Representing Teleological Structure in Case-Based Legal Reasoning: The Missing Link

Donald H. BermanN@hms@n University School of Law

400 Huntington Ave., Boston MA 02115 USA

Carole D. HafnerNortk@ern University College of Computer Scienee

360 Huntington Ave., Boston MA 02115 USA

Wearguethat robuateam+asd models of legal knowledgethat represent tlw way in which pmeticing profbasionalsuse legal decisions must contain a deeper domain modelthat represents the proposes behind the rules articulated inthe cases. We propose a model for representing theteleological components of legal decisions, and we suggesta method for utilking this representation in a HYPO-likeframework for case-baaedlegal argument.

1. Introductii.

Bytheuseofhme+med Wucturm, transition nets,semantic networks, issuease discriminW on trees, andcxmnectionist models, AI researchers have made significantadvances in mod4ing various ways that lawyers reasonfrom previously decided cases [Ashley 1990; Ashley &Akwen 1991, 1992; Branting 1989, 1991; Gardner 1987;Goldman 1987; Hafher 1981, 1987; Rissland 1987;Rissland & Skalak 1989,1991; Voaaos et al 1991] Severalmodels generate outputs that to some degree simulatehuman adversmial diaeourae. [Ashley & Aleven 1991;Ashley 1990; Branting 1991]. Another model representsthe adversarial techniques for interpreting open texturedstatutory predicates with an elegance and thoroughness notfound in any comparable legal text [Skalak & Rissland1992].

To date, these case-basedcomputational models ofjudiciaiopiniona represent the knowledge as a cmcatenation ofdisembodied symbols divorced from the real world ofclients, lawye~ money, changing social values, history,politics, etc. In these models an objective symbol or set ofsymbols found inacurrent caseismatc~ornotmatche4 against symbols fbund in a prior case. But thesesophisticated indexing pattern-matching algorithms lackthe robustness needed to represent intelligently the thoughtPermission to copy without fee all or pafi of this material is granted providedthat the copies are not made or distributed for direct commercial advantage, theACM copyright notice and the title of the publication and its date appear, andnotice is given that copying is by permission of the Association for ComputingMachinery, To copy otherwise, or to republish, requixes a fee and/or specificpermission.

@ 1993 ACM 0-89791-606-9/93/0006/0050 $1.50

P~ ofpractieing lawyers becauaethey do notincorporate the purposm underlying the decisions. Theneed fordeepstructme inndebased systems hasalreadybeen noted [Smith & Deedman 1987]. In this paper webegin to examine the requirements for adding ateleological component to case-based reasoners. UnlikeSmith & Deedman who formulated their deep stmcturemodel without reliance on the doctrines or conceptsiuticuhted in judicial opinions [Smith& Deedman 1987,p. 89] we suggest a method of representing teleologicalarguments in a case-based system for legal reasoning byincorporating the teleological arguments that we believewere actually made or would have been made by skilledadvocates. Thus, we treat teleogoical arguments in amanner similar to argument moves described elsewhere[Skalak & Rissland 1992; Klare 1992; Kennedy 1991;LLewellyn 1951].

2 Teleological Concepts in Legal Arguments

From their first day at American law schools, studentslearn that the relevance of facts can not be divorced fromthe proposes behind legal rules. Consider the followingthree caseswhich arestndiedby many(ifnotmost)American law students [e.g. Casner & Leach 1984, pp.lO-21]. The plaintiff in Pierson v. Post (Pierson), 3 CM. R175,2 Am. Dec. 264 (Supreme Court of N.Y, 1805), wasfox hunting on open land. While the plaintiff, with horseand houn4 pursued the fox the defendanz well knowingthe fox was so hunted...didj in the sight of Px to preventhiscatchingthe same, killandcarry itoff. The Cmutruled that to recover the plaintifF had to have gainedpossemion of the fox. In ruling that the plaintiff had notgained posseaaionbecause he neither captured nor mortallywounded the fox the Court explaine4

We so hold] for the sake of certainty, and preservingpeace and order in sociely. If the first seeing stinting,or pursuing such animals... shotdd afford the basis ofactions...it would prove a fertile source of quarrels andlitigation.

5()

The Chutarticulated a vision that iftheydidnot establishstrict @Mineafwobtain@ title towildanimals thentheComtmightbe swamped withdiagnmded hlmtersarguingOverwhofirst sawtheanimal.

lhediment a,inaninten seeilbrttojustitjf acontrmyresu.lc ar~

Whenweretlect ...thattheintmsto fourhusbadmqthe mostusefulofme ninanyczummmity,willbeadvmwedby thedestmdionofabeast sopeMMousand incO@ibl&we eamtoterr, insa* that a pursuit like thepreaent ...czmfkrs ...aright totheobjectofit...

