AI and Legal Reasoning
Proceedings of the Ninth International Joint Conference on Artificial Intelligence. General. Pages 1254-1259.
© 1985 International Joint Conferences on Artificial Intelligence
Abstract
Thil paper prelelltl a IWDID&ry of the ~ of a panel to illuel 011 Al and lepl ,.,.,..lna. Tbe panel CODiilted of: Bdwina L. Rillland, Chair (UDivenity of Maaacbuaetts), Kevin D. Atbley (UDivenity of Maaachuaetts), Michael 0. Dyer (UCLA), Alme v.dL. Oardner (Stanford), L. Tbome McCarty (Rutpn), and Doa.ald A. WatenDan .(RAND). Amooa the illuea ~ by the panel were: 1. What are the characterlldca of the lepl domaln that make it llltereldq or ameoable to AI approechea - wbat Ia lpOCial about lt; 2. Tbe opeo·tatured oature of lep1 cooc:ep11 and the lmpllcati0111 thll bu for llliDa Al·tec:bniquea, eapeclaUy knowledp repre~CDtatioo; 3. Tbe c:omplemeatulty of rule-bued and c:ue-bued reuooiDa - bow cuea are uted, eapecla1ly wbea the rulea "ruo out"; 4. Tbe pervuive role of aoaloay lD lep1 reuonlna; S. Tbe lpOCial role played by bypotheticala lD the lepl domaln and bow bypc11 belp wlth arpmeatatioo and stratqlc cue plannina; 6. Tbe interleavllla of )lstlficatioo, aplanatioo, and arpmeatatioo; 7. How c:ommoo law tylteml can be leell to be systems wbich leuo from cuea; 8. Tbe apprcpriateoe. and feuibility of lllteßlaent aldl for practicina litlptors and otber lepl experts; 9. lmpllcati0111 for otber domailll - llte medldnethat use c:ue-bued ,.,.,..lna; 10. Methodoloaical and otber illuel. For each illue COillidered, the c:ommeats of the panellats are summarized.
- Albiey, K. D. (1984). "Reuonina by Analogy: A Survey of Selec:tecl AJ. Research with lmplicatiODJ for Legal Expert Systems". In Proceeding1 Firn Anltlllll Coriference tHJ Law and Tecllnology, Univenity of Houstoo, Houstoo TX. To be ioduded io a collectioo to be publilbecl by West Pub. Co. io 1985.
- Aupst, S. B. and Dyer, M. G. (1985). "Underltandina Analopet in Editorials." UCLA AI Lab Memo.
- Birnbaum, L. (1982). "Argument Molecules: A Functional Representatioo of Araument Struc:ture". In Proceeding1 AMI-82.
- Pithburgb, PA.
- leBessooet, C. G. (1984). CCUPS, TM Civil Code Legal ln/ormDtioft Proce~ling Syltem. Law and AI Projec:t, Louiliana State Law Institute, Batoo Roup, LA.
- Corbin, A. L. (1961). The 1udicial Proc:al Revilitecl. Yale Law JDUTMl 71, 195-201.
- Dyer, M. G., cl Flowers, M. (1984). "''oward Automatins Lepl Bxpertile." In Proceeding1 of tM Firlt Law and Tecllnology Coriference. Univenity of Houstoo, Houltoo, Texu.
- Flowers, M. {1982). "'n 8einJ Cootradictory". In Proceeding1 of AMI-82.
- Pittlburp, PA.
- Flowers, M. and Dyer, M. G. (1984). "Really Arpi.na with your Computer in Natural Languap." In Proceeding1 of NCC~.
- Lu Vegu.
- Flowers, M., McGuire, R. and Birnbaum, L., (1982). "Advenary Arauments and the Loaic of Penonal Attacks." In Lebnert and Rioale (ecll.) Strategk1 for Natural Lturg1111ge Proce~ling. LBA Prell, NI.
- Gardner, Anne v.dL. (1984). All AnificUil lnlelligence ApprDGCII to Legal RetutHiing. Doctoral Oillertatioo, Deputment of Computer Science, Stanford Univenity.
- --(1983). "The Dellp of a Lepl AnalytiJ Program." In Proceedillg1 AMI-83.
- Wublnatoo, DC.
- Goldman, S. R., Dyer, M. G. • and Flowers, M., (1985). "Prec:edence-Balecl Leaal Reuootna in Contract Law ." UCLA AI Lab Memo.
- Hobfeld, W. N. (1919). "Fundamental Lepl Cooceptions u Appliecl in 1udicial Rea.,.,lna." Yale U. Prell, New Haven, CT, Bd., W. W. Cook.
- Kubn, T. S. (1970). Tlle StrlletUre of Scknlf/ic Revoluttoru. Sec:ood Bditioo. Univenity of Cbicaao Prell.
- Lakat01, I. (1976). Proofs and Refutatiotu. Cambridge University Press, Londoo.
- Levi, E. H . (1949). An lntroduction to ugal Reasoning. University of Chicago Press.
