Metrics and measures have always been the subject of quite a lot of research works in the Requirements Engineering (RE) community, including about intentional models of Goal-Oriented RE (GORE) such as those of i*. However, using recent developments of the Measurement Theory, in this paper we show that the concept of Measurement Framework (MF) for soft-systems is useful for the analysis of business service systems that need long-term service agreements based on consistent measurements at all stages of their life-cycle (from inception to operation). We show that with two kinds of goals and softgoals based on MF, it is possible to improve (a) the elicitation of functional and non- functional requirements, (b) the structure of the i* models, and (c) the consistency between run-time measurements and the model-based assessments of business services at early stages of RE.
IT based business systems are enablers to create business opportunities across
business entities boundaries that belong to complex business constellations. Being
able to analyze those opportunities contributed to the successful application of the
requirements engineering activities based on i*, in particular with the use of the
concepts and analysis techniques based on the Strategic Dependency Diagrams and
the Strategic Rationale Diagrams. However, for the business services, which are
means often used in those business constellations, agreements between the service
stakeholders (e.g. the service provider and the service client) must be faithful to what
actually happens during the service performance. This is why those agreements, often
called service-level agreements, must be based on empirically valid measurements. If
not, some party will assess negatively its business collaborations and can find other
business services offered on the market. Progress of our current research projects is
reported with an example of a case study followed in the Construction Sector [
Section 2 motivates the objectives of the research in the context of GORE, before presenting in Section 3 our method based on measurement frameworks (MF). We conclude and explain future works in Section 4.
Note that this paper uses some terminology of Measurement Theory that may conflict with GORE terminology (e.g. about model-based evaluations). As explained in the introduction, it is necessary to have consistent assessments of the business services at the different stages of their life-cycle, from requirements engineering to operation. Empirically valid measurements should be the basis of all assessments used thorough the life-cycle of the business services.
Recent developments in Measurement Theory for soft-systems can be used to
support the development of RE techniques in i* that focus on those empirically valid
measurements provided at run-time. Our aim is to provide RE support for complex
measurements having an empirical validity in the context of GORE for business
services. This aim has come up from needs identified during more than 6 years of
research in related contexts: RE for risk management [
Since long Measurement Theory [
The recent developments have pointed out three important characteristics of measurements (see Fig. 1): measurements must be empirical, objective, and intersubjective. Briefly, “empirical” means that the (empirical) attribute must be clearly identified; “objective” indicates that the measurement model must respect all (empirical) properties of the attribute (e.g. measured length must respect the size ordering); and “inter-subjective” expresses the fact that the measurement model (and scales) is a belief shared by everybody using (or referring to) the MF. All three characteristics make a distinction between measures, evaluations, and preferences (see Fig. 1). Measures and evaluations must be empirically validated. During RE elicitation processes sometimes we have to model attributes of systems that are not yet scientifically well understood. For instance, the usability attribute of IT applications: do we have a precise definition of what “usability” is? Research is still needed for validating empirical and objective measurements of usability. According to the Fig. 1, if neither empirical, nor objective, a wide place for (implicit) preferences is left in those models.
This motivates us to define specific types of softgoals and goals: when their description is complete, those types of goal and softgoals can be specified on the basis of measurement frameworks. For those two specific types of goals and softgoals, a distinction can be made between them: the fulfillment of a goal is measured, whereas the fulfillment of a softgoal is evaluated. (For now on in this paper, we consider only goals and softgoals of these types.)
Our research works present two methods: one [
A partial view on the case study in the Construction Sector [
Finally, at the bottom, both MF are instantiated to the specificities of the collaborative work needed in construction projects, in this case, the sharing of documents and expertise. The three horizontal lanes at the bottom separate the high-level strategic diagram, the MF resulting from the instantiation of the two preceding MF and one “solution” (at the very bottom). The actors and their dependencies are not shown. The solution describes functional requirements that are derived from the measurement. The evaluation procedure imposed by the MF rigorously structures the arguments of the model-based assessments: see the shaded area on the evidences showing that the goal “reactions time are monitored” of the MF (i.e. an outcome) will be fulfilled.
Using goals and softgoals based on MF has practical consequences for RE activities and requirements models. First, the measurement model of the MF provides constructive insights to derive functional and non-functional requirements (see the top and middle models). Second, our specific kind of goals and softgoals basically imposes an expected measurement profile to the result of a MF that can be applied to detailed requirements of alternatives solutions. This increases the structure of large i* models respecting the separation of concerns of the different MF. This is sketched at the bottom of the Fig. 2. The separation of concerns can be expressed with (a) the attributes (and their MF); (b) the measurement profiles; (c) the solution (as measurement target). Third, the MF provides a complete specification of a measurement system that can be implemented and used at runtime for an objective measurement of the fulfillment of goals and softgoals. (Those measurement systems can also be used during operational tests.) s e s o p r u P s e m o c t u O s e s o p r u P s e m o c t u O R ieuq trSa i-hgH rem teg Lve ten ic le s
Fourth, the same measurement specifications can be used for creating the arguments of model-based assessments of the fulfillment of goals and softgoals as explained in Sec. 3.1. (Measurements are made at run-time, whereas model-based assessments provide expected measurement results.) However, when carefully made the results of the assessments will be consistent with the results of the measurements. (Those model-based assessments can also be used for model-based simulations.)
As seen in the example, the design method is based on the insertion of soft-system measurements between the strategic goals and the solutions. This creates a valid relationship between the (elicited) measurement profiles and alternatives of functional (and also non-functional) requirements. For a bottom-up design process it is possible to collect arguments and evidences from the analyzed solution and derive the measurement results of the model-based evaluation defined by the MF. For a topdown elicitation, measurement results of model-based evaluations can be imposed in order to derive the requirements on the basis of refinement patterns (cfr. KAOS by A. van Lamsweerde) or strategies (cfr. MAP by C.Rolland) that generate alternative solutions (not shown). For instance, concerning monitoring activities (outcome 2.1.3), in the selected measurement profile the monitored activities can include the sharing of documents and exclude the sharing of expertise. The advantage of this design method is to benefit from the constructive insights given by the MF of the soft-systems.
In recent overviews [
Using the same MF for requirements analysis and for defining the specifications of the measurement system and method, and for performing the actual measurements during system operation, this provides a greater consistency between the argument structure used in the requirements model, and the argument structure used at run-time.
In this paper we have shown that the recent developments of Measurement Theory
adapted to GORE methods, in particular i*, gives a sound basis for an interesting type
of goals and softgoals to guide the elicitation and analysis of functional and
nonfunctional requirements of business services. To benefit from the scientific knowledge
of a number of MF defined, improved and validated by the scientific community
(often using social sciences research techniques), the efficiency of the proposed
methods should be improved by providing tools supporting (a) the import of
measurements frameworks into a knowledge base, (b) the definition of refinement
patterns and strategies needed by the elicitation activities, (c) the assessment of the
requirements models [