Process modelling has over the years become an essential skill in Information Systems and Business Process Management practice. Consequently, more and more training programs have evolved, teaching different process modelling languages. Two popular process modelling languages are being compared in our experimental study. Experiment participants received extensive training in one language but not the other, leading to the expectation that learning outcomes would be better in the case of the familiar language. Our study provides empirical evidence that this is not the case. In fact, it is shown that participants achieved similar learning outcomes when confronted with the unfamiliar language. Our results lead to a fundamental question, namely whether it is actually an important teaching decision what sort of process modelling language is being taught. Our findings suggest that education and research in process modelling should focus on aspects other than the style, nature or features of languages and tools.
Implications for Academia
Our research results have implications primarily with respect to educational aspects. We have shown that users of EPCs understand BPMN diagrams equally well even though they were never exposed to this modelling language before. With respect to the university curriculum it must be concluded that it is neither of much use to include several process modelling languages into a single course, nor is it of much use to impose an obligation on students to learn several process modelling languages in several courses.
Additionally, our research raises another very important and interesting question: Can we build families of process modelling languages with the following properties: Language users of one process modelling language can easily switch to another language of the same family and have difficulties to switch to a language from another family. It seems likely that the results of this study can be replicated for other process modelling languages in which activities and events play a central role. But is it possible to replicate these results, e.g., for Petri nets, or are Petri nets so fundamentally different from BPMN and EPCs that our obtained results would not hold again?
A third implication for academia is to face the question of how process modelling learning outcomes are actually achieved. The cognitive theory of multimedia learning used in this paper suggests three elements involved in the process: content, content presentation and user characteristics. We have focused the element content presentation in our study and controlled for content and user characteristics. The next step would then be to study different types of content and different types of user characteristics, respectively. Could it be that certain types of process modellers are more receptive to certain process modelling tasks? Some prior studies in IS, indeed suggest that user differences in cognitive abilities, application domain knowledge and method knowledge affect the way that conceptual modelling and information acquisition is conducted. An important research question would then be o investigate how these user characteristics impact the way process modelling is conducted.
Implications for Practice
The main implication for practice is the insight that a new process modelling language does not pose an economic threat to an organization if the majority of BPM actors within this organization are users of a different process modelling language. It would appear that there is no immediate need for organizations to embark on extensive training courses every time the process modelling language in use has to be changed. Instead our findings suggest that a set of analysts equipped with adequate skills in one process modelling language will be fit to understand other process models too.
On the provider side our results suggest that carefully managed changes to process modelling languages are not unlikely to be accepted by a customer base. Such changes may always be necessary in certain situations and should be seen as an opportunity rather than a problem. For instance, providers may find the need to enhance the expressive power of a process modelling language to be better equipped for future and advanced process modelling needs (e.g., advanced workflow execution, support for web service specification etc.). The resulting differences in expressiveness and complexity of the language appears to be well-absorbed by the existing language user communities.
About the Study
This study has been conducted by Jan Recker and Alexander Dreiling. We would like to thank all involved participants. The full paper can be downloaded here.