonobjectsatthefirstlevel. Theserulesaredefinitionsofthemajorconcepts192 inthelawofofferandacceptance. Thethirdlevelisasetofexampleswhichexplainthosepredicatesintherulelevelwhichareundefined,butwhoseresolutionisclear;thisincludes(non-legal)commonsenseknowledge.Gardner, citing the legal positivists, assumes that open texturedoesnotrenderallcaseshard:

    Thefirstthreelevelsoftheprogramaredevotedtoidentifyingthehardquestions. Fromanoppositeviewpoint,thefirstthreelevelstrytoidentifytheeasyquestionsandresolvethem.193

    Thefourth(unimplemented)levelwoulddealwiththehardquestions.

    Susskindcommentsthat:Ultimately, the human expertise that Gardner tries to encapsulate in heuristic form is the ability to assess, in advance (in asense), whether a case does indeed raise easy or hard questions.Thiswillindicateifaconclusionmaybeinferredwithoutfurtheradoorwhetherhumanjudgementisrequired. Shemakesabraveattempt, but . . . I was left wondering about the generalisabilityof her ideas and their applicability in other, far more extensive,branchesoflaw.194

    Ashley,too,iscriticalofGardnersapproach:. . . an attorney distinguishes hard from easy questions in termsof comparing the strengths of the best argument he or she canmakewiththebestargumentsanopponentcanmake. Gardnersprogramprovidesnomeasureforevaluatingthestrengthsofcompetingarguments. . .

    Even if Gardners program evaluated case-based arguments, itcould not do so very realistically given the way cases are represented. Distinguishing, for example, is not possible because thereisnothingtodistinguish.195UnlikeMcCartyandBench-Caponetal.,Gardnerdoesconfront

    theargumentsofthe(American)Realists:If legal realism is right, it appears to make the AI paradigm ofrule-basedexpertsystemsinappropriate,atleastwithanysimplemapping from legal rules to knowledge-base rules. There are

  • 7/29/2019 Pragmatik 1

    53/405

    1.4 Legalexpertsystems 33

    severaldirectionsonemightgoinstead. Onedirectionwouldemphasizethe ideathat it is individualdecisions, not generalrules,that

    have

    authoritative

    status

    as

    law.

    With

    this

    emphasis,

    one

    might look for a method of reasoning from the decisions in pastcasestoa conclusion in a present case . . . 196 But this cannot bethewholestory. Ifitwere,whywouldpresent-daylawprofessors,thoroughly aware of the insights of realism, continue to expecttheirstudentstoknowrules?197

    This isastrangeargument; itsaysmoreabouttheway inwhichthe law is taught than it does about whether a rule-based approach to the law is appropriate. (And it must be rememberedthat the American Realists were not true rule sceptics, as discussedabove.198)

    Gardner identifies two other directions to take if legal realismisright: emphasizingthebehaviouristicsideoflegalrealism(e.g.thejurimetricsresearchdiscussedabove199),andathirddirectionwhichGardnerherselfadoptstoretainanimportantplaceforlegalrulesbuttoreinterprettheirsignificance.200 Sheclaimsthat:

    Rulelikesentencescanbeunderstoodasusefulcognitiveconstructs,neededtofindorderin(orimposeorderupon)anunwieldymassofindividualdecisions. Oncearticulated,theycanprovideguidanceastohowfuturedecisionscanbekeptinsomeroughconformancewiththisorder;or,ifthearticulatedruleseemstobeabadrule,it can suggest a way of saying how the course of decisions oughttobechanged.201

    Thisdoesnotreallyrepresentareinterpretationofthesignificanceof rules. Only the most ardent rule-follower would suggest thatruleswereanythingelse.

    Gardner develops a rule-based model for offer and acceptancecases,butmakesthesignificantconcessionthat:

    . . . a rule-based model of case law must be understood, like anyacademiclegalwriting,asasecondarysource. Theofficialsourcesare the decisions. There has never been agreement on what itwouldmeanforrulelikegeneralizationsfromdecisionstobebothaccurateandappropriate. Thus,thebasisforevenuncontroversialrulesremainsundefinedinlegaltheory.202

    This isaninteresting,andquitepragmaticconcession,comingasitdoesfromsomeonewhoseworkhasbeencharacterizedasfairlyclearlypuristinnature.203

