How incorporating neural networks and symbolic reasoning can improve the quality of legal decision-making?

Abtract With the development of pretrained language models, the legal domain has seen a recent surge of interest in using machine learning to support legal decision-making. However, while these models can provide some insights, they are often limited by their reliance on statistical methods and lack of understanding of the underlying legal reasoning. We propose a framework that combines the strengths of neural networks and symbolic reasoning to improve the quality of legal decision-making. Our framework first uses a neural network to learn the relevant concepts from a training set of cases. The concepts are then represented as symbolic representations, which are used by a symbolic reasoning system to predict the outcomes of the cases. The reasoning system can also provide justifications for its predictions and improve the transparency of the system. Workshop Presentation at Inpretable AI Workshop at JAIST, Nomi, Ishikawa, Japan in September, 2022.