Dependencies among intentional actors is a fundamental feature of i* Modelling. By depending on others, an actor can achieve much beyond what it can by itself. At the same time, the dependent actor becomes vulnerable to the failings of dependees. However, even when dependees are fully fulfilling expectations, over time, depending on other actors can result in structures that are hard to change. By analyzing second-order dependencies, i.e., dependencies among (first-order) dependencies, we determine the extent to which a dependency is depended on by other dependencies. The more a dependency is depended on by other dependencies, the more likely it is to become a barrier to change. Our approach is a model-based formulation of the concept of rigidity in the study of dynamic capabilities in strategic management. The i* models are used to analyze resistance to change in socio technical structures.
The importance of dealing with change and enabling adjustment to changing
requirements has been studied in both management and Information System (IS) design
[
While many researchers in IS and software engineering have attempted to
overcome the first challenge, not many approaches exist that can analyze social and
technical inflexibilities in an enterprises [
Two classes of research are presented that deal with adaptation of information systems. The first category focuses on the context and changing requirements while the second category addresses architectural reconfiguration and modification.
Souza et al [
Researchers in software architecture analysis address change with a particular
focus on the effort and process required to enable implementation and modification of a
software system to accommodate changes in stakeholder requirements. For example,
Bengtsson et al [
Both of the discussed categories enable adjustment of information systems in
accordance to changing context, hence focusing on overcoming the first adaptation
challenge. However enterprises face emergent needs that arise as a result of interactions of
social and technical entities within the organization and its ecosystem [
The i* modeling framework is known for its ability to capture intentions of different actors when modeling enterprise requirements. As part of i*, dependencies among actors is modeled to enable analysis regarding how actors rely on one another to satisfy goals and softgoals, acquire resources, and perform tasks. However such dependencies entail commitment among actors and can introduce barriers to change. In this section we propose a method to investigate the most influential elements in i* networks of dependencies to enable identification and analysis regarding highly influential dependencies that can cause inflexibilities.
To determine which dependencies have a higher potential of being barriers to
change the degree of coupling among dependencies is (algorithmically) computed
using second-order dependencies in i*. The approach enables identification of
dependencies that have high impact on the overall network of dependencies. In addition
it can enable qualitative and quantitative analysis regarding how a certain change will
impact the network of dependencies and enterprise capabilities. An example of
qualitative impact analysis on IT and organizational capability alignment is discussed in a
related work [
A second-order dependency is defined as the reliance of one dependency to another to the extent that it cannot perform with the required quality unless the former dependency is satisfied. In other words second-order dependencies refer to dependencies among (first-order) dependencies.
To extract second-order dependencies one can investigate the strategic rationale model of i*. If the dependee-side element of an i* dependency (element which resides in dependee) such as D 7 in Fig. 1 (the dependee-side element in D 7 is Resource Allocation), and that element itself is dependent on some other actor as is the case with Resource Allocation (which is dependent on the IT support capability), then D 7 is dependent (second-order) on D 10.
If the source element of a dependency is comprised of sub-elements where subelements are identified through contribution, decomposition and mean-end links (for softgoals, tasks and goals respectively); then a second-order dependency exists from the dependency to each of the dependencies of sub-elements. For example Virtualization (a task of Infrastructure Management capability) contributes to the dependee-side element of D 12 (Scalable Resource Pool) and depends on Virtualization Expertise (resource) which is provided by the IT Support (capability), therefore a second-order dependency exists from D 10 to D 12. The second-order dependency exists as Easy Resource Allocation (D 12) relies on setup and engineering of the virtualization infrastructure which provides Scalable Resource Pool.
The dependency propagation graph presented in Figure 2 enables analysis of the rippling effects of dependencies among actors in i*. Its construction can be automated in a tool using the rules described earlier in this paper. The directions of arrows depict the path in which rippling effects of a change can propagate, i.e., the opposite direction of the dependencies in the SR model. In this graph the dependencies are grouped into rows according to the dependee actors. The grouping facilitates visual analysis regarding how changing certain capabilities or systems will impact the overall network of dependencies. With additional information the graph can serve as a roadmap to quantify economical contribution of each dependency and its role in value creation.
According to graph presented in Figure 2, D 11 which refers to Virtualization Expertise provided by the IT Support capability to the Infrastructure Management capability of Figure 1, is a sensitive element as deficits in resources to design and govern a virtual infrastructure can have extensive impacts on the functionality and use of information systems across the enterprise. Furthermore making changes to the process (i* task) by which this resource is provided, i.e., Expertise Development in IT Support, can impact many other applications and organizational dependencies. Hence when making decisions regarding its evolution, one should carefully consider consequences and alternatives. 4
Building the flexibility required to enable enterprise transformation is a major concern in both management and IS research. While many have proposed approaches to deal with automated adjustment of IS, there is a lack of methods that allow analysis regarding inflexibilities that arise in a socio-technical context. An approach that enables analysis and identification of potential inflexibilities is introduced by investigating second-order dependencies in an i* model of enterprise capabilities.
As future work, in order to fully recognize causes of rigidity, one needs to investigate the degree of impact that a certain sensitive dependency has. This can be achieved through assignment of quantitative measures to the edges of the dependency propagation graph. The measures can be assigned as a weight to depict the importance of the second-order represented by the edges. If the dependency is resulted from a softgoal, the contribution links can serve as a roadmap for assigning values. Such quantitative measures assist human judgement regarding the sensitivity of an element and how it can cause barriers to change.
The results of the analysis can be used at design time to enable accurate planning and mitigation of the risks imposed by any potential inflexibility. In the case presented in this paper, careful planning and consideration in training human resources with the skillsets to manage a virtual infrastructure should be a major concern at the design time. Furthermore the dependency graph can be used at runtime to monitor and measure potential inflexibilities in order to alert the changes in the probability of some dependency causing inflexibility.
Analyzing and interpreting the significance of second-order dependencies without tool support that points to the source i* elements is difficult and reduces the practical usage of the method. Furthermore as the models scale and enterprises grow creation of the graph requires automated tool support that can take an i* model and produce second-order dependencies based on the proposed algorithm.