Abstract for Elga’s talk

When a system consisting of many interacting parts (such as an electrical power grid or a banking network) starts failing, it is
tempting to make the system more robust by linking its parts together. For example, one might allow an overloaded power generator to “borrow” power from its neighbors, or one might allow banks to heavily insure each other.  Using a generalization of Bryan Skyrms’s notion of resilient probabilities, one can see that such changes tend to destroy (stable) risks associated with system failure, making it easy to get caught off guard by a large cascading failure.  Furthermore, market forces, political incentives, and psychological tendencies all push in the direction making such changes, and hence tend to create systems that are both fragile and *seemingly* robust.  The moral is that the existence of stable risks is itself a public good—a good that is subject to a serious free rider problem, and so is worth protecting through legislation and carefully designed incentive systems.