In 2013, Karagiannis and Schwab introduced the hybrid modeling approach. The approach is motivated by the need for “modular construction systems“ and “open modeling approaches“, covering the “divers business requirements in a fast moving environment“ thereby amplifying flexibility (KARAGIANNIS AND SCHWAB, 2013, p. 3). The authors define two major require- ments for their hybrid modeling approach: (1) “the result of hybrid modeling is a modeling method that falls in the category of a graphical semi-formal modeling method based on a meta- modeling approach“, and (2) “the hybrid modeling method is implemented and offered in form of a modeling tool“. Due to this focus, the authors also define a “deliberate procedure compris- ing different steps“(KARAGIANNIS ANDSCHWAB, 2013, p. 1) towards the implementation of a modeling tool for a hybrid modeling method - in the following referred to as conceptualiza- tion.
3.4.1 Conceptualization Life Cycle
The process of implementing a hybrid modeling tool in accordance to a hybrid modeling method is divided into the phases create, design and compile. The conceptualization is dependent on the ADOxx meta modeling platform (see section 7.3.1 for a detailed introduction to the ADOxx platform). Figure 19 illustrates the conceptualization by means of a life cycle.
In order to foster the implementation of hybrid modeling tools, the authors state, that “the better the resulting hybrid modeling method is conceptually composed, the smoother the trans- formation of the concepts to the codes can be done“ (KARAGIANNIS AND SCHWAB, 2013, p. 3). The integration of different building blocks of multiple modeling methods requires the analysis of the different abstraction levels. Three “states in a way meta models vary“ (KARA- GIANNIS AND SCHWAB, 2013, p. 3f.) have been identified: vertically, if they show different
Figure 19: Hybrid modeling conceptualization life cycle (KARAGIANNIS ANDSCHWAB, 2013, p. 5)
levels of abstraction; horizontally, if the concepts of the building blocks are on the same ab- straction level but they describe semantically different aspects; and both, if a combination of the former two states is given. These different states are explicitly considered in the conceptu- alization of a modeling tool, described in the following:
3.4.1.1 Create Phase
The first phase of the hybrid modeling approach “is related to the application scenario and the need of the user and refers basically to the selection process the user performs for identifying the existing modeling concepts within the hybrid method“ (KARAGIANNIS ANDSCHWAB, 2013, p. 4f.). Basically, the first phase collects all scenarios that should be supported by the modeling method to be built. These scenarios lead to the building blocks that need to be integrated in the hybrid method. Having the building blocks defined, the relationships and dependencies between the concepts of different modeling methods need to be specified. “Depending on the maturity of the underlying building blocks the description can be of formal but also informal kind or can for example include the definition of a consistent meta model of the hybrid modeling method on a sole conceptual level“ (KARAGIANNIS AND SCHWAB, 2013, p. 5).
The creation of a hybrid modeling method is strongly aligned to the components of modeling methods introduced by Karagiannis (cf. Figure 8). The hybrid modeling approach therefore analyses all three major parts of a modeling method: modeling language, modeling procedure, and mechanisms & algorithms.
3.4.1.2 Design Phase
In the second phase, the input of the creation phase in combination with the intended platform for the development of the hybrid modeling tool is used to elaborate “the preconditions for the later implementation respectively customization phase“(KARAGIANNIS AND SCHWAB, 2013,
p. 6). The development platform, and the functionality it provides, influences the design deci- sions of the method engineer in the design phase and the implementation phase. The purpose of the design phase is the definition of two meta models, a conceptualization meta model and an implementation meta model. The separation of those two meta models reflects the platform- independent integration resulting from the create phase and the platform-dependent realization in the design phase. This realization is performed by mapping the concepts of the hybrid mod- eling method meta model to the generic concepts provided by the ADOxx platform by means of the ADOxx meta meta model.
