Using empirical studies to mitigate symbol overload in iStar extensions

Modelling languages are frequently extended to include new constructs to be used together with the original syntax. New constructs may be proposed by adding textual information, such as UML stereotypes, or by creating new graphical representations. Thus, these new symbols need to be expressive and proposed in a careful way to increase the extension’s adoption. A method to create symbols for the original constructs of a modelling language was proposed and has been used to create the symbols when a new modelling language is designed. We argue this method can be used to recommend new symbols for the extension’s constructs. However, it is necessary to make some adjustments since the new symbols will be used with the existing constructs of the modelling language original syntax. In this paper, we analyse the usage of this adapted method to propose symbols to mitigate the occurrence of overloaded symbols in the existing iStar extensions. We analysed the existing iStar extensions in an SLR and identified the occurrence of symbol overload among the existing constructs. We identified a set of fifteen overloaded symbols in existing iStar extensions. We used these concepts with symbol overload in a multi-stage experiment that involved users in the visual notation design process. The study involved 262 participants, and its results revealed that most of the new graphical representations were better than those proposed by the extensions, with regard to semantic transparency. Thus, the new representations can be used to mitigate this kind of conflict in iStar extensions. Our results suggest that next extension efforts should consider user-generated notation design techniques in order to increase the semantic transparency.