Thediasenter hadthevisionofmany additiomdtisravaginglocdfiulns unkashuntcrs arelewarddint.heir-@.

Since fktual relevance is determined within a broadphilosophkal and jurisprudential conte@ judges may notadhere to a uniform definition of possession in all wildanimal casea. Given the dissenters justificatio~ he mighthave agreed with the n@oritys result had the hunter beenpumuing a quail rather than a fox. Similarly, the majoritymight have reached a contrary decision had violence-~~wadtofmhConfrontation behavior.

Students next read Keeb/e v. Hickerimzill (Keeb/e), 11East 574, 103 Eng. Rep. 1127 (@ltXS1SBenc4 1707).There the plaintiff owned a pond upon which he placedduck decoys. The defendant intending to injure theplainti.tl% liveli.ho@ used guns to scare away the ducks.Even though the plaint.itThad neither wounded nor-tititi~m-h,fdforbplaintiff by reaaonin&

~hereaviolent ormalicious actisdonetoamanswcupat@profnorway ofgettingalivelihocxLthere anactionliea inaUcaaes. Butifamandothhimdamage byusingthe same employmen~ as if Mr.Hickeringill hadsetupanother decoy onhisownground neartheplaintitl%, andthathad spoiled thecustom of the plainti& no action would lie...This islikethecase ofll H.4,47. Oneschoolmaster setsupa new school tothedamage ofanantient school,-ti~alew-titiwwwtimmtiti-.~donwkldtimtimtie)Butsuppose Mr. Hicke@@shouldlie inthewaytithhisgmw and fkighttheboy fkomgoingtoschool ...surethat schoolmaster might have action.

*****

And when ...decoy[s] have been used...in order to be

taken for profit of the owner of the pond...and wherebythemarkets of thenationmaybe furnMM@thereisgreat reason togive

~ -to..*

Oncethepurpoae oftheruleisunde@3@ analogous easessettingftithe n@tsofscbl mastenbmmmlevantthan easesdealing with foxes.

Indealingwi ththeaecas eslawstuden tsalsodeveloptheability to isolate and match the relevant thctors. ~ViOUSsimihuities exist. Both cases involved unfettered wildanimals and both involved deft!ndants motivated by malice.Students dis@@sh Pierson from Keebfe on the groundsthattheplaintiEioPierson washuntingthe fox OnOpeOWwWk~~hKeebleti-tiWsontisown bu@andtheplaintiff in Pierson washuntingfbrsport while the plaint.iHin Keeble was purdng hislivelihood.

The importanceof this type of symbolic mapping becomesevi&nt when the students move to an examinahon ofYoung v. Hitchens (Young), 1 Dav. & Mer. 592, 6Q.B.606 (1844). In Young the plaintiff, a commercialfishermaQ spread a net of 140 Mhoms in length across aportion of open ocean. After the plaintiff had closed thenettoaspace ofaf-fhthoms thedefkndant wentthroughtheopeningand spread itsnetandcaught the fish.Studen@ f~ing on the plaintif?s need to make a living,map to Keeble which would @end a plaintiils victoIY.Students alSOmap to the open land in Pierson whichsuggests that the defendant should prevail. In Young theCouit found for the ddendant.

Figure 1 shows the relevant fhctors in the wild animaldomaim and the relationship among the three casesbasedon those fiwtors, following the approach of ptissland andAshley, 1987, Ashley 1990]. ThiS data stmctum can beused to generate the following 3-Ply argument [Ashley1990, p. 70-71]

Argument I.

Where: Plaintiff aeekingcapture ofawildanimal anddidnotmortallywound oreapturethe animtdthedefendant should win a claim for trespass. Cite:Pierson v. Post.

Rabvaat Futma:H Not CauSMw &4ertaU Wouwbd (NOTCAUGHT, Favm D)

, pinrlvsovnL8ndpP%2ff

& Favcm D; OWNLAND, FavomP)Pmuing Liwlibxl (LiYELlliOOD, Favm P)

Dofardant in Cemptitien vith PlaMfff (COMPETE, Favorr D)

OPENLIVELIHOODCOMPETE

tlPimrm (b)NOTCAUGHTOPENFigaN 1. Ftirs amt Cam Anal@a in Wiid hid Domain

COunterexample:

Keeble v. Hickeringill held for the plaintiff eventhough theplaintiffwas seeking to capture wildsnimalsand didnotmortally wound oreapturetheanimals.

