| For artificial intelligence, understanding the forms of human reasoning is a central goal. Legal reasoning is a form that makes a new set of demands on artificial intelligence methods. Most importantly, a computer program that reasons about legal problems must be able to distinguish between questions it is competent to answer and questions that human lawyers could seriously argue either way. In addition, a program for analyzing legal problems should be able to use both general legal rules and decisions in past cases; and it should be able to work with technical concepts that are only partly defined and subject to shifts of meaning. Each of these requirements has wider applications in artificial intelligence, beyond the legal domain.;The dissertation also describes an implemented program based on the framework. For definiteness, the program is given the task of analyzing law examination problems concerning the formation of contracts by offer and acceptance. The program performs adequately on a narrow class of such problems and appears to be extensible to other areas. Thus, the research opens up a line of investigation that may prove fruitful for both artifical intelligence and legal philosophy.;This dissertation presents a computational framework for legal reasoning, within which such requirements can be accommodated. The development of the framework draws significantly on the philosophy of law, in which the elucidation of legal reasoning is an important topic. A key element of the framework is the legal distinction between hard cases and clear cases. In legal writing, this distinction has been taken for granted more often than it has been explored. Here, some initial heuristics are proposed by which a program might make the distinction. |
