The Design of an Expert System for Domain Knowledge Engineering and Decision Making: A Case Study in the Criminal Justice System

Abstract

In this paper, we discuss a tool for eliciting domain knowledge (specification) of a decision support system. In particular, we focus on a decision support software system (DSS) which employs domain knowledge of recidivism in the juvenile justice system. Using the elicited domain knowledge, the DSS tool uses deductive reasoning techniques to make inferences and provide suggestive courses of action to support the investigatory functions of police, attorneys, or probation officials. The motivation for developing the system is manifold: (1) the activities of the officials are repetitive and their procedures mostly manual; (2) investigations usually result in large volume of biographical data; (3) the need to link several, related case files; (4) officials seldom have concurrent access to case files — causing delays in resolving cases in the court system; among others. Developing a software system to support the investigation and decision making of criminal cases is in itself a daunting task, which makes the system specification a critical input to the development process. Hence, the correctness of the resultant domain knowledgebase and the underlying deductive/support system depends on logically consistent and sound methods. In the paper, we describe the rationale for developing the DSS system, why we focus on the criminal (juvenile) justice system, the methodology for eliciting DSS domain knowledge, and a scenario of what we are implementing as a proof-of-concept system. A series of elicitation sessions which epitomize the DSS system are discussed in the article.