Legal knowledge-based systems: new directions in system design

Abstract

This thesis examines and critiques the concept of 'legal knowledge-based’ systems. Work on legal knowledge-based systems is dominated by work in 'artificial intelligence and law’. It seeks to automate the application of law and to automate the solution of legal problems. Automation however, has proved elusive. In contrast to such automation, this thesis proposes the creation of legal knowledge-based systems based on the concept of augmentation of legal work. Focusing on systems that augment legal work opens new possibilities for system creation and use. To inform how systems might augment legal work, this thesis examines philosophy, psychology and legal theory for information they provide on how processes of legal reasoning operate. It is argued that, in contrast to conceptions of law adopted in artificial intelligence and law, 'sensemaking' provides a useful perspective with which to create systems. It is argued that visualisation, and particularly diagrams, are an important and under considered element of reasoning and that producing systems that support diagramming of processes of legal reasoning would provide useful support for legal work. This thesis reviews techniques for diagramming aspects of sensemaking. In particular this thesis examines standard methods for diagramming arguments and methods for diagramming reasoning. These techniques are applied in the diagramming of legal judgments. A review is conducted of systems that have been constructed to support the construction of diagrams of argument and reasoning. Drawing upon these examinations, this thesis highlights the necessity of appropriate representations for supporting reasoning. The literature examining diagramming for reasoning support provides little discussion of appropriate representations. This thesis examines theories of representation for insight they can provide into the design of appropriate representations. It is concluded that while the theories of representation that are examined do not determine what amounts to a good representation, guidelines for the design and choice of representations can be distilled. These guidelines cannot map the class of legal knowledge-based systems that augment legal sensemaking, they can however, be used to explore this class and to inform construction of systems.