Willkommener gast   logon       English, Dansk   
 

Expert System

Introduction, Situation Analysis, Instructions, Technical Background

The Historical Background of Expert System


This model has been developed over a 40-year period. It began when Victor Vroom was a professor at Carnegie Mellon University in the 1960's and continued after his move to Yale University in 1972. The research began by asking people in positions of leadership to describe two decisions which they had made--one of which had been successful and the other, a clear failure. For each decision, these leaders answered questions about the degree to which they involved their teams as well as about other features of the situation in which the decision took place.

Over a thousand decisions were obtained from a wide range of managers working in contexts as different as government, military, NGO's, community organizations, and the private sector. Wide geographical distribution was achieved by obtaining decisions from managers from Asia, Europe, Latin America, and the Middle East. Each case was recorded on a 5 x 7 inch card, and the data coded with respect to information on the success or failure of the decision, the leadership style used, and a set of situational variables which have been suggested as influencing the effects of participation. Early in this research it became apparent that there were no simple relationships between the leadership style used and its success or failure. This appeared to conflict with prescriptions made by social scientists such as Likert (1960), who advocated participative groups as the model for effective leadership without qualification.

In 1972 Victor Vroom published a book, Leadership and Decision Making, with Philip Yetton, then a doctoral student working with him at Carnegie Mellon. This book outlined a preliminary model for incorporating situational factors and leadership styles into a method for matching leadership styles with situation demands. The Vroom-Yetton Model was expressed as a decision tree which became a standard fixture of management textbooks in the ensuing decades. The model on which the decision trees were based grew out of an analysis of the successful and unsuccessful decisions described above and also on laboratory and field studies of participation in decision making, many of which were carried out by Kurt Lewin and his colleagues.

Vroom and Yetton used the terms Model A and Model B to refer to two different decision trees. Model A could be termed "managing for today" since it focused on short-term results by recommending the process which utilized the least amount of time necessary to obtain an effective decision. Model B ignored time in favor of longer-term developmental benefits of participation. In recent revisions of the model, decision trees have been replaced by decision matrices, and Model A has become the Time-Driven Model and Model B, the Development-Driven Model.

In the 1970's Peter Fuss, a director at Bell Labs, who had studied the model in a course taught by Vroom, built a "black box" which duplicated the Vroom-Yetton Model. The box could fit comfortably in one's hand and contained a set of eight switches with labels corresponding to each of the model's situational factors. Once each switch had been set, corresponding to the presence or absence of the factor, a button could be pressed which illuminated one or more of five lights corresponding to feasible alternatives.

The Vroom-Yetton Model elicited varied reactions from the academic community. For example, Miner (1984) concluded that no leadership theory surpassed the Vroom-Yetton Model in its scientific validity and predictive usefulness. In contrast, Field (1979) suggested that the model was too complex to be of much practical value.

During the next decade there were six studies assessing the validity of the model, each conducted by different investigators. The results were summarized by Vroom and Jago (1988). They show that the decisions that are consistent with the model are approximately twice as likely to be successful as those which are inconsistent with the model. In addition to supporting the basic idea of a normative model, these validation studies pointed to ways in which the predictive validity of the model could be substantially increased, albeit by increasing its complexity. Vroom and Jago took on the challenge of building a model which was substantially more complex than its predecessor but was, in fact, easier for managers to use since it was expressed in a computer program. Unlike the black box, the computer had the capacity to take more differentiated inputs, perform more complex operations on these inputs, and produce much more complete and useful information for the user.

One limitation of the Vroom-Yetton Model lay in its treatment of situational factors as dichotomous rather than continuous factors. Expressing situational factors as either present or absent was conducive to the decision-tree format but did not do justice to the fact that each of them could be more accurately thought of as a matter of degree.

Closely linked to the dichotomous situational factors was the concept of rule, integral to the original model. Rules provided the basis for eliminating processes from consideration based on threats to either decision quality or acceptance. Rules were crude devices which have been compared with commandments because of their "thou shalt not!" quality. Rule violations were not matters of degree. Trade-offs between quality and acceptance were not allowed. All alternatives which violated rules were deemed equally unsatisfactory, and those which did not violate were deemed equally effective. Clearly neither conclusion was warranted. The latter has been identified by critics as particularly bothersome. On a typical portfolio of situations, almost 50 percent of the time all alternatives remained in the feasible set. What good is a decision aid which provides no guidance concerning the choice to make?

Clearly enough is known about the underlying processes to know that there are likely to be differences in the outcomes of processes within the feasible set. To take advantage of this knowledge, it was necessary to replace the rules with equations which perform the same functions but which took advantage of the multi-level nature of the situational factors.

The first use of multiple levels of situational factors and of equations for estimating the joint effects of leadership styles and situational factors on decision outcomes was undertaken by Vroom and Jago in their 1988 book, The New Leadership. Twelve years later Vroom built upon advances in computer technology to create Expert System, a computer program contained on a CD-ROM. While in the "spirit" of the black box, the incorporation of graphics enabled one to visualize the effects of the style of leadership on decision quality, implementation, time, and development. It also introduced new situational factors as well as changes in terminology. After more than a decade of experience with Expert System, we now seek to make it more user friendly and more helpful in providing managers with the information they need in making informed leadership choices.

For additional information on the model: