Adversarial Stress Testing of Lifetime Distributions
This work addresses a theoretical problem in probability and decision-making, likely incremental as it adapts stress testing concepts to a new context without demonstrating broad impact.
The paper tackles the problem of stress testing an individual's confidence in a probability distribution by proposing a game-theoretic framework with adversarial and amicable players, using de Finetti-style bets and Kullback-Leibler discrimination to maximize expected utility, but no concrete results or numbers are provided.
In this paper we put forward the viewpoint that the notion of stress testing financial institutions and engineered systems can also be made viable appropos the stress testing an individual's strength of conviction in a probability distribution. The difference is interpretation and perspective. To make our case we consider a game theoretic setup entailing two players, an adversarial C, and an amicable M.The underlying metrics entail a de Finetti style 2 sided bet with asymmetric payoffs as a way to give meaning to lifetime distributions, an adversarial stress testing function, and a maximization of the expected utility of betting scores via the Kullback Liebler discrimination.