® InfoJur.ccj.ufsc.br
Paul Bratley[1], Daniel Poulin[2], Jacques Savoy[1]
[1] Département d'informatique
[2] Centre de recherche en droit public
Change is a fact of life, in law as elsewhere. The law must be stable if it is to regulate effectively the affairs of those subject to it, yet it must be open to correction when its rules need revising. For designers of legal expert systems and legal hypertext systems, coping with change is a constant challenge.
A review of the recent literature suggests that this problem is just beginning to be taken seriously. We describe some of the difficulties which must be overcome, and review some suggested solutions. Our conclusion is that most of these provide little practical help, and that the problem is still entirely open.
Change is a fact of life, in law as in any other field. The law must be stable if it is to regulate effectively the affairs of those subject to it, yet it must be open to correction when its rules need revising on account of errors, changing social reality or change of policies. In the rapidly-evolving modern world, such corrections are increasingly frequent, and a major item of social legislation may be amended several times a year, or even several times a month.
For designers of systems intended to give legal advice, or to help lawyers prepare cases and opinions, coping with change is a constant challenge. In a legal expert system, every change in the law can have repercussions on the knowledge base. In the worst case, it might require the entire knowledge base to be rebuilt. In a legal hypertext system changes in the law cause links to be deleted, added and changed. In both applications it must be possible to reconstruct the state of the system as it was at any previous time. The challenge is to make the necessary alterations at a minimum cost in effort, errors and inconvenience.
The long-term aim of our group in Montreal is to produce a system for lawyers which will integrate a legal expert system with a corresponding hypertext. For example, modules of the expert system may help in navigating through the hypertext and answer questions about the legal concepts involved; links in the hypertext will provide a way for the expert system to produce explanations of its conclusions and may help when the system has to be updated. Although we are still far from this goal, pilot projects are under way in the area of unemployment insurance. The need for a detailed study of the problems created by changes in the law was brought home to us last year, when the Canadian Unemployment Insurance Act was substantially amended. Overnight, many of the rules in our expert system needed revising, and a sizeable collection of cases which we had analysed in view of experiments in case-based reasoning became largely irrelevant.
Although most published work on legal expert systems pays lip service to the idea that they must be designed to cope with change, a review of the literature provided little practical guidance. The present paper is therefore an attempt to pose the problem more concretely. We first outline rapidly the kind of changes in the law that may be encountered, and indicate their effects on both the expert system and the hypertext components. Next we summarize the attitudes to change which we found in the literature. Finally, although we have no magic solutions to suggest, we indicate where future work should concentrate.
The law is created and evolves through statutory changes, through bilateral or multilateral instruments, through the application of customs, and through precedents. For the purposes of this paper, we restrict ourselves mainly to the first form of creation of the law. The changes we encountered when the Unemployment Act was amended seem to fall into four rough categories.
In one section of the Act, the order of the paragraphs has been inverted between the old and the new provisions. What is considered the main rule now precedes the exception. The terminology has been made "gender-neutral". In another, the words "who qualifies to receive benefit" have been added in a context where they were already implicit. In a third, the requirement that an applicant have twenty weeks of insurable employment before qualifying for a certain category of benefits is replaced by a table.
In some cases tests or provisions already present in the Act are reused in a different context. For example, section 10(4) has been modified to include a provision already found elsewhere dealing with interruption of earnings.
The old Act provided that where the applicant refused work without good cause, lost his employment as a result of misconduct or left voluntarily, benefits would not be payable for a "penalty period [...] not exceeding six weeks." This provision now reads that the number of weeks of disqualification "shall be not less than seven and not more than twelve." On the surface, this looks like a minor change. For the expert, this is not the entire answer. Since the disqualification now lasts at least seven weeks, the expert expects that authorities deciding whether work was refused without good cause, whether the applicant misbehaved or whether he left voluntarily will interpret these clauses more restrictively to avoid imposing an unduly harsh penalty. Technical changes aimed at "tightening up" the Act thus lead to conceptual changes.
Section 28(1) of the Act reads: "A claimant is disqualified from receiving benefits under this Part ... if he voluntarily left his employment without just cause." The voluntary character and the absence of just cause are concepts that are not directly observable but are left to be specified by decisions rendered in individual cases. They require "tests" in terms of directly observable facts.
