i* is the most popular goal modeling language and, therefore, has several dialects that can interpret its concepts in different ways. iStar 2.0 was designed to lessen the misinterpretation problems of its constructs. Concepts can be well-defined in the language but, if not used properly, the produced models may become inconsistent. An ontology can be used to verify and help the construction of correct and consistent models. GORO is an ontology about GORE that was built based on different goal modeling languages. This paper interprets some iStar 's constructs and proposes some discussions in the light of GORO.
Due to the popularity of iStar , different extensions have been proposed, leading
to different interpretation of its constructs. This led the community to propose
iStar 2.0 [
Hence, when analyzing models produced using iStar , it is possible to identify inconsistencies that are not related to misconstructions of the language, but to improper use of its constructs, due to misunderstandings regarding its constructs’ semantics. Moreover, we argue that a syntactic analysis of the language is not enough. We need a deeper understanding of the constructs with a well founded artifact that provides a common and unique interpretation of concepts: an ontology.
Several interpretations are given to the iStar constructs, which lead to models with multiple meanings and confusion among beginners. This motivated us to analyze the definitions of the language elements in the light of a well-founded ontology.
One of the main focuses of ontologies is to provide a non-ambiguous
conceptualization of a specific domain. As a consequence, modeling languages that
are created with support of an ontology will have a very precise definition of
its constructs. Indeed, giving an ontological interpretation of modeling language
constructs has been recognized as a good practice in modeling language
development [
In this paper, we propose the use of the Goal Oriented Requirements
Ontology (GORO) [
To put it more precisely, in this paper, we provide possible interpretations of some iStar ’s constructs in the light of GORO. By analyzing different interpretations of constructs, we seek to solidify the definitions of concepts in a consensual manner. Thus, this paper intends to raise discussions about some constructs, namely: (i) the Refinement Link; (ii) the Task/Goal refinement interpretation given the different modeling phases to which they belong; and (iii) the Role element definition and its relation with intentionality.
The remainder of the paper is organized as follows: Section 2 presents a brief theoretical background on GORO. Section 3 proposes the ontological interpretation of some iStar ’s constructs based on GORO, also presenting some modeling guidelines based on such interpretation. Finally, Section 4 concludes this paper. 2
Due to the large number of languages proposed since the first GORE approach
emerged [
The Goal Oriented Requirements Ontology (GORO) [
Figure 1 presents an excerpt of GORO. The ontology focuses on the concept
of Goal and defines it as a Mental Moment, which is a mental property
existentially dependent on a single individual. A Goal in GORO can assume four types
of classifications, according to [
A Task is a subtype of Action Universal (Plan), which is a specific way to do something. In this context, a Task describes a process that a stakeholder wants to follow in order to achieve a Goal. A Task can be an Atomic Task or a Complex Task. A Complex Task can be broken down into other Tasks, whereas an Atomic Task cannot. A Task can require and/or produce a Resource.
Assumptions are domain properties expected to be true in a specific
context (the context in which a Goal is expected to be achieved). Assumptions are
classified according to [
We emphasize that the entire process of building GORO and mapping
constructs of GORE languages to GORO (including both versions of i* /iStar) was
carried out with the support of a group of domain specialists, as described in [
This section discusses the interpretation of different concepts of iStar and is divided in three parts. Subsection 3.1 describes the multiple semantics of the Refinement Link, while subsections 3.2 and 3.3 propose open discussion topics about the Task/Goal refinement and the Role element, respectively. We use sans serif and slanted to represent concepts and relations of GORO, respectively, and boldface for iStar constructs. 3.1
With the aim of alleviating the complexity of the language and thus facilitate its adoption, iStar proposes gathering multiple relations between goals and tasks in just one relation called refinement link, which can be of two types: AND or OR. Indeed, this choice simplifies the language’s syntax; however, this can also lead to confusion regarding semantics. In iStar , a goal can be (AND/OR) refined in goals and tasks. Also, tasks can be (AND/OR) refined in goals and tasks. With GORO, we can explore the very nature of all these relations.
The iStar AND refinement between goals is interpreted as GORO’s And Complex GBRA, which represents a complex Goal (an implicit concept derived from Goal-Based Requirement, analogous to And Complex GBRA) decomposed in sub-parts (sub-goals). Being Goal a Proposition, it is possible to separate it in different parts that, together, compose the original (G () G1^G2^G3^: : : Gn). That means, G is satisfied by exactly those situations which satisfy G1 : : : Gn conjunctively and, so, satisfying all subgoals implies satisfying the supergoal.
Moreover, the OR refinement element in iStar can be used to represent GORO’s Or Complex GBRA indicating alternatives, in which Goals are not broken down in sub-parts but in alternative Goals in which a situation that satisfies each of them (separately) also satisfies the main goal. Hence, considering S(G) as a situation that satisfies G: S(G) () S(G1) _ S(G2) _ : : : _ S(Gn).
The iStar refinement link can also be used to refine a goal into tasks. If the goal is OR refined into tasks, then the child task is a particular way for achieving the parent goal. In GORO, these relations can be interpreted as the intends to operationalize relation between a Task and a Goal. This relation denotes that the successful execution of a Task — i.e., the action of executing this specific process — will interfere in the reality, producing a post-situation that may satisfy the Goal. However, if a goal is AND refined into tasks, the child tasks are sub-tasks that must all be executed to achieve the goal. In this case, there is an implicit Complex Task that intends to operationalize the Goal and is OR refined in the child tasks, as explained next.
Regarding task decomposition/refinement, iStar tasks can also be refined in other tasks. When OR refined, the child task represents a specific way to execute the parent task, whereas when AND refined, the child task is a set of steps that need to be performed in order to complete the parent task. In GORO, Tasks can be Complex Tasks or Atomic Tasks. The former can have subparts that, when executed together, achieve the whole Task. The iStar AND refinement of tasks is in agreement with GORO’s concept of Task decomposition, whereas the OR refinement of tasks is not covered in GORO. We do not consider OR refined tasks in GORO as we propose a different approach for this type of modeling: if a task has alternative ways of being achieved, then it is actually a Goal and hence, its subtasks are Tasks that intend to operationalize such Goal. Consequently, a task that is OR refined into other tasks should be modeled as a goal that can be achieved by performing different tasks.
In iStar , it is also possible to refine tasks into goals. This specific type of refinement will be discussed in Subsection 3.2.
iStar refinements have different meanings depending on the elements they connect. Ontologically speaking, grouping different semantics into one syntax leads to construct overload, which may cause these elements to be misinterpreted by modelers. Nevertheless, we here try to minimize the impact of this issue, by providing explicit ontological interpretation (i.e., ontological commitment) of the languages’ constructs. We firmly believe that such interpretation helps clarifying the meaning of its construct, and ultimately lead to better iStar models. 3.2
In iStar , a task can also be refined in goals. GORO does not predict this kind of relation. As GORO defines, Goals are the propositional content of an intention to achieve a specific situation (state of affairs) in reality. This means that Goals refer to desired situations in reality. On the other hand, Tasks describes specific processes executed with some intention.
A possible interpretation is to consider that goals that are children of a task refer to the post-situation brought about in the reality after that task’s execution. However, it is our belief that being goals and tasks of different ontological nature, the trade-off of allowing such refinement is not a positive one, possibly leading to much confusion. After all, how can a process (i.e., a Task be decomposed into propositions (i.e., Goals)?
Yu [
The Role Element
iStar propose actors as active and autonomous entities that aim to achieve their
goals and may be further classified as roles or agents. In GORO, an actor
corresponds to the Requirement Stakeholder Kind concept. A role, in its turn, is
interpreted as a Requirement Stakeholder Role. The distinction between a kind
and a role, according to UFO, is that a kind represents essential properties of
objects (they are also termed rigid or static types [
As stated before, GORO interprets a Goal as the propositional content of an
intention, which is a mental property existentially dependent on a single
individual. Thus, a Goal is related to an individual Requirements Stakeholder. However,
being a type, a role represents expected behavior patterns of several individuals.
To explain that, we resort to the ontological clarification made by Guizzardi et
al. [
Among the various GORE modeling languages, iStar has been one of the most popular ones, further leveraging the GORE field. Therefore, conceptually standardizing all concepts common to this field becomes an increasingly important need. This is the goal of iStar 2.0. This standard can be considerably improved with the support of ontologies, such as GORO, making explicit the semantics behind the language’s constructs. In addition, ontologies can be used to guide users in adopting best practices in iStar modeling, thus preventing modelers from creating inconsistent, low-quality models. We emphasize that an ontological analysis is not intended to limit or restrict the use of a language, but to provide a well-founded and unambiguous interpretation of its concepts.
In this paper, we propose an ontological interpretation for iStar role, we argue that agent is not a necessary concept, and we explain why it is not a good idea to refine tasks into goals. Moreover, we presented different ontological interpretations of the iStar ’s refinement link. We acknowledge that this element groups different semantics in order to reduce the complexity of the language. However, we emphasize the need to find a balance between simplification and loss of meaning. We understand that this is a hard task and that the line that defines a good balance between both sides is thin. Therefore, the use of an ontology to mitigate the risk of having the language’s construct misunderstood and misused may be a suitable approach.
We are currently using GORO to analyze a greater number of languages, also facilitating their interoperability. In addition, GORO can be used to solidify GORE concepts, so that they can be properly used, thus allowing new adepts to learn GORE in a simpler and faster way.