Managing requirements@run.time with a linguistic decision making approach

In the seminal work about requirements, Zave and Jackson established that if specification models hold the correctness criteria then, they can be used instead of requirements to make decisions (e.g. select an architectural configuration to implement requirements). Unfortunately, during runtime for systems under changing environments, domain assumptions may change and if they are not properly maintained (synchronized) the correctness criteria may becomes not useful to detect when the specification model is not anymore a valid representation of requirements. Thus, requirements may be violated but not properly detected. In order to avoid specification models become obsolete during runtime, we already proposed reify requirements into abstract specification models. In this paper we extend the correctness criteria to requirements@run.time and we propose specifically the linguistic decision making (LDM) models to represent these abstract models. We present an illustrative example of how our approach works. The main contribution of this approach is obtained during runtime, when the false negative rate error on determining when requirements are violated is reduced.