Analysis and Evaluation of Decision Aids in Organizations

Abstract

A methodology for the modeling and evaluation of decision aids is presented. Petri Net models of human decisionmakers and decision aids are described that can be interconnected to represent organizations. The approach is illustrated by modeling and analyzing the effects of a class of decision aids that is based on artificial intelligence and especially expert systems. This paper presents a procedure for assessing to what extent the measures of performance of an organization are modified when an expert system is introduced. First, a model of symbolic computation with fuzzy logic, using Predicate Transition Nets, is presented to model the most common kind of expert systems: the consultant expert systems. This model allows to evaluate its response time for a given input. An example problem involving a hierarchical two decisionmaker organization, where the expert system is used as an aid in the fusion of inconsistent information, is then presented. A strategy involving the use of the expert system is compared to two other strategies expected to be used by a decisionmaker responsible for this task. Measures of performance (workload, timeliness and accuracy) are evaluated for each of these strategies. The results show that the strategy involving the use of the expert system improves significantly the accuracy of the organization, but requires more time and increases the workload of the decisionmaker using it.