=> ~ for ~e~t

Keeble v. Hickeringill is distin@sMle because: InKeeble the plaintiff was conducting his business on hisown land.

Keeble v. Hickeringill is also distinguishable because:In Keeb/e the defendant was acting maliciously whilehere the defendant was engaged in business-ti~.

Thou@ XCl@t@ transmibing IIMlly kiW school CkMsdiseusshs tbmxielisineomplete beeauaeatsomepointthestudent mustanswerthe question -whicheaseshouldgovern and whyP. Even though attameys wield theiradversarial f~s with considerable dexterity judg~, when_legalargumen@ donotactas refereesatanintellectual fencing match. Rather, judges make rules thatsignificantly affiwt human lives and their decisionsIKZCSW@yembody their views (or prejudices) as to whichrules improve the quality of life in society. Therefore,lawyers, in addition to arguments based on factualdistinctions, suggest to judges various policy argumentsthatsWaffectth edecision.He rearesOmeofteleological or policy arguments that the court may

have considered in Young. 1

ARGUMENT II:

=> p~t for ~-t

You should apply tk rule in Pierson becausetheuncemln . ty about what eonstituta propertyrights to fishmimndng intheopenseawillcause endless controversy.

~~ for ~e~t

Unlike the defendant in Keeble this defendantdefendant was not acting maliciously and was merelyengaged in vigorous eanpetition like the schoolmaster whomerelysets upaeompeting sehooIortheowner of adjoining land who also deploys deeoys.Society benefits from such competition.

S~*~ for ~e~t

Guidelines for determining whether competitionamong iishermen is fiir or unt%ir should be left to thelegislature otherwise there could be endless lawsuitsattempting to establish what fisherman mayor maynot do.

1 We have inferred from the reported opinions the policyarguments that were emsidered in Young because easeswere argued orally, and written briefs are not available.Also, since opinions were delivered orally and tmnseribedby reporters the opinions tended to be rather short. Finally,common law judges used a form of verbal shorthand thatembodied their policy determinations. As evidenee thatthese kinds of arguments are routinely considered in lawschool classes sec @3ukeminier & Krier 1988, pp. 8-20].

52

Lord Demnansopiniea in Young indhteathatkwaamoved bytkplaidffs argument butdeniedrelief becaulwthepMntiffhadmadc apmccduMerror. Judge RMtcaonscOncuningopinklQ in Yo8mgmggeata thathewaainfluenced bythethrustoftk defenM#ssurmbu@. Itwould appear that Judge Wightmans concuming opinion-b-c~.

3. b heumeat uf Existing sy8tem8

C!urrent CBRmodelasutl%r iiomtheaame limitationsasSHRDLU ~inograd 1972, pp. 8-11], a program that asksqucstionl%makes ata@n@aandissues cOmmmdstoarobot thatarranga blockaefdifkent *_andculors. Onecantell SHRDLUthat Idontownanythingwhich supporta apymmid hmwhich SH.RDLUcimreport that I dont own a box.

...although SHRDLUS answer to the question is quitecorre@thesystem cannot besaidtounderatand themeaning of own in any but a sophistic sense.SHRDLUS teat of whether something is owned issimply whether it is tagged owned. There is nointentional test of ownership, hence SHRDLU knowswhat it owna, but doesnt udmtand whatitistoownSoMhing. [Simon1977, p. 1064]

An analysis oftk outputofKari Bmntings sophisticatecasebd reasoner, GRI?BE ~ranting 1991], generatedin response to the f~owing hypoth4d demonstmtea thelimitations of unintentional symbolic maniptdatkm.

Joan and Donald were employed by the school districtas teackrs at a middle school and carpooled together.Eachworkday, thedriverofthe carpoolwaareqonaibIeforpicking upsomeaandwicks ontkwaytoworkfbrboth teachers toeatbecause therewaanocafkteria attheschool. Donald picked up Jaanatherhouaeand drove toward theschool. Dcmakithendeviatdikomthe direct route totheschool onhiawayto the sandwich ShOp. Befbre reaching the sandwichshop, Donald had an automobile accident in whichJoan was injured. Does Joan have a claim for workerscompensation against the school district? @ranting1991, p. 150]

