AI-Assisted Court System: System Design and People's Attitude

This paper presents our interdisciplinary collaboration among legal scholars, socio-legal scholar, A.I. scientists, Bayesian statistician to develop an A.I.-assisted court system. In this system, the fact-finding is based upon Bayesian network system, the legal decision is based upon D.N.N. and N.L.P., and the legal reasoning is based upon logic programming. We also conducted several surveys about the attitude of Japanese people toward A.I. judge system. The Bayesian decision theory and hence the Bayesian network are the only well-defined and logically coherent decision model. The problem is that it is difficult for legal professionals to understand and utilize it. It also needs a huge amount of data to deal with various facts and evidence. The Deep Neural Network is the most promising model for the natural language processing, as is shown in the development of ChatGPT-4o. The problem is it needs a huge amount of data since it is based upon the large language model. The legal reasoning means the inference with several legal norms to reach the conclusion, i.e., the judgment. The logic programing is suitable for the legal reasoning since it is logically perfect (see the logic programing language "prolog," which is based upon the first order predicate logic). The problem is only human experts can, so far, translate legal rules and precedents into logical rules. People are very reluctant to use court in Japan (and also any countries). We conducted several surveys to find out whether and to what extent people expect and/or worry the utilization of A.I. court system.