  • 7/29/2019 Pragmatik 1

    54/405

    34 Apragmatic legalexpertsystem 1.4

    1.4.4 SusskindSusskinds

    work

    on

    the

    importance

    of

    jurisprudence

    to

    the

    developmentoflegalexpertsystemsisdiscussedindetailabove.204 He

    chose the Scottish law of divorce as his first experimental legaldomain.205 He has also been involved in the development of theLatentDamageSystem: alegalexpertsystemconcernedwiththetimeperiodswithinwhichclaimantsmaystartnegligenceproceedings where they have suffered latent damage or loss.206 CapperandSusskindclaimed in1988thatthiswasthefirst legal expertsystembuiltintheUK bylawyersforlawyers.207

    Susskindbelievesthatstatutesandsomecasestheclearcasescan and should be represented using rules. With clear cases,heclaims,itispossibletodrawlegalconclusionsonthebasisofliteral interpretationsoftheformallegalsources.208 However:

    . . . itshouldnotbetakenforgrantedthattheentirecommonlawsystem can be reduced to a collection of rules . . . Simpson hasforcefullycontendedthatsuchareductionistmodelmisrepresentsthecommonlawandisinconsistentwithitsdevelopment,content,andscope. . . 209

    Other methods, he says, would be required in order to representcaseswhicharenotclear: methodswhichreasonwithuncertaintyanddrawprobabilisticallyphrasedconclusions.210 Yet,hecontends:

    . . . althoughthecommonlawmaynotbesufficientlyrepresentedintermsofrules,itcannotbedoubtedthatitisinvariablypossible,desirable,andnecessaryto interpret individualcases inthe formofindividuatedrules.211

    I doubt that it is possible, dispute that it is desirable, anddemonstratewith SHYSTERthat it isnotnecessarytorepresentcasesusingrules.1.4.5 KnowledgeacquisitionandrepresentationFeigenbaumwrotein1981that[t]herearemanyimportantproblemsofknowledgerepresentation,utilisation,andacquisitionthatmust be solved, but the acquisition problem is the most criticalbottleneck problem.212 As discussed above,213 Bench-Caponetal.claimthatthereisnosuchbottleneckforknowledgeacquisition from statute law because legislation is already formulated

  • 7/29/2019 Pragmatik 1

    55/405

    1.4 Legalexpertsystems 35

    andwrittendown.214 However,thisdemonstratesaseriousmisunderstanding of the law. Knowledge acquisition is as much aprobleminthelegaldomain(statuteandcaselaw)asitisinanyother. Iproposeasolutiontothisproblem, at least in regard tocaselaw, inchapter2.

    Inordertoavoidtheknowledgeacquisitionbottleneck,someresearchershaveexaminedthepossibilityofthemachine-processingofstatutes. Iftheonlyinputisthewordsofthestatuteitselfthenthisproblemisoneofunderstandingnaturallanguagearesearchfield inwhichresultshavenotfulfilledoriginalhopes.

    Amorefeasibleapproachistoconvertthewordsofthestatuteintoamachine-readableform. Allenhasdevelopedonesuchformandclaimsthatifstatutesweredraftedusingthisform,notonlywouldtheautomaticlogicalanalysisoftheircontentsbepossible,but humans would be able to read and work with them moreeasily.215 Unless and until legislation is expressed in machine-readable forman extremely unlikely event, the desirability ofwhichisfarfromclearbuildersoflegalexpertsystemswhichdealwithstatute lawmustusehumanexpertisetoextractknowledgefromstatutes.

    Theprocessofacquiringknowledgefromstatutesisusuallyseenas a process of writing rules. However, Shannon and Golshaniwarnthatdoingthisinanadhocfashionisunsatisfactorybecausetherulescannotbecheckedforcorrectness,andsuchanapproachmay lead to dissimilar rule formulations which do not work welltogether.216 Instead they recommend following precise and consistentmethods(likethoseofAllen)toformulaterules. This,theysay,wouldreducethescopefordissimilarruleformulations.217

    Shannon and Golshanis belief that there is a correct interpretationofastatute issharedbymanyresearchers inthisfield.Inbelievingthis,theygoevenfurtherthanthelegalpositivistsonwhosetheoriesmuchoftheirresearch isbased.2181.4.6 FactrepresentationThe representation of legal facts in a rule-based systemindeedany expert systemis difficult. As with the representation ofstatutes,researchershavedevelopednormalizedformsforfactrepresentation. An example is deBessonet and Crosss atomicallynormalized form (ANF).219 Shannon and Gols