GREBE retrieved three relevant cases. Tk ~ Janak V.Texas Emploprsl.. Co, 381 S.W. 2d 1976 (1964)(Janak), involved an accidentthatoccumd during adwiatkmt%omtkdirect mutetoanoil driuingsite. Thedriver andtheinjured pwwnger Worked onthedriuingcrew. TkPurpeaeofthe devbtimwast ogeticetocoolthe crews &inking water. In holding that the travelingwas in the fmtkmna ofemploymen4 the Texas Supreme

Court reaso@inp@ thattheice wasreasonablyew!dialfbrtk driuingactivitie sbecauaeoftkextremely hot working conditions. Jmak also decided thatthepasaenger intkcarpool wouldbedeemed incoumeof~~

ifthe driver were in the course of his*-

Sean@ Vmghn v. Highlands Underwriters (Vaughn),445 S.W. 2d 234 (1%9) held that Wk the

employee Wsatold by dispatckr to go get somethingtoeatwhile hiatruckwaabeing unload@ andwhowould not have made trip in absence of such order,was in course and scope of his employment whiletraveling ...fiom place ofhia employment to themstamant... [Vhughn, p.234].

Thid American General Ins. v. Coleman {*Coleman),303 S.W.370 (1957) which set forth the general rule that

an injury incurred in the use of public streets orhighways ingoing toandretmdngfiom theplaceofemployment is not a compensable iqjmy because notincurred in the course of the employment as requiredby ... Texas Statutes,.. [Coleman, p. 375]

From Jhnak GREBE concluded that Joan had thestronger argument. GREBE inferred that there is astrong argument that Joans trip to the sandwich shop wasin furtherance of her employment because she was amember of a carpool and her case was analogous to Janslcstrip to obtain ice water which was held to have been iniirtherance of his employment. To reach that ConclusionGREBE had to conclude that obtaining sitXldWiCkS wasreasonably essential to employment because the court inJivmkhadconcludedt hatobtaMngthei cewaterwaareamably essential for oil &Wing. But since thehypotkticd did not spec@ whether fd was necemay forteachin& GREBE then used Vlghn to support theproposition that food was reamnsbly essential toemployment. GREBE then noted a difference between thehypotktical andJanakbaaed onthefactthatthetempxature of the school was unrelated to the need forfood. ~mn~ pp. 151-52].

GIWBE illustmtes a sophisticated algorithm that triggereda plausible analysis of a complex legal problem. However,tklastpart of theanalysis would seemoddtosome akilledadvocates because few judges would base their decision onthe temperature of Joans workplace. More to the point anunderstand@ of the underlying purposm of workerscompensation law would suggest to many practitioners thatJoan has the weaker case.

The confluence of four 19th century judge-made rulesmade it very diflicult for injured workers to collect

dama@s from their employers. F- the employer had tohave acted negligmtly. Sect@ the injured worker had befree from contrihtmy m@igence. l%ir~ the employeecould not have vohmtmily assumed the risk ofinjmy.Finally, the Cll@OyOfwas M mqmnsible iftk! negligentSctsweremmmtted byanotberen@oyee @rson1992,54.30]. These rules were harsh becauaetheunitedstatesdidnot(and stiUdoesnot) haveabroad social satktynetsothatmostcithns badnomedical coverage andnoincomeprotection when i@red.

Aa pmt ofa 20th century legislative compromise employersweredenied these fwde&nsea inexchange fwworkersreceiving areducedamount ofdamagesin allcasesofinjuries thatoccumed during thecouraeofempbyment.But WOrkOl%~ onwasnot intended asaformofgeneral social inmmnce. The Coataofinjuries toanemployers workers were seen as a cost of doing businas@ therefore, should be born by the employer.Conversely, an employer should not have to bear the costsof injuries that were not work related -n 1992, $4.30].

Since legislators had no intention of requiring employers toinsure against all injuries

~ by people who wereemployed judges navigate between a visceral desire toassist injured workers who may have little or no medical ordisability inwance and their duty to avoid convertingworkers compensation into a system of general socialinsumnm. State judges are also aware that undulykmasingtbe cost ofworkers compensation immmncemay make their states businesses less competitive.

Whenthelangnage of thestatute sustained inthecourseof employment is read in light of * pwposes moatcasesare easily decided. If the worker slips atthefkcto~and breaks his leg he is covered but if he breaks his legwhile going down his fkont stepson the way to work he isnot. But there exista a class of cases like injuries swminedwhile commuting that involve some elements of personaland business activity. In this class of case the workershows a but for caumtion between his injmy and work -but for my job I would not have been commuting to worka@ therefore, would not have been at the place where theaccident occurred. But it is almost uniformly cxmcededthatworkers compensation was not intended to cover workersinjured during normal commuting @rsou $15.11]because compensating so many workers would constitute aform of social inmmnce that und4dy burdens employers.

The fbllowing lawyerlike argument that mirrors thesepolicy concerns would convince most judges that thepurpose of compensating injured employees rather thanproviding some form of general social insumnce wouldNOT be furthered by permitting workers to recover forinjuries sustained while buying their lunches on the way to

WO*

To permit Joan to recover would logically requirecm@nsadonfbrtheworker whodevia&afkunherroute togotothe wpnwketthe night befbretobuycold cuts and sandwich rolls. [See Lamom3515.12(b)].

Texas case law is consistent with this policy analysis. InJanakthecourt statecL

If the deviatbn... had been forthepurpose of pickingupteolsessential tothedriUingoperatiou thetravelwould clearly be impliedly dirated by the employer.On the other hand, if the deviation...lmd been forthepurpoae dpermitting oneormoreofthecrewmemberatobuyap artwlar kind of hamburger forhind, * travel would just aa clearly not beimpliedly directed by the employer (emphasisSUppkd). [J-p. 182].

Vaughn, decided five years later, does no~ as GREBEsuggests, support Joans case because the court in Vaughnwent to great lengths to make clear that the employee couldrecover only because he made a trip that he would nototherwise have made and the trip resulted from a diredrequest from the employer. Vaughn follows the generalprinciples set forth in Coleman that commuting injuries arenot compensable unless the transportation is fiun.ished bythe employer or the transportation is paid for by theemployer or the employee takes on a special mission forthe employer or performs a service for her employer withthe express or implied approval of the employer [Coleman,p. 375-76].

It has been suggested that Joans case becomes stronger ifthe facts demonstrated thattheschool waslocatedinaremote area similar to the drilling site in Janak.2 We donot believe that the remoteness of the job site wouldneceaw@ control the result. To prevail, Joan mustdistinguish her or Donalds activities from the daily actionsOfthemanytbowmds of Texas employees who pickuptheir lunches on their way to work. For exmnple, her casebecmnes winnable if the fact that the driver of the carpoolwas responsible for picking up some sandwiches isinterpreted to mean that her employer either expressly orimplicitly requested her or Donald to pick up sandwiches.3

2 This suggestion was made by an anonymous reviewer ofthis paper.3Wedonotargne thata Texas Court would nothavendedfor Joan. Since the Texas court decided Janak in 1964 andVaughn in 1%9 several other state courts extendedcoverage to claimants in Joans position. seeHornyak v.GreatA&P Tea Co., 305A. 2d 65 (N.J. 1973). We only

54

To hold otherwise would permit recovery for an injuryS@ainedwhile travellingto thesupermarket thenightbefore. Me@ the Texaa Coaut in Bnployers Casul& Co.v. Mii Jme Hutchinson 4 recently Eaf6rmdthisneedtolinktheinjured wtxkersactivity toactionsoftheemployers

Aniqjuryreceived whikusingthe public streetaiscmpenAkwhen theemployec hasunkrWenaspecial mission at the direction of the employer or isperhrmingaserviceinMhemnce of the employersbuaim%swiththeexpresa orimpliedappmval of theemployer...

4. Modelling Tck#o@cd Coacepts

Todevelopa deepnwdeloflegalpurp meaiaoneoftlwukimate goals of the Al and Law fieW however, we arenotattemptingto dothatinthis paper. Inste+@weattemptto make our case4wsed reasoning system a little smarter byattaching Sollw infibrmation about the purpmes involved inits caseknowledge base. To do this, we must consi&r whatthese pmposeaare, howtheyare related toeachotherinthedomainofb andhowtheycanbe usedwithinthe fi%unewcukofcmebed legal argument.

Discovering therelevant legal purposmina domain isthegoal (along with finding the relevant authorities andthctors) ofa skillfd legal researcher. Ey reading judicialopinions and consulting appropriate commentmy, it isgenerally possibk to @emtandwhat pmpoaesthecourtsaretryingto advance. Ourapproach toselectingpurposesinalegal domain will bethesameone usedtoselectfhctors or dimensions _ and Ashley 1987] incurrent legal CBR systems: that b a knowledge

maintain thatanyfhir reading of thethree Texas casesused by GREEE and the policies underlying thosedecisions donotsupport sucharuling. Indeedjthelanguage of those casessupports a contrary ruling.

