Predicting Outcomes and Identifying Patterns in Law

Student: Benjamin Johnston

Supervisor: Guido Governatori

Category: Information Technology Honours Project

The market for intelligent legal information systems remains relatively untapped, but it is a domain for which information systems could offer dramatic cost reductions whilst simultaneously improving accuracy. Even though there does not appear to be any technical or ethical reasons that might make the development of such systems infeasible, none of the recorded attempts in this domain have reported overwhelming successes. An analysis suggests that the fundamental failure of these prior attempts is a poor balance between powerful knowledge representations and efficient methods of encoding knowledge - a problem that can be solved with defeasible logic.

Defeasible logic is a non-monotonic logic with proven successes in representing legal knowledge that overcomes many of the deficiencies of prior efforts. Unfortunately, an immediate application of technology to the challenges in this domain is an expensive and computationally intractable problem. So, in light of the potential benefits, we have developed a practical algorithm that uses heuristics to discover approximate solutions to the challenges. The algorithm integrates defeasible logic into a decision support system by automatically deriving its knowledge from databases of precedents. Experiments with the new algorithm are very promising - delivering results comparable to and exceeding other approaches.

Integrating this work into a meaningful operational context is a difficult challenge, but that can be solved using semi-structured data formalisms such as XML. Some work has been done in this area, but much of the routine implementation remains. However, the future of this work is extremely promising and worthwhile of further research and development.

Integrating this work into a meaningful operational context is a difficult challenge, but that can be solved using semi-structured data formalisms such as XML. Some work has been done in this area, but much of the routine implementation remains. However, the future of this work is extremely promising and worthwhile of further research and development.

Thesis Document (PDF)