Stone Rice Big dot Net

Finally, I got to read an interesting working paper by Oliver Thomas (2006) about the concept of reference models. The paper is written in German (there is an English version of the paper available, though) but since I want to train my English …

Reference models

For all of you how have never heard the term reference model: it is a unique research area of German Information Systems Research and refers to conceptual models (i.e. semi-formal descriptions of a system) that can be used as templates for creating individual (i.e. specific for a company) descriptions of a desired system (the so called application model ). So reference models work very similar to recipes - recipes for designing business processes or information systems.

So what is the problem with reference models? Well, our discipline is called Information Systems Research . Hence, we should be able to present some kind of proof for attaching the prefix reference to our conceptual model. However, this requires an exact definition and conceptualization of what a reference model is and what benefits it provides.

As usual in Information System Research (and probably other disciplines as well) no commonly agreed definition of reference model exist. Most authors describe reference models by explaining three distinctive characteristics:

1. Instead of solving a specific problem (e.g. the supply chain management system for company XY) they focus on a class of problem (e.g. supply chain management systems in the automotive industry).
2. That is why authors who call their models reference models at least implicitly claim that the reference model has some degree of universality/generality for the intended domain - which should be evaluated and demonstrated.
3. Furthermore, reference models contain recommendations on how to build the underlying type of system efficiently and effectively. Since reference models are used to design new information systems or adapt COTS (Commercial of the Shelf) software the desired information system is not yet built. By claiming some degree of universality, it becomes evident that reference models try to provide recommendations for such future systems. Again, this claim should be established.

However, establishing claims of reference models is very difficult problem from a theoretical and empirical perspective. For instance, from a empirical point of view, evaluating the utility of reference models has to eliminate confounding factors.

Usage-centric reference modeling

Now, Thomas (2006) argues that the characteristics of universality and recommendation are not constitutive for the concept of reference model. Thomas (2006) provides the following support for his argumentation:

• The characteristic of universality is bound to the specific domain a reference model was built for. Hence, it cannot be universal.
• The characteristic of recommendation is not operationalizable, i.e. it cannot be measured. Furthermore, this characteristic is highly subjective as it depends on the actual situation of the user of a reference model.

That is why Thomas (2006) proposes a new constitutive characteristic: acceptance of the character of reference by the actual users . He argues that only the actual users of a reference model can decide whether it was useful or not. The claim of reference by the authors is (almost) irrelevant. Thus, Thomas (2006) coins the term usage-oriented reference modeling.

Evaluation of reference models would be simplified to identifying users. While the concept of usage-oriented reference modeling sounds promising it has some weaknesses (Besides, the emergent character of reference eliminates the need of a sub-discipline called reference modeling in the sense of a systematic construction of reference models and reference modeling methods and tools - every model could be a reference model. So in the best case reference modeling is the same as normal conceptual modeling):

Models already are usage-centric

According to Stachowiak (1973) a conceptual model is already usage-centric. Without any usage and utility a description does not become a model.

But when is someone using a reference model? The user has to gain some advantage by exploiting the reference character:

• faster design process,
• lower costs,
• better quality

are typically proposed benefits of reference models - the same benefits usually stated in discussions of patterns, software components, frameworks, etc. However how can a potential user assess the character of reference before applying it?

Evidences or trust are necessary for accepting the character of reference

The key issue is reusing other work. But how do users pick models they can use as reference? Either they can constructed the models them self or someone claims the model at hand is a reference model. In the case that the user trusts the claim the model becomes a reference model, according to Thomas (2006). Again, when does a user trust a reference model? Either many users have already used the model (e.g. the IT Infrastructure Library) and can report some benefit from it (are there any studies that can establish benefits of applying ITIL) or the reference model supplier has to provide some evidence. So in the end users will apply reference models only if there is either trust or evidence of the reference model’s capabilities. Hence the user needs information about the utility of a reference model.

Reference model markets

Usage of reference models requires established claims of reference. Furthermore, as discussed above, reference models contain recommendations about future information systems. It is very unlikely to get users apply a reference model without any evidence that it contents are beneficial.

Still, the usage-centric perspective on reference models can be fruitful. It requires some conceptualization of a reference model market or marketplace. The market can serve as an coordinating instance that provides necessary information on utility of reference models. Potential users may use market information to assess the potential value of a reference model. However, as an initial requirement reference model constructors still have to establish reasonable claims of the benefits their reference model provides.

Conclusion

The proposed usage-centric view on reference models is not constitutive of reference models. A model does not become a reference model just by reusing it - it gets its reference character from established (this requires usage of the model) claims and documented benefits. Hence, the intersection of intended and perceived utility for a class of users is the constitutive characteristic of a reference model. Usage is required for the utility-oriented perspective on reference models but not sufficient. So evaluating the claims of reference models is the key challenge in reference modeling.

References

Thomas, O. (2006). Das Referenzmodellverständnis in der Wirtschaftsinformatik: Historie, Literaturanalyse und Begriffsexplikation (Arbeitsbericht 187). Saarbrücken: Institut für Wirtschaftsinformatik (IWi), Deutsches Forschungszentrum für Künstliche Intelligenz (DFKI).

Stachowiak, H. (1973). Allgemeine Modelltheorie. Wien, New York: Springer.