® InfoJur.ccj.ufsc.br
Instituut voor Taal- en Kennistechnologie
Institute for Language Technology and Artificial Intelligence
On november 26 the University of Twente hosted the annual Jurix-conference 1993. This year's conference addressed the topics `Intelligent Tools for Drafting Legislation, and Computer-Supported Comparison of Law,' two relatively new themes in the field of AI and law.
Jurix is the Dutch acronym for The Foundation for Legal Knowledge Based Systems (stichting JURIdische KennnisSystemen). Ever since 1987 the Jurix-Foundation provides a forum for scientists in the Netherlands in the field of law and artificial intelligence. To this purpose Jurix organizes regular national meetings on research topics shared by its members and an annual international conference on fundamental topics in the field. The international Jurix-conferences are generally visited by (legal) computer scientists, (legal) practioners and lawyers.
Some 70 participants attended the 1993 Jurix-conference which was held at the campus of the University of Twente. In comparison to previous editions of the Jurix-conference a relatively high number of legal practioners and legislators participated. Especially the Dutch Ministry of Justice and the Ministry of Science and Education were well represented. Furthermore guests and contributors from abroad, with a wide range in nationalities (Australia, Venezuela, Norway, Italy, UK etc.), attended the conference.
The deliberations during the conference were divided into five sessions. The first four sessions dealt with the topic of Intelligent Tools for Drafting Legislation. The last session was dedicated to the theme of Computer-Supported Comparison of Law. During each session papers contributed to the conference (and published by Vermande publishers and presented at the conference-day itself) were orally presented and sequently discussed.
Although almost all of the presentations and discussions were interesting I will not deal with each and every contribution, but I will only discuss some of the highlights. For those interested in a more detailed overview of the topics and contributions I refer to the published conference-papers. The book, containing the papers can be obtained by contacting the Jurix-address listed here below.
The bulk of papers and presentations during the conference was dedicated to the theme of intelligent tools for drafting legislation. As a research topic the issue of computer-assisted legislative drafting is, in fact, not all that new. Especially the pioneers and `founding fathers' of AI and law, like Layman Allen (See L.E. Allen, Language, law and logic; plain legal drafting for the electronic age, in: B. Niblett, Computer Science and Law, Cambridge: Cambridge University Press, 1980, p. 75-100.), saw as early as the 1970's, attractive perspectives for the use of AI-techniques and knowledge based systems in the field of legislative drafting more. The tempting promise of intelligent legislative drafting-systems, built by using AI-techniques, was however for a long time not fulfilled in practice: as of yet there are hardly any really functioning knowledge based legislative drafting systems available.
The first invited speaker to the conference, dr. Philip Eijlander of the Dutch Ministry of Justice, argued that the possibilities of putting AI-techniques and knowledge based systems to use, in order to benefit legislators, are inherently limited, due to some distinct characteristics of the drafting-domain itself. Eijlander considered it inconceivable that computer technology in particular knowledge based systems will take over any of the legislators core activities in the near future. The core of these drafting activities simply involves too many aspects and necessary considerations and the drafter's reasoning is too complex and too reliant on different kinds of knowledge (e.g. socio-economic reasoning, political reasoning, and legal reasoning), particular to a specific drafting project, to be automated. This does, according to Eijlander, however not mean that intelligent IT-drafting tools cannot make very practical contributions to the rationalization of the drafting process and assist drafters in their work. In fact Eijlander sees two major possibilities for drafting-systems using AI-techniques. First of all AI-techniques can be used to guide users methodologically through a body of drafting knowledge or relevant legislative information, secondly knowledge based systems can be used to check the internal (deontological) consistency of drafts.
