On several levels, humans depend on the functioning of complex networks, including the molecular networks in cells, ecological food webs and technical infrastructure nets. In such systems the complex interplay of distributed components can result in cascading failures.
Large complex networks are likely to be dynamically unstable and recent work showed that trying to stabilize the network can lead to large-scale failure. In this talk I will investigate both the propensity of networked systems to fail and the scale of the failure once it occurs.
I will show mathematically that naturally occurring networks have predominantly localised instabilities that are confined to small parts of the network, but attempts to stabilize the network can lead to a delocalisation, such that instabilities become less likely but affect a larger number of nodes.
These results shed new light on the apparent stability of ecological food webs and the causes of systemic failure in artificial technical and organizational networks. Furthermore they have implications for the planning of interventions, such as vaccination campaigns against epidemic diseases.