Legal reasoning by analogy with neural networks: The Value Prediction Network

This paper introduces the Value Prediction Network (VPN), a neural network model designed to perform legal reasoning by analogy. The VPN learns to predict the outcome of legal cases based on their facts and legal rules, enabling it to identify relevant precedents and reason about new cases by comparing them to past decisions. It essentially automates the process of finding similar past cases to predict the outcome of a current legal scenario, a core component of legal practice.