Computers have long been utilized in the sphere of law in the form of word processors, spreadsheets, legal research systems, and practice management systems. Most exciting, however, has been the prospect of using artificial intelligence techniques to create so-called legal reasoning systems—computer programs that can help to resolve legal disputes by reasoning from and applying the law. But the practical benefits of such automated reasoning systems have fallen short of optimistic early predictions and have not resulted in computer systems that can independently provide expert advice about substantive law. This is not surprising in light of the difficulty in resolving problems involving the meaning and applicability of rules set out in a legal text.

Early attempts at automated legal reasoning focused on the doctrinal nature of law. They viewed law as a set of rules, and the resulting computer systems were engineered to make legal decisions by determining the consequences that followed when its stored set of legal rules was applied to a collection of evidentiary data. Such systems underestimated the problems of interpretation that can arise at every stage of a legal argument. Examples abound of situations that are open to differing interpretations: whether a mobile home in a trailer park is a house or a motor vehicle, whether a couple can be regarded as married in the absence of a formal legal ceremony, and so on. Indeed, many notions invoked in the text of a statute may be deliberately left undefined so as to allow the law to be adapted to unforeseen circumstances. But in order to be able to apply legal rules to novel situations, systems have to be equipped with a kind of comprehensive knowledge of the world that is far beyond their capabilities at present or in the foreseeable future.

Proponents of legal reasoning systems now argue that accommodating reference to, and reasoning from, cases improves the chances of producing a successful system. By focusing on the practice of reasoning from precedents, researchers have designed systems called case-based reasoners, which store individual example cases in their knowledge bases. In contrast to a system that models legal knowledge based on a set of rules, a case-based reasoner, when given a concrete problem, manipulates the cases in its knowledge base to reach a conclusion based on a similar case. Unfortunately, in the case-based systems currently in development, the criteria for similarity among cases are system dependent and fixed by the designer, so that similarity is found only by testing for the presence or absence of predefined factors. This simply postpones the apparently intractable problem of developing a system that can discover for itself the factors that make cases similar in relevant ways.