- Llwellyn, K. N. (1930). The Bramble Bush. Ocean Publications.
- ----(1960). The Common Law Tradition: Deciding Appeals. Little Brown, BOiton.
- McCarty, L. T. (1983a). "Intelligent Legal Information Systems: Problems and Prospects." Rutgers Computer and Technology Law Journal 9(2):265-294. (1983b) "Permissions and Obligations." In Proceedings /JCA/-83.
- Karlsruhe, W. Gennany. (1977). "Reßections on TAXMAN: An Experiment in Artüicial Intelligence and Legal Reasoning." Harvard Law Revkw 90:831-93.
- McCarty, L. T. and Sridharan, N. S. (1981a). ~e Representation of an EvolvinJ System of Legal Coocepts: ll. Prototypes and Deformations." In Proceedings /JCA/-8/. (1981b). A Computational Theory of ugal ArgWMnl. Technical Report LRP-TR-13, Laboratory for Computer Science Research, Rutgers University.
- McGuire, R., Birnbaum, L., and Flowers, M. (1981). "Opportunistic Processi.ng in Arguments". In Proceedings /JCA/-81.
- Vancouver, B.C. Polya, G. (1973). lnduction and Analogy in Mathemlllics, Volumes 1 &: 2. Princeton University Press, New Jersey.
- Peterson, M. A . (1983). New Tools for Reducing Civil Litigation E:cpenses. Memo R-3013-ICI. The Institute for Civil Iustice, The RAND Corporation, Santa Monica, CA.
- Reichman, R. (1981). "Modeling Informal Debates". In Proceedings /JCA/-81.
- Vancouver, B.C., August.
- Rissland, E. L . (1984a). "Argument Moves and Hypotheticals". In Proceedings First Annual Conference 0r1 Law and Technology. University of Houston, Houston TX. To be included in a c:ollection to be publisbed by West Pub. Co. in 1985. (1981). E:cample Generation. Technical Report 81-24, Department of Computer and Information Science, University of Massachusetts, Amherst, MA. (1982). "Examples Hypotheticals in Contract Law". Annual Cognitive Science Society Michigan, Ann Arbor. in the Legal Domain: In Proceedings F ourth Conference. Univ. of ----(1983). "Examples in Legal lteuonins: Legal Hypotheticals". In Proceedings /JCA/-83. Karlsruhe, W. Gennany. ----(1984b). "Hypothetically Speaking: &perience and Reuooing in the Law". To appear in Cognitive Science. An earlier version appeared in Proceedings First Annual Conference Oft Theoretical /ssues in Conceptual lnformotion Processing, Georgia Tech, March. (1984c). ~e Ubiquitous Dialectic". In Proceedings Si:cth Ewopellll Conference 0r1 Artificiol lnlelligence (ECA/-84), held in Pisa, Italy, September, publisbed by Elsevier (North-Holland). (1978). "UnderstandinJ UnderstandinJ Mathematics". Cognilive Science, Vol. 2, No.
- Rissland, E. L., Valcarce, E. M., and Asbley, K. D. (1984). "Explaining and Arguing with Examples". In Proceedings AAAI-84, Austin, Texas.
- Waterman, D. A ., and Petenoo, M. A. (1981). Models of Legal Decision Making. Memo R-2717-ICI. The Institute for Civil Iustice, The RAND Corporation, Santa Monica, CA.
References were extracted automatically from the PDF and may contain errors.
Full text extracted automatically from the PDF; figures, tables and equations may differ. The PDF is the version of record.
1. 1'lle C.Deap ud Special Claaracterlltlc ol ... Lepl Daalala
Tbe lepl domaln prelelltl IOIIle very llltereldq challeqea to the Al reaearcher. Wblle lt Ia a domaln wblch bu eatabllabed ltandardl for derivina new truths <··I·· "·· dectN or the doctrlDe of precedeot), lt il more of a "'cruffy" domain than a "Deat" ooe, delpite lts orderly, rule-like IUJ'face veoeer. lt Ia very much an experieace-bued aample-driveo field. Lepl ret1011ina il allo beavüy intertwined wlth oatural lanpaae procellina and commoo te111e l'e&ICIIliq and therefore inberits all the bard problems that these lmply.
