What can work organization learn from fault tolerance?

In this paper a component-based n-version approach for organizing work processes is discussed. It is demonstrated that in general such a modular construction offers a potential to considerably increase the reliability of the work process. Moreover, we describe an optimization approach to further increase reliability by selecting the most adequate candidate components while at the same time observing given constraints. A heuristic search algorithm to find optimal system configurations is sketched, and a modification of this algorithm is discussed. Finally we show how for different parts of the versions of the system, varying levels of reliance can be memorized, according to their monitored behaviour. This is achieved by the introduction of adaptively changeable weight factors for the outputs of the work versions.