Managing Complexity Through Selective Decoupling

C. Jason WOODARD, Singapore Management University


Designers of complex systems are often confounded by the tendency for design changes to increase performance on some dimensions while decreasing it on others. While adopting a more modular architecture may temper these opposing effects, modularization can also deprive designers of opportunities to harness complementarities among system elements. This paper explores this tension using an NK model in which product designers can modify the structure of their fitness landscapes by suppressing or restoring interactions between components. We find that these changes can lead to improved performance by flattening harmful interactions that would otherwise cause search efforts to become trapped on local fitness peaks. Despite the fact that this process, which we call selective decoupling, tends to lower the maximum performance of a given component, it may increase the maximum attainable product performance. Moreover, selective decoupling can trigger the restoration of previously suppressed interactions, enabling the discovery of integral designs that were inaccessible to local search. We discuss the similarity between this pattern of endogenous architectural change and the stylized product life cycle described by the technology and innovation management literature, and suggest that the forces highlighted by our model might play an underappreciated role in technology and industry evolution.