Aland LEGAL REASONING
Report of a panel chaired by
Bdwina L. Rialand Deputmeat of Computer and lnformatioo Scieace UDivenity of Maaachulettl Amhenl, MA 01003
Several paneliJtl empbuize that lepl reuonlna and argumentatioo take spec:ial lkUII and that leunlna to thint like a lawyer requitel COilliderably more than roce memorlzatioo of a larp oumber of casea, a dauotiq tut in itlelf. For lnstaoce, Dyer says:
Modelina wbat a lawyer doea Ia more compla than ~odelina expertl lD techDicaVscleotific domailll. Flnt, all of these c:ompla cooc:eptualizatiODI are apm ecl lD oatural lanpaae, 10 modelioa the comprebensloo ability of the lawyer requites solvllla the oatural Iaoau·ae problem. For aample, aivina lepl advlce ofteo 1tart1 wlth bearioa a "'tory" wbere the clieot wu ooe of the actors. So lepl advlce ofteo preiUPJX*I a story understandlila capability. ID other apert systems, the oatural Iaoauaae problem can be larply fin.ed linc:e the tut (e.a .· diJeue diapolil, recoofiauriol bardware, aoalyzioa dipmetm) rarely lllvolvea the c:ommuDicatioo of c:ompla cooceptualizati0111. Secood, the plaoDioa lD thll domain il compla. Thil Ia a world wbere actor1 obtalo apots, wbere they counterplao, retaliate, attempt to set up sublumptioo ltatea, etc. Tbird, Aqlo-American jurilprudence, at leut, makea use of a larp body of conc:nte BPISODES. 8ecomioa a lawyer requitel leunlna bow to lllda, aeaeralize and apply these coocrete epiiOdel lD service of rea""''"l and arpmentatioo stratepea. Founb, law Ia limply an attempt to formallze COMMON SBNSB reuooioa and ooti0111 of jultke, morality and fair play.
Rialand pointl out the aoalosiea betweea lepl and mathematical reuooioa· ID her aoalysll, learoioa to be an apert mathematician or mathematicl studeot lllvolvea a Iot more than limply learoioa defiolti0111, theorems and proofl. Tbere Ia a wealth of otber knowledp and lkUII oeedecl: for inltance, there Ia a larp c:orpus of aamplea and beuriltica, and there Ia the lmportan.t IkW of belDa able to aenerate aamplea [Rilllaod, 1978]. Slmilar remarb cao be made about the lepl domain [Levi, 1949; Uewellyo, 1930]. ID particular, casea lD law play the aoalosous role to aamplea in other fields. However, in additioo to the usual quelti0111 about aamplea - such u their functioo and aeaeratioo - the law il a domain wbich Ia larply case or aample-bued and thus aamplea are truly central.
lbe law aJJo e:xhibitl many of the traitl dilcuaed by Kuhn with repnl to ICientific dilciplinea [Kubn, 1970). In particular, the dialectic between propoli.na an idea or rule and the teltina of it tbrougb c:oocrete cuea - 10 weil dilculled by Lakat01 (1976) and Polya [1973) - can aJJo be found in the law and other fielda [RilllaDd, 1984c).
law: Watetman notea teVeral special charac:teristica of the
lbe law combinea many different kinda of reuonina proceaa: rule-bued, cue-bued, analogical, hypothetical.
lt hu the unique property of beina peeudo-formalized, i.e., there exiltl a larp body of formal rulea that purport to define and replate activity in the domaln. However, thae rulea are often deliberately ambipous, c:ootradictory and incomplete. Many of the actual rulea uled in legal reuon.ina are rulea about how to aa:ea and reuon with the "formal" rulea. lbe problem this createa il the naive notion (for 10me) that becau.e a body of rulea and replations exiltl, all one hu to do il b'anllate them into executable code to create a legal reuonina proaram.
lbe law il in a CODJtant state of chaap, 10 expert legal reuonina l)'ltems have to be euy to modify and update. Many other application domalns have this characteristic:, but perhaps not to quite the same extent.
Oardner not only joinl others in pointins out the challenp for Al pre1011ted by the intertwlniina of natural lanauaae and common leDJe t.J:i.owledge in legal reuonlna but aJJo lhe remarkl on 10me differenc:ea c:ooc:ernlna the nature of expertile in the law compared with other domalns of expert systems work:
Law il an area for "expert systems" in the leDJe of involvfna profellional knowledge, but it il unusual in that we have to expect the expert~ to dilap'ee. (B.a., oppolina COUDiel; judp wrltlna majority and cU.entina opinions.) Queltions then arilina are: (1) What kinda of thinp can they dilapee on, and what kinda of dilap'eement are beyond tho bounda of profellional competence? (2) How can a rea1001na l)'ltem leave room for the appropriate kinda of dilapeement?
Law il aJJo an area deeply involvina natural lanauaae and common~eDJe unc:lentandiq. In particular: (1) Lepl materlall (such u statuteil and other statementl of legal rulea) are wrltten in natural lanpap, and the problem of saylna how they apply to a particular li.tuation il in larp put a problem of sayina what thil fanauaae mcanJ. 1bil upect of law c:alll for a different approKh to meanlna from that taken in mOit Al propams. (See dilcullion of open texture below.) (2) lbe lituations to which we mfaht want to apply lepl knowledge encompu1 practically the whole ranp of human activity - with a special focus on th010 human lituatiOOI where IOIDethins 8001 WfODI· lbus a tru1y apert l)'ltem wouid be able to repre~ent and reuon about such lituations uli.q a areat dea1 of common~eDJe knowledge u weil u technical knowledge. 1bil lut il true of IOIIlO other kinda of apert l)'lteml too.