Ourassessmentof GREEEsuggeats theneedtostudymethods of validating tlw pelf of AIsystemsinthelegal domain. One might present *problem forconsideration by ten practitione~ haIf of whom representWorkensand the other half who represent employers.Though a theoredcal mechanism for validat@ a system inthelegaldoma@ suchaproposal would meetwithrc%kamx Since lllitlly skiki lawyers would not respondwihoutarathercomple tefilethatinclud edaclientinterview, statcmmts ofco-worke~ statements of theemployer, the medical recorti maps of her route, etc. [SeePeterson & Waterman 1985, pp. 642+53],

4814 S.W.2d 539(Tx 1991) at 543.55

-* approac4 where the human exput determineswhatconcepts toincludein a domain model. Forthewildanimal Xthepwpoaeswewill *=thOsemntioned in Argmnent II above:

atodWne _ninaway thatpromotea---~ -

b. to protect a persons livelihood item Me&exec. to protect* enterprise and competitiond. for the judiciary to respect the powers of thelegishree. protect righta ofpropertyownersfnoproperty rights inresourcesfd onpubliclanda

For the WOXkWS~ Oncasesthe purpmeaarettommpwate iqjured wo~ and to avoid over+deninge@oyerawithclaims notdir@lyrelated toe@oym@.

There are several kinds of relationships among legalpmpoaes, which are illustmted in Figure 2:

a the taxonomic relation (ISA) relates a general purpose,such as the achievement of certainty and predictability inthelaw, toamorespedic exemplm, auchastheachievement of certainty and predictability in the legaldefinition of pmseasion of wild animals. The ISA relationalsoconnectsthe pmposeof ~--tothemore apecificpurpose ofencoun@g fox hunting.

b.theinstmme ntal relation (fN!ST) relates a generalpropose to a purpose which contrii to it for example,the goal of protecting a persons livelihood againstinterference has an instmme ntal relationship to the generalP- Ofen_ging usefd enterprise.

c. Alimitation mlation(LIMIT) qreasesthe f%ctt.hat0ne-mayimposeafitontimtof-~: for example, the purpose ofproteding the publicfiomobacenity impoaesalimit onthepurpose of protect@fkedomofspeech. Inthecurrent example, the~Ofpreventing mdhir competition imposes a limit on theP- of protecting tlee enterprise and competition.

d. the opposing relation (OPP) is the most strMngrelationship among legal purpmew for each purposeti_byo~ti&dale@m,timham_purpose which isadwmcedbytheotherside~1991, p. 101]. The fimdamental adversarial stmctumoflegal argmnent is often analyzed in terms of suchcompeting social pwpoaea.

Figure 2 shows three pairs of opposing goals

i. certainty of legal rules competes with context sensitivity,which permits the encouragement of usetid activity.

Promote eertawginemfting ~pp Allow wntext msitivarulsslegat dafinitiors ad rulw

to aIcoura@ ativities Ussfd~ to wcie~

Probet eeononiccompetition ~

4

Prottxt ValuabladivitiwsuchmasJluJuit.of.livdihood

E-Proteet rights ofproperty omenLIMITNo pro rty righkf?on pub C tadst

LIUIT

Pravmt unhir eompatitkm1--

opp Rm&;yp~$;iry for IFigura 2. K~wlad/P! StrIEUIM for La@ Purposas

currentCwt?Factor 1Fmtor 2F@or 3

\Precadant1 (D) Pracdent 2 (P)

Legal Purpowls:

Fwtor 1 Fwtor 1Fwtor 2 Faxor 3

&Figure 3. Ineorporaiing Lagat Purpsas into Cam Amalysis -tam

ii. protection of live W competes with protection ofWcmomie Wmpe4.ition.

iii, prevention ofunfi+ir competition by judicial deeision

56

competes with reaped by the judickwy forthe regulatory powers of the legislature.

Note that the shwture of legal purposashown in Figure 2 (and legal argument ILwhich is derived from the Young opinion)leads toargument points thatarenotobviously related to the earlier wild animaleases- points invoking unbir competitionand judiciidllegislative competency.Teleological knowledge allows a legalargument system to go beyond factualsimdarities to include thsse broiaderjurisprudential eonwpts.There are several approaches that might beusedtoineorporate legal purposwintothecurrent framework of legal CBR SY- asexemplified by the HYPO system [Ashley1990]. One could create knowledgestmetures that package the reasoningunderlying expWons [of eases].[Schank & Leake] Such expWons mightconstitute a slot in a easeframe and theseslots could be indexed in numerous ways toretrieve both similar and analogous cases.So the explanation of Pierson v. Post mightbe notpossession to avoid unnwessarydisputes while Keeble v. Hickeringillmightbe possession to protect theplaintiil?s livelihood. Such a scheme wouldhave problems since the explanations wouldneed to be unpacked when only some ofthe dimensions of the current easematchthe prior ease, and then re-combined in asystem such as GREBE which is capable ofcombining partial arguments from severalcases.

