An AI Approach to RCT Experiments on Police Focused Deterrence

• In Event: Integrating Artificial Intelligence (AI) and Randomized Controlled Trials (RCTs) in Police Work

Abstract

This study develops CARESim (an integrated agent-based environment for crime analysis and risk evaluation) and uses it to simulate focused deterrence strategies in policing and conduct randomized controlled trials (RCTs). CARESim encompasses an agent-based conceptual model of behaviors for risk estimation, a modeling background using GIS for agents’ mobility and interactions, a data management module, and a user interface for high-risk individuals (HRIs). Moreover, this simulation environment is transparent, data-driven, and promotes accountability. By creating a synthetic population in a jurisdiction of Virginia and modeling their daily activities related to crime and victimization, this study examines the effects of various focused deterrence strategies including place-based and person-based practices. RCT experiments reveal that the experimental group using focused deterrence has fewer crimes and more arrests than the control group. The findings suggest that CARESim is a trustworthy tool for finding the best strategies to improve police effectiveness. Policy implications and future research directions are also discussed.

Author

• Mengyan Dai, Old Dominion University