2. Tlle Opea·Tatared Natare ol Lepl Coaeepta
Unlike certain other domalns ll.ke mathematics or even medicine, c:ooceptl in the law tend to be "open·textured", that il, one cannot provide black and white definitions. Interpretina a legal conc:epc in a new li.tuation dependa on pul Interpretations. To declde whether a 1ep1 predicate appliea to a fact li.tuation, one musc usually find c:ue1 in which the predicate wu applied and focus on li.milaritiea or dillimilaritiea.
Gardner, who acSctre.ed thil illue in considerable detail in her rec:ent doctora1 thelil ( Gareiner, 1984, apecially Sections 3.1-3.3], summarizes:
Legal lanauaae il open·textured in that (1) we cannot state n~ and sufficient c:ooditions for the application of legal predicatea and (2) neither can we take legal predicata u primitiv. in the .._ that a proaram could li.mply rec:opize whether or not the predicate appliea to 10me fact li.tuation. (lbat il, a legal rule (Px impliea Ox) and a fact Pa do not entail legal CODIOQUence 0..)
1bil characterization of open texture coven IOVeral phenomena. One il the variable standard (e.a.. "a reuonable time"). Another il vaauen· (althoup legal rulea do tend to avoid uli.q obviously vape words - leafllators wrlte rulea about people over 65, not about "old" people). lbe m01t important, however, il that legal concluli.ons that appear to follow deductively are in fact defeali.ble. lbat il, Iiven the legal rule (Px impliea Ox) and the fact Pa, the concluli.on 0. il only a default concluli.on.
Alhley pointl out that the law's handllna of open·textured predicata il not Deutrai, that il, the decilion about the conc:epc il tied in with who the declder wantl to win the cue:
In decldiq whether a lepl predicate appliea to a fact li.tuation, a court usua11y aJJo determines wbo sbould win tbe issue or tbe case. The criteria for deciding wbetber tbe predicate applies are rarely completely specifiable in rules. The court must usually find cases in wbicb tbe predicate was applied and focus on similarities or dissimilarities between tbe new fact situation and past cases. lt must decide if tbe romparisoll cuts in favor of following tbe prior case or distinguisbing it. 'Ibis analysis is significandy different from trying to apply some abstract definition of tbe legal predicate. Any attempt to model legal reasoning must take account of tbis aspect of caso-based reasoning.
3. 1be Complemeutarlty or Cue-Based aad Rule-Based Reasoaia1
One way cases are used is to give concrete content to tbe concepts and abstractions in legal rules. Even if one believes tbat tbe law can be captured in rules - wbicb many, particularly tbe legal realists, do not - one needs cases to flesb out tbe meaning and intent of tbe rules. As Gardner bas put it, cases are wbat you use "wben tbe rules run out". Most scbolars of tbe common law, would say tbat cases are wbat you use all tbe time - even if you bave rules and know bow to apply tbem. McCarty bas long campaigned tbe importance of tbis point for AI work on legal reasoning; for instance see [McCarty, 1977; McCarty and Sridbaran, 1981a].
Rissland restated tbe tbeme of tbe intertwining of caso-based and rule-based reasoning, and tbat it is akin to tbe proofs-and-refutations dialectic between conceptlconjecture and examples talked about by Lakatos. In fact, sbe suggests, as bas Gardner, tbat one could carry tbis computationally further for bandling tbe different types of questions discussed by Gardner [Gardner, 1984]: a rule-based approacb for tbe "easy" or black-and-wbite questions and a caso-based approacb for tbe "bard" or gray-area questions. A rule-based reasoner sbould be able to caU on tbe services of a caso-based reasoner wben tbe predicates get murky or tbe rules run out and a caso-based reasoner sbould be able to caU a rule-based reasoner for tbe weU-understood, weU-structured aspects of its problems.
4. Tbe Pervuive Role of Aaalogy
Asbley identifies several uses of analogy in legal argumentation and practice [Asbley, 1984] including:
- Precedents tbat are analogous in some sense to the case being decided are used to focus tbe attention of tbe advocate on tbose issues and autborities wbicb bave a bearing upon tbe argument;
- The precedent provides an example of wbat a reasonable argument in an analogous context Iooks like;
- An existing classification from a precedent may be applied to analogous facts in tbe case being decided;
- A new classification, analogous to a classification in a precedent, may be created and applied to tbe facts in tbe case being decided.
Gardner points out tbat if one takes analogy to refer to comparing facts of cases for tbe purpose of · arguing tbat a new case sbould be decided in tbe same way as an old, similar case, tben tbe crux of tbe issue is wbat makes some factual similarities and differences important and otbers irrelevant; tbis involves more tban a simple comparison of facts.
McCarty bas addressed tbis issue in bis TAXMAN ll researcb, especially concerning tbe famous tax case of Eimer v. Macomber. If one can imagine a sequence of intermediate bypotbeticals connecting tbe two cases to be analogized then the sequence of intermediate transformations provide a key to tbe analogy. For a detailed analysis, see [McCarty and Sridbaran, 198lb).