We have chosen a slightly different scheme,illustrated in Figure 3, which builds on thecentral role of faetm or dimensions inlegal CBR systems. In current systenu$eaeh factor indieatea whether it t%vorstheplaintiff or defendant. One limitation ofcurrent CBR systems is the lack ofexplanatory power as to why a t%ctor favorsthe plaintiff or defendan~ and why thatfhctor is considered legally relevant. In ourscheme, each hctor indicates the legalpurposes which it advances,~ each legal

~ ti turn speifies whether it favors the plaintitTordefendant. Finally, eaeh legal purpose is linked to itsopposite, so that the system knows when purpose A isadve that purpose A is generaUy defeated.

Sevemlibrma ofargument c3nbegenmted uaingthisknowledge atrwtum. Inacaaeauch as Young, anargmnentcanbe genmtedthatreaembb @P@~morethan Argument I. In Argument ~thedefdantpointaoutthatthe iishhadnotbeenmpturedormmtallywound@ juetaein Pierswn. However, withthe@ucture

PmPOaedke (illuatmtedibrthe wildanimalsdomaininFigure 4), anargunwntcanbe created whoeefirstetepmight be:

-> point fa ~

Thiscaaeis like Piemonv.Post,aincepbindffdidnot_m_wotmdthefiItia~todefine pweaaimina waythatpmmotea certainty andav~ ~ -. ~~ f~ ~e~t

Thiscasecanbe dis@uMA fkom Keeble, since inKeebletheinterference waemaliciou& while inthiacasethed&ndant waeineompetition withplaintiiT.Thelawshodd protect tienterpriseandcompetition.

Undistinguishing acase(asthe pUntiffin Young wouldattempt to do), tk adversarial link from certainty to''encourage ueefulenterprise* can beusedtohelp thecaae+based reasoner create teleological argmnents. For example,theplaintiffin Young mightbolsterhis reqOnseby-the~ of fiahingas a moreusefid activitythan fox hunting:

Point for Defendant

l%edefendant should win becauaethistypeefcompetition will encaurage the moat efllcient methodsof trapping fish.

Or believing that a judge would be influenced by mugmentsWribWMetocridcal legal tkmiatethe attorney forthedefendant might argue

= Point for Defendant

Thedefendant should winbecauseprotect@ the-is P- *M* -i---~the expense of smaller entmpmeum.

Though arbitrary andincompleteour propoaedarsenalofteleological arguments more closely approximates realworld adveeacy in mnmon law jurisdictions. Grantex%there isnoineluctable limit tothelevele ofabatmcdonthatonecouldincorporate intoamebmed reaaonersowecanMtreplhtet heildlrange ofargmnent moves fdinthearsenal ofa truly gifted advocate. Nonethekaaj weanticipate thatcaeehadreaaonm thatimqomteteleological argmnentswill prove moreuseful tolessakiUedadvocateeorlegal dwatoradesirioue ofenhanchgtheadvocacy skills of their students [Ashley& Aleven 1991].

57

Relevant Factors:~ NotCaught or Wrtall Woumkl (NOTCAUGIiT)

~ Ad~: mfim Fz~ ta ROMCAC Cartabty. .

Phntlff mating Oa Ovdoplkn Land (OWNIOPEN)~ @--: protect pro~ty Ri@talNo property rights

s** WJlihod$L.ELm.D)~ Advancu+ Protect aluable Activities from htorferenw)

Defemlamt in Compatataon with PMntifff (COMPETE)~ M-: Protut F- ~ ad Co#qnitition

YoungNCITCAUGHTOPENLIVELIHOODCOMPETE

Legal Purposes:

p, p,

Promote mrtainty in crafting OPP Emouraga %tivitieslqal dafiitions iud rules usefultosocis~

(FWcxsDafdant) (Favors Plaintaff)

i A

I ( INST =

m ST !=LIMITtNo pro rty rightson pubic landsI I

Protact frm Sntm@sa Protsct valuable wtivitiesOpp Smh ?Spursuit of liwlihoodml conqditon(Favors

~) from interfemme.(Favors Plaintiff)

-k-Figure 4. Teleological Analysis of Wild Animal Cases

58

s.ConeluiQmWehaveargwdthat tkpurpmeaoflegal ndesareanimpwtmtwmpommt eflegaiar- andhavebeguntoexplorewaysefmpmmtmg auchpuqmses inaeasebd~-we-hhowm repmmWionsehemeibrlegalpurpmcS eddbeueedto makelegalarguments morereaUtie. Theqnx@aWn suggested forlegal purpowsraises manymomqudonet hanitanswem-q-that ought tobctheaubject ofmuehnuxeresearchbythe AIand LaweQmmunity.

