ASSYST - computer support for guideline sentencing

• Authors: E. Simon;G. Gaes
• Venue: ICAIL '89 Proceedings of the 2nd international conference on Artificial intelligence and law
• Year: 1989

ASSYST (Applied Sentencing System) was developed so that judges, prosecutors, defense attorneys, probation officers and other key members of the Federal criminal justice system could easily compute, record, archive and examine the implications of Guidelines promulgated by the United States Sentencing Commission [1].Although ASSYST was not developed using the formalisms employed by most expert system designers, it does meet at least some of the criteria of an expert system [2] and in particular an expert system for legal analysis [3]. ASSYST is a domain specific knowledge representation of the rules for determinate sentencing in the Federal judicial system. It elicits from the user all the information required to make a final determinate sentence. Unlike expert systems which resolve decision conflicts, ASSYST will only suggest tentative solution sets to grouping rules required by Sentencing Guidelines and, in this respect, it is a deterministic rather than a conflict resolving system.ASSYST has been a success. In June of 1988, every U.S. Probation main office received a copy of the software. The response has been very favorable. Officers who have taken the time to learn the system have reported substantial decreases in amount of time spent per case. The software has proliferated throughout the Federal criminal justice system since, as a publication by the Federal Government, it is not copyrightable.In a related paper, Sergot et al. [4] have discussed the formalization of the British Nationality Act in Prolog. In their article, Sergot et al. demonstrated how the language of the statute could be translated into definite Horn clauses which form the basic logic structure of Prolog's programming language.The difficulty usually encountered in the logical interpretation of statutes, according to Sergot et al. is that the statute language is often vague. Phrases such as “having reasonable excuse” or “being in good character” are two examples from the British Nationality Act. The Sentencing Reform Act of 1984 [5], the statutory basis for federal sentencing guidelines, also contains many vague phrases. As an example, the enacting legislation requires the U.S. Sentencing Commission to consider the following offender characteristics: mental and emotional condition, community ties, and criminal history. There is no exposition in the statute on any of these factors. It was the responsibility of the Commission and the guideline writers to translate the intent of the legislation into coherent and consistent rules of determinate sentencing. Analogous to Sergot et al., our programming effort attempted to formalize legislative logic, in this case determinate sentencing, into rigorous programming logic. Unlike Sergot et al., our burden was ameliorated by others who transformed the vague language of this statute into the rules of sentencing guidelines.Sentencing guidelines were proposed by Congress to reduce sentencing disparity among federal prisoners and to provide a consistent set of rules to codify characteristics of both the offense and the offender. In the past, judges have had a great deal of latitude in their sentencing decisions. While some would argue that this discretion was necessary in order to treat the individual circumstances of the crime, others have argued that the degree of latitude introduced inconsistent sentencing decisions.This paper does not attempt to resolve the imbroglio between proponents and opponents of sentencing guidelines. Rather it is intended to show how a computer application became an integral component of the continued development of the guideline process.