Dyer ecboes some of McCarty's and Rissland's concerns, particularly on tbe utility of tbe idea of deformations and standard or prototype cases, witb an added empbasis on tbe key role of memory in analogy:
Functioning successfully as a lawyer requires baving indexed key cases in such a way tbat tbe comprebension of novel cases l~ds immediately to access of key cases. In a sense, legal reasoning and comprebension involves finding "issands" of legal prototypeil from wbicb well-known reasoning patbs can be adapted to tbe novel situations and tben traversed. Mucb of legal reasoning, tberefore, is a memory indexing and searcb problem rather tban a logical or "tbeorem-proving" problem.
5. Tbe Role or HypothetJcals
Hypotbeticals - tbat is, make-believe cases - serve many roles in legal reasoning. Several of tbe panelists, in particular, McCarty, Rissland, and Asbley, agreed on tbeir importance. AU tbree make strong use of bypotbeticals in tbeir models of legal reasoning and generate tbem in a somewbat simi1ar framework (caUed "prototype plus deformation" by McCarty and "modification plus retrieval" by Rissland, wbo also taxonomizes examples into claaes like "model", standard reference, anomaly, etc.).
Rissland enumerates some of tbe uses tbat bypos play in legal reasoning and argumentation [Rissland, 1984a, b). For instance, some general functions of bypos are:
- bypos remake experience
- bypos create experience
- bypos organize and cluster cases
- bypos tease out bidden assumptions
For inltance, tbe lleCOild usage allows one to CODJider laues tbat have not yet been litipted, and may weU be, (e.s., tbe recent cue of tbe frozen embryo which was predeceued by its parents or a malpractice suit apinJt an expert system). lbus hypoa can conuibute to stratepc Plannins by providins what-if lituati0111 - espec:ially contruy ones - to consider. ThiJ can Iead to a debugina an argument or stratesY ahead of time.
Ashley elaborates on three roles played by hypotheticals in tbe context of argumentation:
1. In a lesai argument, a hypothetical can be used to isolate tbe weaknesaea and strenstb· of an attorney·s cue. A hypo can be constructed to emphuize one upect of a cue tbat pusbes stronslY for a decision contruy to tbe proponent's, perbaps by milkins it morc like a line of contrary c:ues or by exaaeratins some common sense objec:tion to followins tbe proponent's classification. One way for tbe proponent to respond il to diJtinsuisb tbe hypo and hil cue. · In so cJoins, tbe proponent il fon::ed to emphuize tbe otber strenstb· of hil cue tbat are not present in tbe hypo.
- A hypothetical can be used to illuminate tbe consequences of a decision. A court may ask, "If we dec:ide in favor of tbe proponent on tbil cue, what will we do when a case witb somewhat lesa sympatbetic facts il posed?" lt il an exercise in line drawins. If no defenlible lines can be drawn, tbe court may choote not to decide in favor of tbe proponent, or at least not on tbe grounds arsued for.
3. In both of tbe above, tbe hypotbetical il used u a sbort band for lines of cases, lines of reuonins. even whole arsuments.
Such uses of hypotheticals in lesai argument can be illustrated by excerpts from oral arguments before tbe United States Supreme Court in which tbe justices poae hypotheticals to make and elicit lesal points and to control tbe presentation of a proponent's cue (e.s., tbe questionins by tbe IUstices in tbe recent civil rishts case of Gomez-Bethke v. U S. Jaycees (1983)) and to justify and explain tbeir c:onclusi0111 (e.s., see tbe opinion of Justice Pitney and tbe dissent of Justice Brandeil in tbe Iandmark corporate tax cue of Eimer v. Macontbu (1920)).
'· IDterteavtaa or Jllltlllcatioa, EqUutioa ud Arpmeatatloll
Waterman reminds us tbat explanation (explainins how you reached a conclusion) and justification (providins a c:onvincins argument tbat tbe metbod used to reach tbe conclusion il valid) aren't u easy in law u tbey are in some otber domains which have clear underlyins (deep) models of tbe mechanilms involved, mechanilms tbat can provide predictive and explicative power. He il joined by McCarty in tbil concern.
Waterman asks us to consider tbe question, "What are tbe fundamental models in tbe lesal domain and what constitutes a real "justification" of an answer?"
Of course, tbe lesal community in its answer would surely mention tbe doctrine of precedent and tbe use of cases u fundamentals in justification and argumentation. The Al community would emphuize representation and models of procesa. Asbley, McCarty and Rissland all cite tbe important role of hypotheticals in justification and argumentation. Ashley and Rissland so further to also say tbat cases and hypotheticals can carry a large pan of tbe burden of explanation u well. Thus Waterman·s question can be re-posed to ask for tbe details of how c:ues are used in justification, argumentation, and explanation.
Dyer points out tbat tbil interleavins hu metbodolosical consequences: process models of lesai expertiJe must be deliped u a whole, where belieft, soalJ, plannins, and arsuins are integrated.