Eijlander's contribution was followed by a speech from the Dutch State Secretary of Science and Education, dr. M.J. Cohen. Cohen fully endorsed the position put forward by Eijlander. Contrary to Eijlander he, however, focused on an already existing system called OBW, which uses AI-techniques to benefit the work of legislative drafters. OBW (OBW stands for `Ontwerpbank Wetgeving'. See for more information J.G.J. Wassink, Kennistechnologie en het ontwerpen van wetgeving, brochure Ministerie van Binnenlandse Zaken, 's-Gravenhage 1992, and A.M.J.J. Goossens, Ontwerpbank Wetgeving, in: RegelMaat 93/5, p. 173 e.v.), or, in a free translation, LegislativeWorkbench, assists legislative drafters by making legislative information available to the users. Like LEDA which will also be discussed in this issue of Think OBW comprises a methodologicalframework which enables users to find their way through a body of relevant knowledge, and, at the same time, offers substantive methodological assistance in preparing a draft. Actual drafting that is putting legislative or policy choices into words is (contrary to the LEDA-system) not supported by OBW. According to mr. Cohen the first experiences with OBW are very promising and justify further research into the possibilities of AI-concepts and -techniques in the area of legislative drafting.
The two opening speeches by the invited speakers were followed by oral presentations from the contributors to the conference. First in the row was Robert Van Kralingen from Leiden University, who, together with his colleagues Eduard Oskamp and Edwin Reurings, advocated a new life for frames in legal knowledge representation. Van Kralingen, Oskamp and Reurings propose to use so called `norm-frames' to represent legal knowledge held within legislation. The idea of using frames to represent legal knowledge in itself is not new, but the way in which they use the frame-concept is. They do not exclusively use their frames, which are modelled according to the instances of norm or norm-sentences, as a means of formalization but as an intermediate language for conceptualization of domain-knowledge. By using the frames for knowledge-modelling they dispose of a powerful language to express all kinds of norm-logic relations within legislation, thus creating a deontological relations-network which represents legal knowledge far more adequate than the traditional isomorphic or 1 on 1 methods (like the production-rules-method) used for representing legal knowledge. Although Van Kralingen c.s. originally designed their norm-frame-language for representation-purposes they see possibilities to put their notions to work in a legislative drafting-setting. Norm-frames can, according to these contributors, also used for legislative drafting. For instance: the frame structures can be used to determine if a new piece of legislation is (norm-logically) complete, coherent and/or provides sufficient information for those who have to apply or understand it. A brief discussion, which followed the oral presentation, made clear however, that there are quite a number of theoretical and practical problems to overcome before norm-frames can actually be used in legislative evaluation-systems. Especially the interrelations between knowledge expressed in a piece of legislation and other kinds of legal and world knowledge poses a difficult problem. The use norm-frames as an analysis tool and an intermediate language for modelling legal-knowledge are however theoretically most challenging.
A group of contributors from Leuven University (Belgium) adopted a somewhat different approach towards the issue of modelling of legal knowledge in order to benefit legislators. In their paper and presentation Van Buggenhout c.s. advocated a renewed use of the decision table technique to accommodate legislative drafting. Decision tables can, according to Van Buggenhout c.s., be a considered a useful formal representation technique for legal rules with a complex decision-structure. When a legal rule enhances such a complex decision-structure and a lot of factual circumstances are involved, decision tables can make the logical relations and the decision-structure of a legal rule transparent (by visualizing it) for a legislative drafter. Not only can decision tables increase the drafter's overview and insight in a draft-rule, they can also be of assistance in optimizing, verifying and validating these draft-rules. In Leuven a tool called `Prologa' was built for the automated modelling, optimization, verification and validation of drafts using the decision table technique.
The decision table technique has however -- as was made clear during the discussion which followed the presentation -- some major drawbacks. First of all it can only be used for those rules which have a (sufficiently complex) decision structure (quite a lot of legal rules do not have a structure like that), the user-guided modelling of draft-rules is time-consuming, and the quality of the legislative result is largely dependant on the quality of the decision tables. Even a more fundamental objection was raised: decision-tables only support reasoning in a two-valued logical scheme, which rules out the representation of some deontic knowledge-modalities and forms of deontological reasoning, which are quite frequent in legal rules. Nevertheless the presentation and subsequent discussion presented a lot of food for thought, not in the least for Van Buggenhout c.s.
