Generative AI in Selected Criminal Justice Applications: Advantages and Limitations

• In Event: Innovative Applications of Computational Methods

Abstract

Criminal justice agencies increasingly use GenAI based applications to improve access to justice services, obtain information needed in criminal investigations and facilitate legal work, thereby improving agency efficiency and effectiveness. Despite noted successes, however, the underlying Large Language Model technology and its variations produce results that are statistical in nature and can raise issues of explainability, data privacy and provenance, bias and hallucination, production of results contrary to known facts. Here we examine effectiveness and trustworthess of GenAI based systems in public facing chatbots, legal services and criminal investigations involving large amounts of data. Based on meta-analysis of existing studies as well as interviews with vendors and law enforcement agencies, we report on recent deployments in each of the aforementioned areas. Analysis indicates that appropriate deployment and informed user guidelines are critical to ensure benefits and mitigate inherent GenAI issues.

Author

• Douglas Salane, John Jay College of Criminal Justice, CUNY