Unlocking the Policy Laboratory with AI: A Comprehensive Database of Criminal Justice Policies in U.S. States

• In Event: At the Frontier: Exploring Applications of Large Language Models in Crime Research and Public Policy

Abstract

Despite decades of research and numerous policy initiatives, empirical evidence on the causal effects of criminal justice reforms remains limited to a narrow subset of policies, leading to many broad gaps in knowledge. These gaps are not a result of insufficient effort, but rather the inherent challenges of studying diverse policy implementations across various times and locations. The federalized U.S. states, often referred to as "laboratories of democracy," provide a unique and mostly untapped opportunity to address these challenges. However, fully harnessing this potential has been challenging, as policies are encapsulated in collections of unstructured text that resist conventional statistical analysis. This project aims to transform our ability to evaluate criminal justice reforms by creating a comprehensive, AI-driven database of state-level policies spanning several decades. The database will leverage advancements in large language models to analyze and classify legal texts, enabling systematic analysis of policy evolution with human oversight. Beyond evaluating existing policies, this database will help identify promising patterns and contexts where specific reforms have succeeded, allowing researchers and policymakers to make more informed decisions about which interventions merit further investigation and investment.

Authors

• Elliott Ash, ETH Zurich

• Zubin Jelveh, University of Maryland, College Park