Towards an iStar 2.0 Extension for Analyzing Goal Variability

Sotirios Liaskos, Shakil M. Khan, John Mylopoulos

Abstract: Goal models of the i* family have been shown to be suitable for concisely representing and analyzing goal variability. In standard i*, goal variability emerges due to the presence of OR-refinements, which describe alternative ways by which stakeholder goals can be fulfilled. On closer inspection, however, variability exists in goal models beyond what can be represented by OR-refinements. Firstly, given one set of tasks that fulfill the root goals, there are many alternative orderings by which those tasks can be performed. Secondly, tasks can have many alternative outcomes. In the past, we have proposed various approaches for analyzing these two types of variability, each accompanied by an extension to the i* notation that allows translation of the visual goal model into a formal specification suitable for automated identification of goal-fulfilling alternatives. In this chapter, we propose a consolidation of these extensions into iStar T, a unified extension to the current iStar 2.0 modeling language. iStar T proposes a minimum set of added modeling constructs and a core set of rules that allow translation of models into specifications usable by different reasoning systems. We describe the extensions, and, as examples, we sketch the rules to translate into Hierarchical Task Network and Golog specifications and the kinds of automated analysis that can be done with the result.