In the past few years, the community that develops i* has become aware of the problem of having so many variants, since it makes it difficult for newcomers to learn how to use the language and even to experts to efficiently exchange knowledge and disseminate their proposals. Moreover, this problem also delays the transfer of the i* framework to industrial settings. Our work is one of the current attempts to promote interoperability among the existing variants, and it does that by investigating the semantics behind the i* core concepts. For that, we apply a foundational ontology named UFO, which is used as a semantically coherent reference model to which the language should be isomorphic. In this paper, we report on the steps we have pursued, what we have accomplished so far, also setting the context for the work ahead.
Nowadays, the community that develops i* is relatively big and these developers,
who are geographically dispersed, tend to ascribe different (and sometimes
conflicting) meanings to its constructs. It is argued that this flexibility is part of the
framework’s own nature, and in fact may be considered one of its key success
features. But on the other hand, it is our belief that this represents a clear risk in terms
of promoting the framework, creating serious issues, such as: a) hampering the
efficient communication of knowledge among experts of the community [
In the past few years, the community has become aware of this problem and
several attempts have been made for facilitating the access and uniform use of the i*
language. One of these initiatives is the creation of a common repository and
collaborative environment, namely the i* wiki
(http://istar.rwth-aachen.de/tikiview_articles.php). In particular, there is a session in the wiki called Guidelines,
aimed at collecting and making explicit the different approaches to the language.
Works on metamodeling have also tried to make it clear the meaning ascribed to the
distinct constructs [
Going beyond syntactic issues, since 2006, we are involved in an attempt to define
a common ontology for the core concepts of the i* language. We believe this may
assist in clarifying the semantics of the language’s concepts, thus generating a number
of modeling guidelines targeted at enhancing the language’s usability. We do this by
applying the UFO foundational ontology as a reference model and our approach
prescribes that the i* metamodel reflects such model [
This paper reports on what we have accomplished so far and also presents what lies ahead of us. For that, the remainder of the paper is structured in three sections besides this introduction: section 2 presents the objectives and methodology of our research, also pointing at which stage we currently are; section 3 describes the main contributions and challenges involved in this initiative; section 4 discusses some conclusions and presents our future research agenda.
The general objective of our long-term joint research is to provide an ontological foundation to the i* language core. In the current research stage, we are clarifying the semantics of i* intentional element links: means-end, decomposition and contribution. In this context, it is necessary not only to determine accurately the meaning of these constructs, but also to provide methodological advice on their use, since in some cases differences may be a bit subtle. To sum up, the general objective of understanding the meaning of i* intentional element links can be refined into a series of more concrete research questions: 1) For every type of intentional element link: Which are the logical conditions and implications of a link established from an intentional element to another? 2) In particular, which i* intentional elements may appear as root and target of such an intentional element link? 3) Which ontological properties must these elements fulfil? 4) Which are the methodological guidelines that drive to the application of a particular type of intentional element link? 5) What is the modeler’s agreement on the aspects above? The rest of the paper reports the current results obtained so far and outlines the most immediate future work in relation to these research questions.
Our work has started with the semantics analysis of the core i* intentional elements,
such as actor, goal, task and resource1. The results of this analysis are reported in [
The i* literature shows that modelers sometimes use OR-decomposition and
means-end to express the same phenomenon. This is also reflected in some dialects.
For instance, GRL does not allow OR-decomposition, thus only means-end links are
applied [
In our work, we see the means-end link and OR-decomposition as two different relations. For example, a goal is as mentioned, a proposition, thus it is not possible to decompose a goal into tasks or resources. A goal may be only decomposed into subgoals. The means-end link, on the other hand, is generally applied between tasks and goals, for example. But how can we formally define the means-end link? To understand that, one must first review the ontological meaning of intention. Intentions are mental states of Agents which refer to (are about) certain Situations in reality (i.e. states of affairs the agent aims at achieving). Now, we may define the notion of deliberately achieving a goal as follows: task(a) goal(G) deliberately-achieves(a, G) achieves(a, G) (i: intention(i) is-reason-for(i, a) implies(propositionalcontent(i), G)) 1 From now on, we use different font to highlight i* and UFO elements. For i* elements, we use bold and italic arial while for UFO elements, we apply courier new
In other words, a task (action in UFO) a deliberately achieves a goal G iff this task achieves G (i.e., causes the world to be in a state which makes G true) but also this task must be motivated by an intention (intention i is the reason for task a) whose propositional content implies G. Informally, we can state that a is performed with the intention of achieving G. Now, we are able to define the means-end link as:
task(a) goal(G) ME(a, G) deliberately-achieves(a, G)
In other words, a means-end link between a task a and a goal g holds if the execution of such task leads to goal achievement and is deliberately performed by the agent in whose perspective the relation is defined. Please note that this definition makes a clear distinction between the means-end link and the OR-decomposition relation, previously defined.
Another common case of construct overload in i* is given by the indistinct use of
the means-end and contribution links. For instance, let us make us focus on a make
contribution example (Fig.1).
In it, a Car Passenger agent executes the take a car sick pill task in order to
prevent himself from being sick during the journey he is making (Means-end link to
car sickness prevented goal). As a side effect of this medication, the Car
Passenger also goes to sleep (Make Contribution link to asleep fallen goal). Let us
now carefully analyze this example. Both goals depicted in the model are equally
accomplished: following the proposal in [
As result of the mapping from i* tasks into UFO actions, every task is associated with a causing intention whose propositional content is a goal. In other words, we execute a particular task in order to accomplish a specific goal. In i*, the association between the task and the goal in this case is made by a means-end link (e.g. take a car sick pill task as means to car sickness prevented goal). On the other hand, this same task can also generate some other goals to be accomplished, without however, being intended by the choice of this particular task. In this case, a make contribution link is established (e.g. take a car sick pill task as means to asleep fallen goal). By using the notion of deliberately-achieves previously defined, we may also make clear the distinction between the means-end link and the make contribution link (MakeCont).
action(a) goal(G) MakeCont(a, G)
achieves(a, G) deliberately-achieves(a, G) As can be noted by the above definition in comparison with the means-end definition described a few paragraphs before, the only distinction between the two links is given by the causing intention. Table 2 summarizes some guidelines resulting from the analyses presented in this subsection for the use of the OR-decomposition relation, means-end link and make contribution link.
For reasons of lack of space, we refrain ourselves from presenting the results of the analysis of the break, help and hurt contribution links.
In this paper, we have presented the long-term objectives and main results of our effort to provide an ontological foundation to the i* language core. Our work so far has been conceptual, providing a clear formalizations to ensure that the semantics of the i* constructs can be well understood and properly used. This work is not finished because i* is a powerful language allowing the expression of actor’s theories about effects, side-effects and obstacles of actions. However, to move forward, we now need to do empirical work to validate the usability of our ontology, measured in for example ease of learning and effort to use, and to validate the utility of the ontology for particular stakeholder purposes.
This work has been partially funded by the ProS-Req Spanish project (ref. TIN201019130-C02-00). We wish to thank the reviewers for their very constructive comments. We are also grateful to the support provided by FAPES (PRONEX #52272362/2010) and CNPq Productivity Grant #311578/2011-0).