This study assesses game theory’s potential to contribute to understanding the North Korean nuclear crisis. Previous APL work suggests that game theory provides a useful framework for formulating and analyzing multilateral nuclear stability issues; however, it seldom provides unique policy-relevant insights. For it to do so, game theorists must work closely with policy and subject matter experts. Thus, APL invited game theorists and nuclear and regional experts to (1) discuss the strengths and limitations of game theory applied to the North Korean nuclear crisis; (2) address the policy community’s skepticism about game theoretic analyses; and (3) explore mechanisms for collaboration. We found that game theory, when correctly applied, is a rigorous framework for understanding interactions in international conflicts; however, its predictive capability is limited. Misguided expectations have led to skepticism about its utility. While collaboration between these communities might increase understanding of nuclear crisis dynamics, obstacles include resolving motivational and communication issues, regulating expectations, and avoiding improper applications of game theory.
APL has published the event summary for the inaugural National Health Symposium, which brought together more than 160 experts from government, academia and industry to discuss ways that advances in research and development can translate into better delivery of health care.
National Security Report
We establish that the US system for nuclear deterrence is a complex system in the formal sense, that nuclear deterrence must be regarded as a system-level function, and that the consequence of this is that there is the possibility of system-level failures not obviously connected to any component failures. These are emergent properties not predictable from an understanding of each of its components and interactions that may be candidates for Taleb’s black swan events. To understand the potential risk of failure of the US nuclear deterrence system as it exists in the United States and in the larger context of multiple state actors, it is necessary to understand the potential interactions of components and command authority. For the analyst, this means constructing models that attempt to capture the non-linearities of interactions, the existence of which is increasingly apparent.
National Security Report
The binomial distribution is widely used across many different disciplines. In cases where data can be represented with a binomial distribution, an estimate for the binomial distribution parameter (for the probability of success) is often produced. However, uncertainty surrounding this estimate is only sometimes reported, partly due to the opacity of the various methods available for determining this uncertainty. Failing to appropriately analyze uncertainty can lead to erroneous, or at least incomplete, conclusions. Here, we explore both Bayesian and frequentist methods for quantifying uncertainty in the binomial distribution parameter, and discuss each method's various advantages and limitations. Our work is motivated by nuclear crisis outcome data. While nuclear crises have been studied to determine the likelihood of the nuclear-superior, compared to the nuclear-inferior, state winning in a crisis, there is great uncertainty in these estimates for the probability of winning. We demonstrate methods that appropriately quantify such uncertainty and use nuclear crisis outcome data to illustrate applications of the methods we present, as well as to demonstrate insights that can be provided by explicitly analyzing uncertainty.