Expert system for administrative decision making: evaluating applications
- 1995
- p.263-71
- Summer
The introduction of expert systems to public administration has been viewed with both excitement and scepticism, and there has been considerable discussion and description of potential or actual applications. Much less common are reports that compare the decisions of an expert system with those of a human, making an empirical evaluation of the system possible. Such a comparison is described in this report. In 1990 the School of Public Administration at the University of Victoria began an experiment on the application of expert system technology to an administrative decision-making process. The process chosen for the test was that of evaluating applicants for admission to the School's Master of Public Admin istration program. The experiment was set up so that the expert system's decisions could be compared with those of human decision makers, and those of the human decision makers compared with one another. The resulting system was tested on three groups of applicants, spanning a period of five years. In all three cases the results showed strong levels of agreement with the human decision makers. In fact, the expert system showed as high or higher levels of agreement with human decision makers as the human decision makers did with each other. Across the tests the system displayed three attributes that might be required of a viable operational system. It appears to be reliable, stable across time, and it does not require highly qualified or experienced staff to run it. It also appears to have the potential to treat applicants with greater consistency than might two different human decision makers. - Reproduced
Public administration - Canada Public administration