Gardner feels that the majority of existing legal expert systems are inadequate because these systems do not distinguish between the solutions experts agree upon and the solutions they disagree upon. She believes that such ambiguities arise from the open texture of legal predicates. To overcome this problem, a distinction between “clear” and “hard” cases is implemented in an AI program. This program identifies the issues raised in a fact situation dealing with the offer and acceptance problem (contract law). The program distinguishes those situations in which there is enough information to solve the problem from those in which there is not. She states that her system has produced a description of legal reasoning that is applicable to appropriate subproblems.

The author’s program reasons from given facts and distinguishes between those cases it can solve and those it cannot. To distinguish clear and hard cases, she devises sophisticated heuristics. She correctly argues that the problem arises not so much from ambiguous rules as from the difficulty of applying rules to particular situations.

Whether or not the author’s heuristics distinguishing clear and hard cases would contribute to an understanding of legal reasoning is uncertain. While she has successfully solved the computational problem, it is questionable whether she is entitled to claim success in solving a problem that is the subject of debate by leading authorities on jurisprudence.

Gardner has made a courageous attempt to solve a significant problem basic to understanding the legal reasoning process. However, her main finding has limited application to other AI research in law because it lacks theoretical validity.

The human interface becomes the central issue here. Gardner is skeptical of user input. “The user may not know, and may wrongly assume that he or she does know” (p. 189).

An alternative approach to the clear and hard case distinction would shift the problem’s solution from the program to the user. Such an approach assumes that legal expert systems are designed for use only by lawyers (experts) and not by technicians. The human decision maker would specify the relationship between facts and rules through an interface. Although two experts could specify different fact-rule relationships and reach opposite results, this should cause no concern: the legal system itself is adversarial.