Most of the (knowledge-based) concepts and systems which have in the recent past professed to be potential beneficial for legislators do not concern the totality of drafting activities but only one activity, namely `(logical) legislative evaluation'. Because the predicate-logic and deontologic tradition in AI and Law until recently was very dominant, a lot of concepts and ideas for knowledge-based legislative drafting support mainly aimed at the logical evaluation of draft rules. Concepts, tools and systems developed in this logical tradition profess the combination of AI-techniques and (deontic) logic representation-methods to check the logical consistency of drafts. Quite a number of contributions to the 1993 Jurix-conference were set in this key (Den Haan, Groendijk/Herrestad, Barrag‡n, Svensson). It cannot be denied that these concepts and ideas, which in most cases are transpositions of more general concepts and ideas developed for the purpose of (semi-) automated law-application, do bear relevance for legislative drafting issues as well. There is however a major difference between the reasoning involved in the making/drafting of a legal rule, and the reasoning involved in the actual application of a legal rule in an individual case by a judge or an administrative body; when a legal rule has to be applied, legal reasoning is predominant in the legal decision process, when, however, a legal rule has to be drafted legal reasoning is not predominant at all. The legislative process is equally dependant on different forms of reasoning like socio-economic and political reasoning. From this it follows that the reasoning concerned in legislative drafting is not equivalent to legal reasoning, and, from an AI-point of view, cannot be dealt with in the same way as legal reasoning. Making a total representation of the reasoning of legislators is virtually impossible because it is to dependant on the specifics of the `drafting-case' at hand.
This is the dilemma which faces the advocates of predicate and/or deontic logic, combined with AI-techniques, in the field of legislative drafting: they cannot represent and thus support the drafter's reasoning as a whole, and the parts which can be represented and supported are often trivial. For instance: when a legal rule has to be drafted it is not possible to fix substantively the correct reasoning or correct deontic expression within that rule in advance. It is however possible to represent the reasoning already laid down in a draft rule for purposes of logical evaluation. In most cases however a logical evaluation of a draft-rule is not in great demand. What a legislator wants is a political evaluation or a socio-economic evaluation of his draft. Logic doesn't always pay in a legislative drafting environment. It is therefore not altogether conceivable that systems that require extensive and time-consuming user-guided representation and formalization will meet with much success in practice, if the only reward for the trouble is a mere logical-evaluation.
Construction of knowledge driven-information environments for legislators can mean a way out of the dilemma mentioned above. During the Jurix-conference two papers, containing ideas and concepts inspired by the database-paradigm, were presented (Hage, Voermans/Verharen). The general idea behind these presentations was that legislators are more interested in questions of efficiently and effectively finding arguments to support their case (i.e. a draft) than they are in finding correct answers to legal questions. In order to support legislators in finding (legal) arguments knowledge-driven, interactive, information systems are needed. For this a model or a method for the way in which a legislative draughtsman can or has to find the (relevant) arguments supporting his case in a domain is needed. In order to do this, according to Hage, at least three forms of representation are necessary: first of all an extensional representation, secondly a intensional representation in a formal language and thirdly an intensional representation in natural language. Clear and traceable knowledge concerning legislative issues (like procedural or technical knowledge) then could be used to lead the way through an information-network which contains different kinds of arguments and which in its own turn can generate new arguments, applying the represented and formalized knowledge parts.
In the LEDA-system, which was demonstrated by Verharen en Voermans, an actual implementation of such a knowledge-driven information network was illustrated. I will not further discuss the LEDA-project because a description of the system is already included in this Think-issue.
The discussion following the presentations of Hage and Verharen/Voermans made clear that the more data-base oriented approach towards legislative drafting issues might well be more rewarding than the strictly logic/knowledge-based approach.
The second theme of the Jurix-conference was `computer-supported comparison of law'. The two papers addressing this subject were dealt with in the fifth session of the conference day. Computer-supported comparison of law is a very new item in AI and Law. So new in fact that none of the two papers dealt with the subject at length. Both the contributors on this subject had directed their papers to the neighbouring domain of micro-simulation of social security legislation. Their contributions were both very interesting, but what was compared (using different techniques and different systems) was financial and or socio-economic impact of social benefit legislation, and not the legislation itself.
Readers who are interested in the work of the Jurix-foundation and/or its publications (including the 1993 conference-proceedings) are invited to contact:
mr. C.N.J. de Vey Mestdagh
University of Groningen
Computer/Law Section
Oude Kijk in 't Jatstraat 26, 9712 EK Groningen
Phone: +31 (0)50 635790
Fax: +31 (0)50 635603
E-mail: Sesam@rugr86.rug.nl