Anders Sandberg has won an AXA postdoctoral fellowship in the area of socioeconomic risks for his project “Robustness in managing risk uncertainty”.
The project seeks to improve techniques for analysing risk in situations where low-probability but high impact events occur. Such situations are hard to evaluate not only due to data scarcity and model uncertainty, but also due to dominant “argument uncertainty”: the uncertainty inherent in methods of measuring, modelling, and reasoning about risk done by biased and fallible humans. In many cases the probability of arguments or even simple calculations being mistaken are significantly higher than the risk probabilities themselves. The project aims at understanding and reducing uncertainties due to fragile or limited reasoning, following up on earlier work at FHI about low-probability risks and cognitive biases.
The project will investigate methods of effective robust argument composition (how arguments can be made more resilient by combining independent sub-arguments), heuristics for estimating the independence of arguments, “red team” risk assessment (where investigators try to find flaws in proposed solutions) and other adversarial methods for improving robustness, and the application of these methods to global risks and models of societal collapse that are being developed at FHI and in the Oxford Martin School.
Visit AXA Research Fund website: http://www.axa-research.org/en/