An empirical investigation of the use of expert systems by groups

Groups make a large number of decisions in organizations. Incorporation of expert system concepts within GDSS has been suggested in the past as a means of supporting the decision-making activity of groups. This study aims to find what group factors affect the use of expert systems by groups within a decision-making setting. Specifically the study examined group commitment, satisfaction, past success and cohesion as variables of interest. Competing alternative models, incorporating these variables, are proposed to explain the group use of expert systems.

An expert system that supports reasoning by analogy was developed for the business simulation game IMAGINIT. The use of the expert system was found to significantly increase the problem understanding, and the average and ending performance. The expert system was also found to help in the strategic decision-making phase by the groups.

A laboratory study was conducted to measure the voluntary use of this expert system. Results show support for one of the models. In this model each of the variables (commitment, satisfaction, cohesion, and past success) were viewed as independently affecting the use of expert systems. Of the variables investigated, past success was the only variable to significantly affect the use of the expert system. Groups that performed better in the past tended to attribute their performance to the usage of the expert system. This attribution resulted in an increased usage of the expert system. The other variables (commitment, satisfaction, and cohesion) did not significantly affect the use of expert systems.

Because many of the group variables did not significantly affect the usage of expert systems by groups, a data-driven model was developed to derive the relationships among the group variables, and usage of expert systems. The data suggests that the correlation between the variables (commitment, satisfaction, past success, and cohesion) is caused by a common second-order factor (group behavior). This second-order factor did not significantly affect the use of the expert system.