An examination of Dutch research on legal case solving revealed that few law students get systematic instruction or testing in the technique of legal problem solving. The research being conducted at the Department of Computer Science and Law at the University of Amsterdam focuses on identifying the different functions in legal reasoning tasks in order to develop computational models which realize these functions. The major assumptions of the theory are: (1) in artificial legal reasoning, a separation should be made between reasoning about events in the world and reasoning about legal consequences; and (2) the "real" legal reasoning (when no reasoning about the world is conceived) should be viewed as a process of rule application and conflict resolution, rather than drawing logical inferences. The consequences of these assumptions for knowledge representation are, that in representing regulation knowledge, knowledge about the actions, agents, and objects should be separated from the representation of the regulation. In an intelligent tutoring system (ITS), the system and the student perform the task simultaneously; the way the system reasons has to be functional for educational purposes. The domain under study is administrative procedural law. The main educational goal of a student solving a case is to learn to handle the theoretical concepts in a specific field of law, to learn to find and apply the specific body of norms, to learn to plan the courses of action and to learn to construct a solution which is legally correct. A conceptual model of assessment, a typical task in the domain of law, is under development as part of the library of interpretation models; an interpretation model is an abstract conceptual model of a set of problem solving methods in terms of inference steps. The model can be used in the process of acquiring knowledge for building artificial legal problem solvers. (AEF)