Section 28 was further changed by the addition of a new paragraph, which reads:
"(4) For the purpose of this section "just cause" [...] exists where, having regard to all the circumstances, [...] the claimant had no reasonable alternative to immediately leaving the employment."
This redefinition of the concept of "just cause" contains two new concepts, "no reasonable alternative" and "immediately," which must be included in our system.
Although these isolated examples cannot serve as a base for wide generalization, it does seem clear that the law generally tends to change incrementally, which means, in a majority of cases, without destroying the knowledge base on which it functions. Rather, the law builds on the existing base, accumulating and superposing rule upon rule. Moreover, it appears that statutory changes do not effectively change the law overnight. In many cases, the expert has no clear answer to the legal difficulties raised by a modifying provision. In such cases, he integrates the new rule into his knowledge base, but without necessarily drawing definite conclusions on its real impact, waiting for subsequent judicial or administrative determination.
Besides the numerous changes caused by major amendments to the relevant law, systems for legal applications must also deal with a constant flow of smaller ajustments. Since 1985, for instance, the Canadian Unemployment Act has been amended seven times, including the major overhaul discussed above. The associated Regulations have been changed on 35 occasions, or about once every two months. Evidence that this is not unusual is provided by [Bench-C. 91]. The authors of this paper, concerned with the field of work-related injuries, estimate that they will have to deal with about 20 new Statutory Instruments and 10 technical instructions each year, as well as several dozen relevant judgements, policy changes by the employer, and so on.
Insignificant changes in the law have -almost by definition- little effect on legal applications. In the expert system, at the most, some explanatory texts may have to be changed. The hypertext must also be brought up to date. Minor changes can be handled easily. As far as the expert system is concerned, they may require the addition of one or two local rules, but the knowledge base does not have to be restructured. Again, the hypertext component must reflect the changes.
Similarly it does not seem difficult to introduce new uses of existing concepts. This may simply require adding a reference to an existing set of rules in the expert system, or adding links to an existing module in the hypertext component. However it will often be necessary to review the existing rules and modules to ensure that they are sufficiently general to function in the new context. In every case re-using an existing concept requires that an expert jurist check to see that two apparently analogous concepts do not conceal an essential legal difference.
Other changes in the law are less easily handled; it is to these problematic instances that we wish to draw attention. Modifications whose effect is apparently minor, but which in fact change the law in unforeseen ways, may be encountered at any moment. Indeed the existence of such phenomena serves as a warning against adopting an unduly descriptive attitude to the law. Furthermore they underline the fact that the law may often be uncertain, and that this uncertainty must be reflected in the computer system.
Finally the introduction of new concepts, or major changes to existing ones, both require considerable effort. If this effort seems inevitable in the former event, since there is no way to avoid giving new rules to the expert system, and adding links to relevant jurisprudence in the hypertext component, in the case of a modification we may hope to do better. If expert lawyers bring their knowledge up to date by building on the base they already possess, it would seem wasteful to rebuild the expert system's knowledge base or the hypertext component's existing links by starting again from scratch. Yet it is not obvious how to do better than this. Even incorporating a restriction on an existing concept, as when "just cause" is defined more narrowly, seems to necessitate a complete restructuring of the rules and the links. To make matters worse, a truly useful legal expert system or legal hypertext system must be able to answer questions not only concerning the current state of the law, but also about the law which applied at some previous date. In any new structure of rules in the knowledge base, or of links in the hypertext component, it must therefore still be possible to reconstruct the previous states.
The design of a computerized system usually requires that the procedures to be computerized be relatively stable; alternatively, if the system lifetime is short, the profit to be realised makes the game worth the candle. Similarly, certain branches of the law, although subject to frequent modification, are nonetheless worth studying by researchers in artificial intelligence. This is particularly the case where the law affects large numbers of people, where the sums of money involved are enormous, and in areas which have resisted attack by more traditional techniques. Tax law and much social legislation are among those heavily used and frequently changed domains.
In such rapidly-evolving systems the methods used in artificial intelligence are often preferable to the procedural techniques of traditional computer programming; they are easier to adapt to less stable problem areas. In particular, artificial intelligence uses an explicit representation of the necessary knowledge, making it relatively easy to determine those areas of the knowledge base which correspond to a particular part of the problem domain. Thus when this domain is modified, it is easier to modify the corresponding structures in the knowledge base.
While it is certainly true that modifying a knowledge base is easier than modifying a program where the rules are represented implicitly, this does not mean that such modification is simple. The following paragraphs survey some proposals which have appeared in the recent literature.
[Smith 87] deals with an area which is particularly stable. The rules used in his "Nervous Shock Advisor" are constructed on the basis of a thorough analysis of all the relevant cases, of which there are only 74 this century. Such a system might almost be designed never to change. Any major modification of the law would require the whole analysis to be reworked.
The area chosen for Schlobohm and Waterman's EPS system [Schlo 87] is the rapidly changing law to do with estate planning. The knowledge is provided by an expert. However as tax law evolves rapidly, this approach led to problems. Schlobohm and McCarty subsequently described them in the following terms:
[EPS] does not contain any knowledge about how the rules were obtained [...] As a result, human experts would have to modify the heuristic rules whenever the law changes, and the entire system containing the new rules would then have to be debugged. Finally, EPS [...] would be of little use to expert estate planners, since experts only need assistance when the law changes or when dealing with a novel client situation. [Schlo 89, p. 9]
More recently Schlobohm and McCarty proposed a new estate planning system: EPS II. In EPS II, the expert's heuristic knowledge is complemented by an explicit representation of the rules laid down by the Internal Revenue Code. Both these kinds of knowledge are expressed in the form of propositions in LLD, the Language for Legal Discourse [McCarty 89]. When the system is used, prototype solutions furnished by the human expert can be modified or "deformed" to satisfy the needs of the client and the constraints of the law. The authors show how certain kinds of change to the Code can be handled by EPS II. However not all changes can be taken care of in this way. Any modification of the Code which adds a new concept or radically changes an existing one is beyond the powers of EPS II.
In the area of statute law, one favourable approach to obtain an easily - modifiable system is to use a formalization which makes articles of the statute correspond to rules in the knowledge base. In a legal expert system based on such a formalization, every rule has an explicit link back to some clause of the legislation, of which it is a formal paraphrase. This is the line taken by Kowalski and Sergot and their team at Imperial College.
From the point of view of ease of maintenance, this is ideal. Modification of a paragraph in the text of the law simply implies that a corresponding modification has to be made in the associated rule in the knowledge base. However such a logical formalization cannot serve as the base of an expert system unless one adds a layer of heuristic knowledge, and other general knowledge of the kind used by jurists. Kowalski and Sergot admit that their model represents "a layman's reading of the provisions. This in itself renders our British Nationality Act program of limited practical value" [Kowalski 90, p. 207]. Such a program will need the help of a human expert, to correct, modify and augment the formalization; but once the formalization is augmented in this way, modifying the system following a change in the law implies modifying implicit or explicit rules which do not correspond directly to paragraphs in the law.
At the University of Liverpool, Bench-Capon's group also insist that the knowledge representation must be closely aligned to the sources of knowledge. In a recent paper [Bench-C. 91], they explicitly acknowledge the problem of coping with change, stating that "the greatest barrier to the routine use of knowledge based systems techniques for practical legal applications lies [...] in the problems associated with the maintenance of such systems." The group aim to provide tools to handle day-to-day changes due to minor modifications of the legal texts and the application of the law. It is still too early to judge the value of their techniques.
In the legal field, the case-based approach has been explored principally by Rissland's group at the University of Massachusetts. According to [Skalak 89], the weaknesses of legal expert systems that can be alleviated by case-based reasoning include their inertia, that is, the difficulty they have in assimilating new knowledge. The essential idea is that it is easier to add new cases to a collection than it is to add new rules to a rule-base. However, new "dimensions" of a legal concept may make their appearance, making it necessary to re-index all the cases using this new criterion.
