Robustness in managing risk uncertainty

Assessing the Robustness of Methods to Better Manage Risk Uncertainty Wouldn’t you agree that once-in-a-million events occur too often for comfort? This stems from the fact that our models of risk assessment have limitations, especially when they deal with low-probability, high-impact events such as the Fukushima Nuclear Power Plant accident.
Dr. Anders Sandberg thinks our brain’s biology is partially at fault here. We have not evolved to be good at this kind of analysis, so our software is buggy, our plans imperfect and our risk models regularly deceived by cataclysmic chains of events. In many cases, there is a far greater likelihood of the model being wrong than of the risk occurring. In order to better deal with this sobering fact and help decision- and policy-makers use more robust risk scenarios, Dr. Sandberg is exploring how to construct “fault-tolerant” reasoning: models with built-in redundancy, using competing groups to improve robustness and reduce the impact of our biases. This may not stop the once-in-a-million events from happening, but at least we will be more ready for them.
My research focuses on improving methods of analysing risks under conditions of data scarcity and model uncertainty. In particular, to improving the robustness of reasoning about risks when model, calculation and argument errors are likely. Due to human cognitive biases and limitations on rationality such reasoning is often erroneous, leading to bad decisions. However, methods from reliability theory and ‘red team’ security gaming might be able to improve the robustness of risk assessment and reduce cognitive biases.