The Reassessment of Preferences of Non-Functional Requirements for Better Informed Decision-making in Self-Adaptation

Abstract

Decision-making requires the quantification and trade-off of multiple non-functional requirements (NFRs) and the analysis of costs and benefits between alternative solutions. Different techniques have been used to specify utility preferences for NFRs and decision-making strategies of self-adaptive systems (SAS). These preferences are defined during design-time. It is well known that correctly identifying the weight of the NFRs is a major difficulty. In this paper we present initial results of a novel approach that provides a set of criteria to re-assess NFRs preferences given new evidence found at runtime using dynamic decision networks (DDNs). The approach use both conditional probabilities provided by DDNs and the concept of Bayesian surprise. The results show that our approach supports better informed decisions under uncertainty by identifying new situations where the current SAS preferences may need to be re-evaluated to improve the levels of satisfaction of NFRs.