Due to the fact, that the hybrid modeling approach is directed towards the implementation of graphical modeling tools, the notation of the concepts included in the hybrid method plays a mature role. Karagiannis and Schwab identified two combinations that lead to inconsistencies and therefore to design decisions for the method engineer (KARAGIANNIS AND SCHWAB, 2013, p. 9): (1) multiple, inconsistent notations due to the existence of more than one concept that need to be merged; and (2) missing notations, as the precise and unambiguous specification of the notation might be missing for one or more modeling concepts.
Along with the discussion about the assimilation of the possibly differing notations goes the consideration about the semantic alignment of different concepts. In order to decide about the semantic mapping between the building blocks, the authors propose the usage of a semantic comparison table. On each axis of this table, the concepts of one modeling method are posi- tioned. For each cell of the resulting table, i.e., for each pair of concepts of modeling methods, the semantic correspondence is defined. The scale ranges from (−), “not applicable - com- parison of modeling class and relation class“, over (!=), “unlike, does not correspond at all“, and (∼), “natural language description and use show similarities“, to (1:1), “identical in their natural language description and use“ (KARAGIANNIS AND SCHWAB, 2013, p. 10). Table 2 shows an excerpt of the comparison of i* and BPMS modeling method concepts.
The third mapping task of the design phase (after mapping the notations and the semantics) is concerned with the syntactic mapping of modeling concepts derived from different modeling methods. The syntax “describes the dependencies and constraints in between the modeling concepts and is furthermore represented in the description of the properties of these in form of attributes“ (KARAGIANNIS AND SCHWAB, 2013, p. 10). The syntactic mapping therefore must define the integration of the attributes of semantically identical concepts (e.g., identified by a 1:1 value in the comparison table between these concepts).
3.4.1.3 Compile Phase
In the compile phase of the hybrid modeling approach, the concrete realization of the designed hybrid modeling method is performed using the ADOxx meta modeling platform. As several steps do not require an implementation but a customization of predefined platform functionality,
Table 2: Semantic comparison of i* and BPMS modeling method concepts (KARAGIANNIS AND SCHWAB, 2013, p. 10)
Modeling Concepts i* classes and relations
Actor Agent Position Association Dependency
Link Link BPMS classes and relations Organizational unit ∼ != != − − Performer ∼ ∼ != − − almost 1:1 Role ∼ != ∼ − − Position != != != − − Is subordinated − − − ∼ != Belongs to − − − ∼ !=
Legend: !=, unlike, does not correspond at all; 1:1, identical in their natural language description and use; ∼, natural language description and use show similarities; −, not applicable - comparison of modeling class to
relation class
this phase is also referred to as customization phase. The main task is the decision about the most appropriate representation of the hybrid modeling method to the modeler, e.g., using a modeling app for mobile devices, a locally installed stand-alone application, or a web-based client-server modeling environment.
The compile phase does significantly depend on the development platform, e.g., a meta mod- eling platform. The concepts and functionality provided by the platform must be used in order to implement all components of the designed hybrid modeling method. The result of the com- pile phase is a complete hybrid modeling tool, enabling human beings to create, interpret, and process the generated models in an appropriate way.
3.4.2 Discussion
The approach concentrates on the combination of “building blocks“ of modeling methods in or- der to generate flexible hybrid modeling methods. Especially the create phase show a significant amount of similarity compared to the Modeling Scenario provided in the MUVIEMOT method (cf. section 6.3.1). Both consider the specification and integration of modeling views that op- tionally originate from different modeling languages or methods. The emphasis of the hybrid modeling approach is on the integration of given modeling methods by stepwise integrating the building blocks of the modeling method framework illustrated in Figure 8.
The modeling procedure of the resulting hybrid modeling method is not considered appropri- ately by the approach. This is a major drawback, as the way of carrying out multi-view modeling plays a mature role when it comes to utility and efficiency of the tools. It seems unlikely, that merging different modeling methods does not require the integration of the corresponding mod- eling procedures and mechanisms & algorithms. Lastly, the hybrid modeling approach targets the implementation of modeling tools given a predefined set of modeling methods. It is not clear, how the multiple modeling methods are transformed into multiple viewpoints of the tool conceptually.