7. Tlle Law u · LearaiDa System
On tbil issue tbere il some dilagreement. Rissland can view tbe law, witb all its cases, rules, statutes, etc, as a learning system in tbat tbe law responds to its environment which presents it c:ues to be dealt witb and in response chanses its case bue, rules, statutes, etc., u well as its ways of dealins witb its environment. 'Ibis view is elaborated somewhat by Asbley:
Experience in tbe lesai system is accumulated in two places: in its database of c:ues and in tbe implicit rules tbat develop for comparins fact lituations in a particular domain. These rules are leamed by law students in learnins to "thint like lawyers." They are evidenced by tbe choices tbat courts make in applyins precedents but tbey are not autboritatively set fortb in any restatement, statute or case. The rules make it poai.ble to apply tbe precedents to new cases. They are how tbe law learns.
McCarty sees tbe law u a "tbeory construction" system tbat is constructins new concepts.
Dyer feeb strongly tbat it is tbe lawyers who leam:
The law il Nor a learning system; lawyers ARE. Lawyers index lesai epilodel in epiiOdic memory. As new lituations arile, new judsements are made and epilodel are reorpnized in memory. ThiJ il just how tbe law student learns: by being presented witb a cue, applyins whatever lesai and common sense intuitions are currently available, indexins the case aloog rec:ognized abstract legal issues, and then modifying and reindexins as new cases (with alternate outcomes and opinions) are encountercd.
8. Iutelllgeat Aids for Practlclaa IJtJpton
The panel as a whole voiced somewhat cautious optimism that it would eventually be poaible to develop a lawyer's workbench which would include tools rangi.ng from standard retrieval tools like the existing WESTI.A W and LEXIS full text retrieval sy~tems, document generatioo aids, scheduling and calendar managen, to tools needing more intelligence like briefins assistanti and interpretive analysis programs which could understand cases. This cautioo is based upon the nearly common experience of how long it takes to develop a program that can handle a few cases or problems, Iet alooe the plethora occ:urring in real practice.
McCarty sees two main categories of intelligent applications programs: conceptual legal retrieval sy~tems and legal analysis and planni.ng sy~tems. He elaborates:
" ... the most critical task in the development of an intelligent legal informatioo system, either for document retrieval or for expert advice, is the construction of a conceptual model of the relevant legal domain .
... these models will not be easy to formulate and the correspondi.ng information sy~tems will not be easy to construct. For the near term, then, the critical problem will be to select an appropriate Ievel of conceptual detail, and an appropriate Ievel of system complexity." [McCarty, 1983&, p. 286).
9. lmpllcatloas aacl Relevaace for Other Case-Bued Doma1lll
Most ot the paneldts did not expraa much enthusiasm for useful cross-fertilization between law and other domains, like medicine, except for Rissland who believes that there is something lilared in that both domains use cases - albeit in different ways - especially in areas where the diseases are new or not weil understood (i.e., where there might not be rules). All fcel that the details of such cross-discipline enlightment are vague at this point.
Dyer puts it:
I believe that modeli.ng lawyen is a much more complex task than modeli.ng doctors. This is mainly because the expert systems approach has restricted itself to those areas where immediate succea is poaible. H we admit that medicine is not limply rule-based diagnosis over symptoms with weighted and propagated certainty factors, then we open up the barder issues in medicine. These issues are those lilared with law: namely, . how do doctors (lawyen) read and understand medical (legal) texts? What must doctors (lawyen) k.now about the physiological (soc:io-c:ultural) prcx:eaes of the body (everyday affairs) in order to be able to understand, plan and give advice?
Such concerns are also lilared by McCarty . who points out that the lllallow reasooing exhibited by c:ertain medical diagnosis programs, like MYCIN, is not at all like, or sufficient for, the deep reasooing (for instance, with respect to its representatioo and causal models) required in the legal domain. McCarty recapa some of bis rec:ent comments from [McCarty, 1983a] where he uked, "lf lllallow rule-based systems have been 10 succeaful in the medical field, why do I insist that they have such serious limitations in the legal field?" McCarty's answer to this paradox lies in the differences of the nature of the rules involved. In medicine, the rules are empirical, associative, probablistic rules of thumb, which are used cumulatively and which do not reflect any deep causal models, say of bacterial disease. Whereas in medicine, this might be enough to get the job done, in the law it is not except for c:ertain discretiooary legal issues like "reasonable care". Interestingly, rule-based treatment of discretiooary judgements (e.g., the worthwhileness of a civil suit) has already met with some succea (Waterman and Petersoo, 1981].
There is the usual two way relationship between AI and the law. Tbe law is a field which might be able to profitably use AI coo.cepts and technology to further its own aims such as developing ;uisprudential analyses of legal reasoning or developing intelligent Iitigation aids and the law provides an excellent task domain for AI research.
