Some Requirements for a Computer-Based Legal Consultant

Abstract

Although the literature on computer-based consultation systems has often suggested the possibility of building an expert system in the field of law (see, e.g., [2]) it is only recently that several AI researchers have begun to explore this possibility seriously. Recent projects include: the development of a computational theory of legal reasoning, using corporate tax law as an experimental problem domain C61 C73 C81; the development of a language for expressing legal rules within a data-base management environment [Y]; the design of an information retrieval system based on a computer model of legal knowledge [ 31; and the design of an artificial intelligence system to analyze simple tort cases [lo]. This paper attempts to identify the principal obstacles to the development of a legal consultation system, given the current state of artificial intelligence research, and argues that there are only certain areas of the law which are amenable to such treatment at the present time. The paper then suggests several criteria for selecting the most promising areas of application, and indicates the kinds of results that might be expected, using our current work on the TAXMAN project [q] as an example. I. Poter&i & &nlications. One can imagine numerous applications of artificial intelligence techniques, in several diverse areas of law, but most of these would fall into one of the following categories: (I.) m-al Svstema. There are a number of systems in operation today which maintain data bases of statutes and decided cases, in full text, and which are capable of searching these texts for combinations of key words, using standard information retrieval techniques. (For a comparative survey of the principal domestic systems, LEXIS and WESTLAW, see cf21.1 These retrieval systems have turned out to be useful for certain kinds of legal research tasks, but only when used in conjunction with the traditional manual digests and indices, all of which are organized according to a rigid conceptual