Including Defeaters in Quantitative Confidence Assessments for Assurance Cases
Abstract
Assurance Cases (ACs) are prepared to argue that a system will realize desired quality attributes, such as safety or security. It is important to evaluate the confidence in an AC, and several qualitative and quantitative methods exist for this purpose. It is also common to incorporate dialectic elements (i.e., "defeaters") into the AC to mitigate biases. However, existing quantitative methods do not have means to account for defeaters. So, practitioners have to choose between applying a quantitative confidence assessment method or using defeaters. This paper makes two contributions to address this limitation. First, it proposes a theoretical framework, containing rules, for quantitative confidence assessment methods incorporating defeaters. Second, it applies the framework to extend an existing Bayesian Network method to account for defeaters when assessing of belief in the AC's top-level claim. The method is implemented in the Socrates tool and applied to an example.