6. Re4emnees

Ashley, K. 1990. M&lling Legal A-nt: Reasoningwith Cases md Hy@eticals. Cambridgq MAMIT Reee.

Aahley K. & Aleveu V. 1991. Toward an Intelligenttutoring System for Teaching Law Students to Argue withCases. In Proceedings of the Third InternationalConference ofArti$cial Intelligence and Law, 8LNYAsociat.ion fa computing Machinery.

Brant@ L. 1989. Representing and reusing explanationsof legal precedents, In ?%ceedings of the SecondInternational Conference ofArt@cial Intelligence and Law,103. vancouvemkX&tm n for Computing Machinery.

Brant@ L. 1991. Reasoning with Portions of precedent. InProceedings of the Third International Conference ofArtijlciallntelligence and Law, 145. NYAso@ion forcomputing Machinery.

Casner, J. & Lack B. 1984. Cases and Text an Property.Boatow MALittle Brown.

Dukeminier, J. & Krier, J. 1988. TeachersMmual,Property (2d. cd.). Bostoq MA:Little Brown.

Gardner, A. 1987. An Arti$cial Intelligence Approach toLegal Reasoning. Cambridge, MAMIT Press.

GOkhIMQ S., Dyer, M. & Plowers, M. 1987.Preced@+aeed Reasoning and Knowledge Acquisition inContract Law A ProceeeModel. Proceedings of the FirstIntematiomd Conference ofArtljicial.Intelligence andti,210. NY:Association fw Computing Mizchinery.

Ha&r C. 1981. An Information Retrieval System Bawd ona Computer M&l of Legal Knowledge. AnnArbor,MIXJMI Rmearch Press.

Hafner C. 1987. C!onmptual Organizuion of CaeeLawKnowledge Bases. In Proceedings of the SecondInternational Conference ofArtificial Intelligence and Law,

35. v ancxnwemkmiation for Computing Machinexy.

Kennedy, D. 1991. A Semiotics ofugal kgument.$vracuse Law Review 42(1):7S-116.

Klare, K. 1992. Torts Su@ement (on file with theauthors).

LarsoIL A. 1992. Zhe Law of Workmens Compensation,New York Matthew Bender.

PetereoQ M. & WatermraLD. 1985. An Expert systemsApproach to Evaluating Roduct Liability Cases. InComputingPower and Legal Reasoning, ed. Waker, C.,629-59. St. Paul, -West.

Rkda@ E. & Ashley, K. 1987. A Case-Baeed System forTrade Seem@Law. In Proceedings of the FirstInternational Conference ofArti~cial Intelligence andLuw,60. New YorkAsociation for Computing Machinery.

Riasla@ E. & SkalalL D. 1991. CABARET: StatutoryInterpretation in a Hybrid Architecture. In InternationalJournal of Mm -Machine Studies 34(6):839-887.

SchanlL R &Leake D. 1990. Creativity and LearninginaCase-BaaedExplainer. In M4CHINE LEARNING:Paradigms andMethodr, ed. P. Carbonell, p. 354-385.Cambridge, MA:MIT Ress.

SimoU H. 1977. Artificial Intelligence Systems thatUnderstand. In Proceedings IJC41-77, 1064. Menalo ParlLCA:International Joint Conferences on MiticialIntelligence, Inc.

Skalak D. & Risslt@ E. 1993. Arguments and Cases: AnInevitable Intertwine@. Artificial Intelligence and low: AnInternational Journal 1(1):3-44.

Smiti J. & Dee&wuL C. 1987. The Application of ExpertSystems Technology to Case-Based Law. In Proceedings ofthe First International Conference ofArtijicial Intelligenceand Law, 85. New York: Asso&tion for ComputingMachinev Ress.

Winogr@ T. 1972. Understanding Natural Language.Cognitive Pqwhology 3:8-11.

Vossos et. al, 1991. An Example of Integrating Case BasedReason@ with Object-Oriented Rule-Based Systems:IKBALS U. In Proceedings of the i%irdIntemationalConference ofArt@cia! Intelligence and &nv, 195-20Q.mhsOcktion fix Computing Machinely Press.

59