Applying the Redundancy Concept to Administrative Organizations
本文探讨了在行政组织中应用冗余概念以降低系统失败概率的实践难题，指出与机器系统不同，人类组织的冗余要素难以标准化和实验控制，导致可靠性难以评估和改进。
A decade has now passed since Martin Landau' first alerted administrative scientists to the benefits of redundancy and overlap in administrative organizations. Although the need to maintain redundancies in administrative systems is clear in principle, it is not at all clear how one can design optimal redundant administrative systems in practice. The common reason for introducing redundancies into any system, administrative or inanimate, is to minimize the probability of the system's failure, i.e., its operation outside specified tolerances. It is clear that if redundancies are not introduced, then the more complex a system becomes, the higher the probability that some serious failure shall occur. Indeed, the norms of organizational design do prescribe that increased organizational complexity be counterbalanced by increased organizational redundancies. However, in attempting to apply such a norm a practical optimization problem arises. On the one hand, the tendency to create ever more multi-faceted and, hence, complex organizations, may leave too few resources for maintaining the required redundant capacity. On the other hand, the difficulty in developing measurable tradeoffs between complexity (or riskiness) and redundant capacity may create a situation whereby organizations become complex precisely because they maintain too many redundant features. In contrast to the systematic study of redundancy in communication and machine systems in which the concept of redundancy was first successfully applied, the design and evaluation of organizational redundancy is not a simple matter. As we have stated elsewhere,2 there is a basic difference between redundancy in inanimate systems and redundancy in human systems. Redundant machine components, for example, can be made identical and can be subjected to controlled experiments, hence having a predictable level of reliability. However, redundant elements in organizational systems composed of people can never be made identical and can hardly be subjected to properly controlled experiments, making their reliability hard to estimate and, hence, hard to improve. As a consequence of this difference, redundant machine components are mainly designed in order to provide backup or reserve capacity;