5.3 Summary and Evaluation of the Approaches

First, a summary of the nine approaches to enterprise modeling is given according to the dimensions of presentation from section 5.1.3. Then comes an evaluation of the approaches in terms of the criteria outlined in section 5.1.1, and finally, some reflections are made upon observations from the survey.

5.3.1 Summary of the approaches

The assessment of state-of-the-art approaches to enterprise modeling is summarized in table 5.1 according to the six variables from section 5.1.3. In addition, a variable called "Fit" has been introduced, being a coarse estimate of how appropriate the modeling approach is related to the perspective on enterprise modeling taken in this thesis.
Name of approach  Background  Purpose  Type of work  Worldview  Models  Modeling  Fit 
F3  ISE  I, II  RD  Cons  Yes  Yes  Hi
Enterprise  KE  I, II, III  Obj  Yes  No  Lo
TOVE  CIM, KE  I, II, III  MI  Obj  Yes  No  Lo
Caesar  OD  I, II  Obj  Yes  No  Mid
Metis  SE  I, II, III  MI  Obj  Yes  No  Mid
SSM  MI  RD  Cons  Yes  Yes  Hi
OMT  SE  I, II  MI  RD  Obj  Yes  Yes  Mid
OOram  SE  I, II  MI  RD  Cons  Yes  Yes  Hi
SAMPO  ISE  RD  Cons  Yes  Yes  Mid
Table 5.1: Summary of surveyed approaches to enterprise modeling

A few remarks on the use of abbreviations in the table:

5.3.2 Brief evaluation according to criteria

A general impression from the survey is provided according to the criteria.

First criterion: Understanding is primary purpose

All of the surveyed frameworks claim to support human understanding, which is the primary focus of this research project. However, the perceived consequences of such a claim varies significantly among the approaches. E.g., the Enterprise approach does not seem to differentiate between human understanding and what may be denoted computer "understanding". This might be a valid assumption with an objectivistic worldview, but not with a constructivistic one.

Another observation is that most approaches have several purposes (and they are not internally prioritized). Approaches that seek to fulfill many purposes necessarily have to be more comprehensive than approaches that only focus on one purpose.

The purpose of enterprise modeling in F3 is both I and II. Hence, requirements to F3 must be more comprehensive than requirements to SSM, which only focuses on I.

Second criterion: Focus on both models and modeling

All the approaches provide extensive discussions of models, with SSM having the least comprehensive discussion. The Enterprise approach and TOVE even define detailed ontologies describing the phenomena that can be modeled. However, modeling is not discussed to the same degree. A frequent and highly illustrative observation is the use of the term method or methodology when one really refers to a language. An example can be found in Auramäki et al. (1988:135, italics added):
"The graphical methods for describing discourses are (1) a discourse graph, and (2) a conversation graph."
A method can never be graphical. A method can never be a graph. Languages, on the other hand, can be graphical and presented in form of graphs. Even if the statement is taken out of context and might represent a "slip of tongue", it is representative for an attitude that plagues some (enterprise) modeling communities: A modeling method is considered equivalent to a modeling language. Focus on the modeling process is required when the perspective on enterprise modeling is reality construction.

The observation that models receive considerably more attention than the modeling process is consistent with a claim made by Rolland and Cauvet (1992:1) in their comprehensive survey of trends and perspectives in conceptual modeling (italics added):

"In contrast to the mature level of research on conceptual models, there is an evident lack of understanding and formalization of the conceptual modelling process."
Furthermore, even if some approaches do discuss the modeling process, they are in terms of the modeling language and not the professional language of the actors in the domain that is modeled. Terms like objects, relationships, processes, speech acts, etc. are appropriate for representation and manifestation of a domain, but not for sense-making within the domain. The exception is SSM.

Third criterion: Type of work that is modeled

The types of work typically modeled with the dedicated enterprise modeling approaches are associated with replication risk. This should not come as a surprise: Modeling of routine work is, at least in theory, easier than modeling of work associated with design risk, as the prime characteristic of design type of work is that neither the product nor the work process are known (at a detailed level of abstraction).

The observation that type of work being modeled is dominated by replication risk is also consistent with the observation that most modeling methods focus on representation and manifestation, and not on sense-making. If the domain that is modeled is widely agreed upon, there is little need for sense-making -- the modelers can focus on representation of the domain.

5.3.3 Reflections on observations from the survey

One fundamental observation from the survey is how the notion of a constructivistic worldview seems to exist at two different levels: In the Universe of Production and in the Universe of Modeling.

Recognizing that reality associated with Universe of Production is socially constructed calls for development of enterprise modeling approaches that are able to model the construction process in the Universe of Production -- i.e., creating modeling languages that are tailored to represent the construction process and alternative local realities. OOram and SAMPO exemplifies these approaches: OOram is well suited for representation of role models that describe various aspects of objects. SAMPO is well suited for representing communication in the Universe of Production.

However, recognizing that the Universe of Production is socially constructed is not sufficient. The main issue that is advocated in this thesis is that the Universe of Modeling is socially constructed. The immediate consequence of this view is that the modeling process necessarily becomes a construction process. Being able to represent the construction processes in the UoP may be considered important, but the vital requirement is to see the UoM as constructed. SSM is the approach that seems to have recognized this to the fullest extent.

Another reflection is that the Universe of Research is constructed as well. To do research on enterprise modeling is reality construction (and the construction process is social as the researcher makes sense through use of already constructed concepts and theories, and communicate with other researchers through dialogue, scientific conferences and papers). Dahlbom (1992:122) discusses the idea of science as construction briefly. These considerations are revisited in chapter 10 when evaluating the research accomplishment.