Fostering the Adoption of i by Practitioners: Some Challenges and Research Directions - Intentional Perspectives on Information Systems Engineering - page 181

soft goals. General knowledge of ontologies in the agent-oriented field [ 33] and
proposals of representation of ontological concepts into the i ∗ framework [ 24] are
worth to explore. On the other hand, the methods shall use a consistent grammar
to name the intentional elements that compose the i ∗ models.

4 Challenge 3: Providing Structuring Mechanisms in i ∗

Context. Modelling is a stepwise process. Some key elements are identified to
build the starting model, and then a series of refinement steps gradually trans-
forms this model into more concrete ones. In the i ∗ framework, the initial key
elements usually are some designated actors (possibly with the main goal that they
fulfil) and the most important dependencies among them. Refinement steps may
yield to new actors and dependencies, and also to the gradual construction of the
Strategic Rationale (SR) i ∗ diagram of each actor by decomposing their main goal
using means-end and task-decomposition links, and establishing the contributions to
softgoals.
Problem. This refinement process has not a clear counterpart in the i ∗ framework.
The only structuring mechanism that i ∗ presents is the concept of actor boundary,
that allows separating the declaration of existence of an actor from the rationale that
it encloses. But the other refinement steps mentioned above are not supported by the
language. Therefore, the final i ∗ model suffers from several problems:

• Difficult to reuse. If a model with some similarities has to be build in the future,

reusability is basically copy and paste the designated elements, which is difficult
and semantically poor. For instance, if a subpart of an SR diagram is a candidate
to be reused, what happens to those dependencies that stem from its intentional
elements?

• Difficult to trace . Since the model does not keep the stepwise refinement his-

tory, the reader is not able to know which elements were introduced in which
stages and why. For an intentional framework like i ∗ is, this is even a more severe
drawback because it hides some rationale.

• Difficult to understand . Since the model is a monolithic unit except for actor

boundaries, the reader has more difficulties than ever to comprehend the full
meaning of the system modelled.

Challenge. The i ∗ language shall include structuring mechanisms for represent-
ing the most usual stepwise refinement operations when developing i ∗ models.
State of the art. There are some lines of research addressing the structurability
issue. The two most ambitious contributions at this respect are the incorporation of
aspects and services into i ∗ .
Alencar et al. [ 3] propose the use of aspects for modelling cross-cutting con-
cerns (see Fig. 3, left). Separation of concerns provides structure to the i ∗ models,
but it does not align with the stepwise refinement process as presented above: