An expert system designed to help the forces of law and order to track down serial killers has been developed by an artificial intelligence researcher at the Rensselaer Polytechnic Institute in Troy, New York, reports Microbytes Daily. What’s unique about the system, called Smart for Serial Murder Analysis and Recognition Techniques, however, is that it approaches the murder from the viewpoint of the killer rather than the detective. The original goal of the project, says Smart (in more words than one) developer Paul Gutwald, was to determine if the concept of rapid prototyping can be used as a viable supplement to the knowledge-acquisition process, a process that is generally regarded as the most difficult part of expert-system development because it requires extensive observation and inquiry. Rapid prototyping, however, basically involves the modelling of a targeted system – the murderer in the case of Smart – and uses the terminology used by the expert – the detective. The key to the success of Smart, Gutwald told Microbytes reporter Jonathan Erickson, is that it uses the same terminology as the expert. This means that the expert can easily understand, modify, and identify changes that need to be made to the system during the development cycle. To assess whether or not rapid prototyping was an effective knowledge-acquisition method, Gutwald chose one of the most difficult areas of research he could find – serial murders. Wicked problem What made this subject so difficult, he explained, was that there are few experts and a relatively small body of knowledge surrounding the subject. In traditional artificial intelligence terms, he said, this is a wicked problem, one that was thought not be suited for expert systems attention – serial murder is classically defined as a killing that occurs when one or more individuals, almost always a male, commits a series of murders that have little or no connection with each other in terms of geographic location, events, or time and without a known motive. From the viewpoint of cost and portability, Gutwald decided to develop the Smart system around a Personal Computer system using Borland International’s Turbo Prolog. Although he said that Turbo Prolog does not offer full Prolog implementation – there were several things, like typing of predicates, that we had to work around – Gutwald credits the program, in part, with enabling him to complete the project in about three months. The biggest advantage that Turbo Prolog provided, he explained, is that we were able to focus on the knowledge, not in the tool itself. Gutwald’s expert was Steven Egger, a New York policeman who is one of the few serial killer experts in the US, and who had completed an advanced degree dissertation on the subject. Although Egger had little, if any, computer experience, Gutwald says that Egger realised that serial murders are a very difficult problem, that he knew there was a solution, and he understood what computers can do. The result of their research is Smart, a menu-driven, user-friendly program that Gutwald expects to see in use at police departments around the US. In general, Gutwald believes that his original question, that is determining whether or not rapid prototyping is an effect knowledge acquisition aid, was answered. Smart shows that rapid prototyping is a viable design strategy for poorly defined problems, he says, and it should enable expert systems to expand to new areas.
