® InfoJur.ccj.ufsc.br
Instituut voor Taal- en Kennistechnologie
Institute for Language Technology and Artificial Intelligence
As with previous Artificial Intelligence and Law conferences, this one began with a number of tutorials. Professor Donald Berman of Northeastern University Law School and Professor Kevin Ashley of University of Pittsburgh Law School gave an introduction to AI and Law. Not only are they two of the leading practitioners/theorists in AI and Law, they are also gifted teachers. Professor Ashley's presentation about the potential uses of case based reasoning paradigms was particularly enlightening.
The other tutorial was equally informative and saw Dr. Howard Turtle of West Publishing Inc. (the proprietors of the mammoth Westlaw database) discussing 'Text Retrieval in the Legal World'. This tutorial concisely developed a theory of legal information retrieval - a vital topic to those interested in the development of legal knowledge based systems; but one which is often overlooked. Dr. Turtle ran through a deeply knowledgeable discussion of the special considerations in legal information retrieval, some information retrieval models and a range of philosophical questions about the ability of using these systems. He was instrumental in building the Bayesian Inference Network model into the WIN (Westlaw Is Natural) user interface, and so his discussion was extraordinarily deep.
At the same time Andrew Jones of the Norwegian Research Centre on Computers and Law, University of Oslo and Marek Sergot, Department of Computing, Imperial College London conducted a workshop with title 'On the Theory of Normative Positions and its Applications'. It was well attended and very successful.
The main conference opened and ran for three days with no parallel sessions, and so one was not caught having to decide between two equally interesting presentations. About forty papers of varying lengths were presented.
Though there were a broad range of topics, approaches and ideas given, it would perhaps be fair to identify two main themes: the use of exotic logics and the development of sophisticated case based reasoners. The exotic logics have largely been and remained the preserve of the Europeans. Thus a number of non-monotonic and deontic logical formalisms were presented, including ones which could handle inconsistencies and, to a degree, case based reasoning. Particularly for those brought up in common law countries the use of exotic logics is unconvincing. The logicians always seem to say that they can find a way to get around the difficulties of representing legal realities (like inconsistencies, open texture, cases, etc.). Yet the approaches they suggest always appear murderously complicated to implement, inelegant, and artificial.
The other major area of interest was the development of a range of case based reasoning systems. Some of these, such as Ashley and Aleven's (described below), allowed for sophisticated legal reasoning on the basis of a given case base. Others combined case based reasoning with rule based reasoning to provide legal decision support for lawyers, paralegals and judges.
We describe briefly below a number of papers which we thought were of fundamental significance.
Thomas Gordon of the German National Research Centre for Computer Science presented a paper entitled 'The pleadings game: Formalizing procedural justice'. It proposed a formal model of civil pleading based upon Alexy's discourse theory of legal argumentation. The sophistication of his approach allows for arguments about validity and priorities of rules at any stage in the 'game'. This accords with the legal practitioner's view of the use of positivistic rule systems, and so may well be the basis for practical systems in the future.
Kevin Ashley and Vincent Aleven of the University of Pittsburgh had a paper expanding upon their earlier work. It was entitled ' What law students need to know to WIN'. The title is somewhat misleading, as we thought that it would deal only with Westlaw Is Natural (WIN), but in fact it was more generallyconcerned with the question of formal models of legal reasoning, and the implementation of a system to show law students these models of reasoning. This system, called CATO, seeks to teach case based reasoning to law students and is interesting for a number of reasons. First, it represents one of the few attempts at implementing legal reasoning formalisms outside typical logical deduction and induction (or abduction for that matter). Further, it showed an interesting diagrammatical interface for students in adopting and distinguishing cases given a set of facts.
Karl Branting, who has written an article for this edition of Think, presented a paper on 'A reduction graph model of ratio decidendi'. This paper showed the legal concept of ratio decidendi as a series of reasoning steps represented by a reduction-graph. Once again, Karl's use of jurisprudential theory marked this paper as a subtle and useful application of AI technology and methodologies to legal reasoning requirements. It further showed fundamental difficulties with the current crop of 'naive' case based reasoning paradigms.
The paper written by Don Berman and Carole Hafner, was entitled 'Representing teleological structure in case-based legal reasoning: The missing link'. It proposed a model and demonstrated an implementation which went beyond the naive fact-based approach in most case based legal reasoning systems. The problem with fact-based systems has always been their inability to match on policy/principles, and hence they have always had limited application. The model enables the retrieval and use of teleological information, rather than simply surface level factual detail. This approach holds great promise for more useful and jurisprudentially justifiable knowledge based systems in the future.
Apart from these papers which were important from the perspective of pure research, there were a number of other paper worthy of comment as being interesting and useful applied systems. Some of these were commercial while others were in the public sector, but their most obvious characteristic was that they were often presented by Australians. The majority of practical systems described in the conference, and the vast majority of systems actually demonstrated as up and running, came from various Australian researchers. Thus there was Mowbray and Greenleaf's 'Privacy Workstation', Softlaw's major expert system shell and hypertext engine together with a number of huge systems built for government departments in Australia and New Zealand, a Chinese law system built by Graham Brown of Newcastle, and the 'Credit Act Advisory System' (see elsewhere in this issue), built by George Vossos and others for Allan Moore & Co., solicitors in Melbourne. Many commented on the fact that it was the Australians above all others who had actually begun implementing systems with the theories previously expounded.
Amsterdam and the Netherlands were an ideal setting in which to hold the Fourth International Conference on Artificial Intelligence and Law. The Netherlands is the home to many significant Artificial Intelligence and Law projects: in Amsterdam (University of Amsterdam and Vrije University), Leiden, Maastricht, Rotterdam, Tilburg and Utrecht. The fifth conference will be held at the University of Maryland in May 1995, whilst the sixth conference will be jointly conducted by La Trobe University and the University of Melbourne, in Melbourne, in early 1997.