Dyer addressc:a whether these two approaches entail different standards and whether this matten:
For a persoo not interested in modeli.ng intelligence, but in supplying the lawyer with smart legal aids, the approach may be different. I believe that this particular subdiscipline for AI is too young to decide that one (or the other) approach lllould not be tried, but I believe that more is to be gained ultimately by modeling the mind in the domain of law than modeli.ng law usina currently known insigbts about intelligence (artifical or otherwise).
Gardner feels that it might be a difference more of degree than kind:
At ooe end, there are people trying to build sy~tems useful to lawyen, or trying to find out how far present AI techniques are applicable io law. The .-dta bere may Iook like lots of law and not mucb that's new for AI.
At tbe otber end, for the purpose of studyioa some AI problem, you could create a microworld and call it law. The .-dtl miabt bave u little to do with a lawyer's view of law u Winltoo's "Sbake.peare world" bu to do with an Bqlilb profe~~or's view of Sbake.peare.
But the reuoo I said the distinctioo wu a matter of depee wu that, to some atent, we can't belp workina with mic:roworlc:ll. That il, every study of law or lepl reuooin&. AI or otberwile, bu to work with an abltracted venioo of ita domain. lt's m01t importaßt to try to be dear about wbat limplificati001 ooe il makina, and to COillider from wbat point of view the.e limplificati001 are reuooable.
Several panelilta feel that progrell in the area of AI and law requirea further progrell in fundamental AI and c:opitive sc:ience raean:b oo IIIUes like repraentatioo. 1bil il a point empbuizecl inc:reuinaiY by Mcearty:
Succellful work in law and AI - botb theory and applicatioo - requirel mucb better "conceptual modelt" of the lepl clomain than the field of AI can currently proviele "off·thHbelf".
Part of Mcearty's re1p001e to . thil problem il bil work oo "permiili.001 and oblipti001" wbicb addreaea the fundamental "Hobfeldian" mecbanilms of riabta, privileges, duties, etc. unclerlylna our lepl system. Otber work, like that of c:leBessooet [1984] oo lepl repraentatioo primitives, also contributes to thil buic raean:b.
Otber IIIUes of conc:ern to the panel induded sociolopcal Qlues, for imtanc:e, the Impact of intelllaent or (semi-)automated lepl services and bow thil will effect society's relatiOillbip with the lepl system (e.a., will lepl services bec;ome ac:c:etlible to more people, will people become more "do it younelf"). Wbat will be the nature of the symbiolil between man and macbine in the lepl domain. Of COUJII), there wu ailo the IIIUe of the law appliecl to AI proarams, for inltance, potential tort, or even criminal, acti001 involvina AI proarams (e.a., a meclical malpractice su.it apinlt a meclical expert system). Many of the.e IIIUes are addre.ecl in the panel oo "AI and Social Relpoosibility". cbairecl by Magie Boc:leo, and reported Oll in the.e proceeclinp.
Albiey, K. D. (1984). "Reuonina by Analogy: A Survey of Selec:tecl AJ. Research with lmplicatiODJ for Legal Expert Systems". In Proceeding1 Firn Anltlllll Coriference tHJ Law and Tecllnology, Univenity of Houstoo, Houstoo TX. To be ioduded io a collectioo to be publilbecl by West Pub. Co. io 1985.
Aupst, S. B. and Dyer, M. G. (1985). "Underltandina Analopet in Editorials." UCLA AI Lab Memo.
Birnbaum, L. (1982). "Argument Molecules: A Functional Representatioo of Araument Struc:ture". In Proceeding1 AMI-82. Pithburgb, PA.
c:leBessooet, C. G. (1984). CCUPS, TM Civil Code Legal ln/ormDtioft Proce~ling Syltem. Law and AI Projec:t, Louiliana State Law Institute, Batoo Roup, LA.
Corbin, A. L. (1961). The 1udicial Proc:al Revilitecl. Yale Law JDUTMl 71, 195-201.
Dyer, M. G., cl Flowers, M. (1984). "''oward Automatins Lepl Bxpertile." In Proceeding1 of tM Firlt Law and Tecllnology Coriference. Univenity of Houstoo, Houltoo, Texu.
Flowers, M. {1982). "'n 8einJ Cootradictory". In Proceeding1 of AMI-82. Pittlburp, PA.
Flowers, M. and Dyer, M. G. (1984). "Really Arpi.na with your Computer in Natural Languap." In Proceeding1 of NCC~. Lu Vegu.
Flowers, M., McGuire, R. and Birnbaum, L., (1982). "Advenary Arauments and the Loaic of Penonal Attacks." In Lebnert and Rioale (ecll.) Strategk1 for Natural Lturg1111ge Proce~ling. LBA Prell, NI.
Gardner, Anne v.dL. (1984). All AnificUil lnlelligence ApprDGCII to Legal RetutHiing. Doctoral Oillertatioo, Deputment of Computer Science, Stanford Univenity.
-- (1983). "The Dellp of a Lepl AnalytiJ Program." In Proceedillg1 AMI-83. Wublnatoo, DC.