Rissland and Skalak have also proposed using such a case-based reasoner with a more traditional rule-based system. In that context, they propose using ID5, a classification algorithm [Utgoff 89], to induce new rules corresponding to the modified case base. They have carried out a number of experiments with this in view [Skalak 90], so far with mixed results. We are forced to conclude with Rissland and Skalak that "there is no quick cure for this fundamental problem."
Legal applications offer an attractive field for the use of hypertext techniques. Shneiderman [Shneider 89] formulates the "golden rules of hypertext" for deciding when these techniques are applicable and when they are not. His rules may be summarized as follows: there exists a large body of information organized into numerous fragments; the fragments are related one to another; and the user needs only a small fraction of these fragments at any one time. A moment's thought is enough to show that legal texts satisfy these criteria. Statutes are divided into sections, paragraphs and articles, related to one another, and usually not all applicable at once; and a collection of cases can also be seen as a set of related opinions and commentaries on various aspects of the law.
It is therefore not surprising that the idea of using hypertext in a legal context is not new. [Yoder 89] provides an example in the area of patent law. Wahlgren [Wahl 89] has also studied different ways of decomposing the law to make a hypertext, and proposed a prototype to be combined with an expert system. A panel discussion at Hypertext '89 was devoted to systems combining expert systems with hypertext. However there seems to be little published work on dealing with change, and more particularly on how to take advantage of the complementary strengths of the two paradigms to deal more efficiently with changes in the domain concerned.
Nevertheless some light is thrown on the problem by [Delisle 87], who is concerned with a system for cooperative authoring of technical documentation. The proposed solution extends the notion of link and node by including a version tag for these components. On another level, some authors make a distinction between form and meaning to allow some links to be constructed heuristically. If such heuristic techniques can be implemented, they will help alleviate the problems caused by change. [Fung 90] establishes links between technical descriptions and the concepts involved; [Savoy 91] on the other hand links such descriptions to users' information needs.
Further references to the problem of dealing with change can be found in [Oskamp 90, Quast 89, Smits 89, Weusten 89] and in our own work [Poulin 88]. Oskamp devotes much of the penultimate chapter of her thesis to the problem of maintenance. Quast and de Wildt have built a rule base to deal with the concept of "commensurate work" in Dutch unemployment insurance law, based on an expert analysis of some 150 cases. They propose a structure which, they believe, will allow the rule base to be updated without a complete reorganization. Smits, Kracht and Weusten characterize the changes which may be necessary in a legal expert system, and propose some elementary measures to facilitate system maintenance.
In a more general context, the problem of change in a formal logical system has been studied by various authors. In classical logic, if new information contradicts previous assertions, then the system collapses completely. The aim of recent work has been to explore how contradictions can be avoided by withdrawing as few conflicting assertions as possible. Katsuno and Mendelzon [Katsuno 89] give an excellent introduction to this area which, while promising, is still far from having significant practical appliations.
The problem of adapting any computer system, expert system or hypertext to changes in the law is very real. Although some changes pose few problems, other essential aspects of the law, such as the existence of loosely-defined concepts, involve complex issues. The services of an expert jurist will be required throughout the life of the system, not simply when it is being created. And once the legal issues have been taken care of, many difficult technical problems remain.
For expert systems, structuring the rules of the system to correspond to the structure of legal texts, and taking advantage of the ability of case-based reasoning systems to integrate certain new cases without too much trouble, are approaches which, while useful, are for the moment limited. In the hypertext community, it is accepted that some systems will have to keep many versions of a text and "filter out" links and nodes which are not relevant in a given context. However there is no agreement on how to design the interface to such a multilayered system, nor how to structure it so that changes are easy. Finally, formal approaches to change in logical representations of knowledge also offer hope. However none of these lines of attack has yet proved its worth.
It remains to be seen if the whole is greater than the sum of the parts. Is it indeed true that combining a hypertext with an expert system can contribute to managing some of the links between the rules and the legal text? In the other direction, will the use of systems built round a knowledge base help edit the links in a hypertext following modifications in the law? These are some of the questions we intend -of necessity- to explore in the coming months.
The work described in this paper is supported by a grant from the Social Sciences and Humanities Research Council of Canada and by the Fonds FCAR of the Québec government under its program of Actions structurantes.