Goldman, S. R., Dyer, M. G. · and Flowers, M., (1985). "Prec:edence-Balecl Leaal Reuootna in Contract Law ." UCLA AI Lab Memo.
Hobfeld, W. N. (1919). "Fundamental Lepl Cooceptions u Appliecl in 1udicial Rea.,.,lna." Yale U. Prell, New Haven, CT, Bd., W. W. Cook.
Kubn, T. S. (1970). Tlle StrlletUre of Scknlf/ic Revoluttoru. Sec:ood Bditioo. Univenity of Cbicaao Prell.
Lakat01, I. (1976). Proofs and Refutatiotu. Cambridge University Press, Londoo.
Levi, E. H. (1949). An lntroduction to ugal Reasoning. University of Chicago Press.
Llwellyn, K. N. (1930). The Bramble Bush. Ocean Publications.
----(1960). The Common Law Tradition: Deciding Appeals. Little Brown, BOiton.
McCarty, L. T. (1983a). "Intelligent Legal Information Systems: Problems and Prospects." Rutgers Computer and Technology Law Journal 9(2):265-294.
(1983b) "Permissions and Obligations." In Proceedings /JCA/-83. Karlsruhe, W. Gennany.
(1977). "Reßections on TAXMAN: An Experiment in Artüicial Intelligence and Legal Reasoning." Harvard Law Revkw 90:831-93.
McCarty, L. T. and Sridharan, N. S. (1981a). ~e Representation of an EvolvinJ System of Legal Coocepts: ll. Prototypes and Deformations." In Proceedings /JCA/-8/.
(1981b). A Computational Theory of ugal ArgWMnl. Technical Report LRP-TR-13, Laboratory for Computer Science Research, Rutgers University.
McGuire, R., Birnbaum, L., and Flowers, M. (1981). "Opportunistic Processi.ng in Arguments". In Proceedings /JCA/-81. Vancouver, B.C.
Polya, G. (1973). lnduction and Analogy in Mathemlllics, Volumes 1 &: 2. Princeton University Press, New Jersey.
Peterson, M. A. (1983). New Tools for Reducing Civil Litigation E:cpenses. Memo R-3013-ICI. The Institute for Civil Iustice, The RAND Corporation, Santa Monica, CA.
Reichman, R. (1981). "Modeling Informal Debates". In Proceedings /JCA/-81. Vancouver, B.C., August.
Rissland, E. L. (1984a). "Argument Moves and Hypotheticals". In Proceedings First Annual Conference 0r1 Law and Technology. University of Houston, Houston TX. To be included in a c:ollection to be publisbed by West Pub. Co. in 1985.
(1981). E:cample Generation. Technical Report 81-24, Department of Computer and Information Science, University of Massachusetts, Amherst, MA.
(1982). "Examples Hypotheticals in Contract Law". Annual Cognitive Science Society Michigan, Ann Arbor. in the Legal Domain: In Proceedings F ourth Conference. Univ. of
---- (1983). "Examples in Legal lteuonins: Legal Hypotheticals". In Proceedings /JCA/-83. Karlsruhe, W. Gennany.
---- (1984b). "Hypothetically Speaking: &perience and Reuooing in the Law". To appear in Cognitive Science. An earlier version appeared in Proceedings First Annual Conference Oft Theoretical /ssues in Conceptual lnformotion Processing, Georgia Tech, March.
(1984c). ~e Ubiquitous Dialectic". In Proceedings Si:cth Ewopellll Conference 0r1 Artificiol lnlelligence (ECA/-84), held in Pisa, Italy, September, publisbed by Elsevier (North-Holland).
(1978). "UnderstandinJ UnderstandinJ Mathematics". Cognilive Science, Vol. 2, No. 4.
Rissland, E. L., Valcarce, E. M., and Asbley, K. D. (1984). "Explaining and Arguing with Examples". In Proceedings AAAI-84, Austin, Texas.
Waterman, D. A., and Petenoo, M. A. (1981). Models of Legal Decision Making. Memo R-2717-ICI. The Institute for Civil Iustice, The RAND Corporation, Santa Monica, CA.
@inproceedings{ijcai1985p250,
title = {AI and Legal Reasoning},
author = {Rissland, Edwina L.},
booktitle = {Proceedings of the Ninth International Joint Conference on
Artificial Intelligence, {IJCAI-85}},
publisher = {International Joint Conferences on Artificial Intelligence},
editor = {Aravind Joshi},
pages = {1254--1259},
year = {1985},
note = {General},
url = {https://ijcai.org/proceedings/1985/250},
}
TY - CPAPER TI - AI and Legal Reasoning AU - Rissland, Edwina L. T2 - Proceedings of the Ninth International Joint Conference on Artificial Intelligence, IJCAI-85 PB - International Joint Conferences on Artificial Intelligence PY - 1985 SP - 1254 EP - 1259 UR - https://ijcai.org/proceedings/1985/250 ER -Copied
Machine-readable versions of this paper, for tools and AI agents: