Model Based System Development/Research Themes/Conceptual Modeling

Conceptual Modelling Methods: Utility, Design, and Operationalisation

The TEE group has a fundamental interest in "conceptual modelling", covering languages needed to effectively represent models, different formal semantics, as well as the processes involved in creating high quality conceptual models (elicitation, conceptualisation, verification and validation). In our research we are strongly inspired by the creation and use of conceptual models in practice. Conceptual models are used in a variety situations covering operational processes in organizations, industry, health care, information systems, etcetera, as well as the analysis, design and implementation of such 'systems'.

Research strategy

Traditionally, the field of computer science has had a strong focus on formal aspects of models and modelling techniques. As a result, there is a tendency to relate the quality of models exclusively to formal properties of models. Our group has a tradition in research on the formal definition of modelling languages for information systems modelling (HPW93, PW95, HPW97). Inspired by practical experiences, however, (including the personal experience of some of our group's members when working in practice (HP02, LP03), we have gradually become aware that there is much more to the quality of a model than its formal properties, making it more of a relative notion depending on the needs at hand. Why is the model needed? Should it serve as a base for analysis? For which audience? Even more, the situation in which actual modelling is to take place has a huge impact on how practical it is to achieve the quality ambitions. How much time/money is available to produce the model? Do we have access to the right informants/stakeholders/...? (JV+03, VHP04, PVH05).

In relation to the above issues, a number of international researchers, mostly from the fields of Information Systems and Enterprise Engineering, has recently been pushing towards a shift in focus from language-oriented foundations of modelling to issues of operationalisation of methods and techniques. This requires a partial shift in research methodology, and creates opportunities for new interdisciplinary connections with established fields like operations research, human-computer interaction, intervention methodology, decision making, problem solving, and so on. Importantly, our goals are practice-oriented but still distinctly academic: we do strive for (applied) theory creation and improvement, taking operationalisation of modelling and model use as our prime object of research. However, such research does require close cooperation with practice (industry, the public sector). Our longer term strategy is therefore to first capture the attention of advanced practitioners and then, through dialogue and empirical testing of theory-based methods and techniques, increase the academic depth of the field. The first stage of this approach is now evidently bearing fruit.

The primary research paradigm for our field is Design Science. Design science includes not just systems design, but also method design. In addition, we tend to look at operational methods for enterprise modelling as interactive systems. The shift in focus discussed above requires us to explore research methods relatively new to the field. In this respect, we have been able to make small but significant contributions in developing new viewpoints on operational modelling and related research methods, and have been among the leaders in that field. This development is still in full swing, but has in already led to the creation of a new international research community, with collaborations, events, and book series of its own.

To be able to gather empirical data underpinning our theoretical work it is crucial to invest in a long term application domain. Based on our personal interests and past experiences, we have selected to focus our empirical efforts in the context of the design and evolution of enterprises; Enterprise Engineering. Our group is prominently involved in the field, and is instrumental in developing it. Fore details, see the Enterprise Engineering section. Separate from our fundamental interests, the priorities shaping our research agenda depend on the maturation process of the enterprise engineering field.

Research themes

In order to improve our own focus within the more general shift towards operational issues in Conceptual Modelling (mostly applied in Enterprise Engineering), we distinguish four key themes in our research:

1. Quality of modelling
2. Increasing the return on modelling effort
3. Act of modelling
4. Reference models as theory

Quality of modelling

Within this theme we are driven by questions such as: What are relevant quality properties of models and the processes by which they are created? How to assess the required quality of a model and its modelling process in a specific situation? How to assess the achieved quality level? What is to be regarded as a high quality model will differ from one situation to another. For example, depending on the specific situation, the requirements with regards to:

• What should be in-/excluded in the model?
• How explicit should the model be?
• Should the model be executable?
• What degree of agreement should exist about the model, by whom?

are likely to differ.

Within this theme we are inspired by work from e.g. Krogstie and Solvberg (KLS95), Stachowiak (Sta73) and Peirce (Pei69). We contributed to this by: PHPR08, BHPW07, PHV05, AB+07, HLP06.

The quality of a model is influenced by the process in which it is created (see act of modelling theme), but it is also enabled by the quality of the modelling language used to represent the model. When using an informal modelling language, only geared towards the creation of high level sketches, it will be nearly impossible to create a model with a high level of mathematical formality. Conversely, a well formalized modelling language may prove to be a burden when creating an draft design of some system in a highly creative environment. In our quality of modelling work, we therefore study the aptness (and design) of modelling languages for given modelling situations. Some of our results are discussed in: PVH05, HPW05. Requirements on modelling methods include the formal definition of languages. However, we have developed a broader view on such requirements, which is reflected in our work over the last 5 years.

The PhD project of Denis Ssebuggwawo focusus specifically on the design of a method, including a set of metrics, for the broad, goal-oriented assessment of modeling sessions. Concepts for such assesments come from existing work on Quality of Modeling, but also from fields like Human-Computer Interaction and Operations Research. For references, see the Act of Modelling.

Increasing the return on modelling effort

The creation of conceptual models takes time and effort, in other words, costs money. How can these investmens be put to good use? How to increase the return on investments put into models? In other words, how to achieve Return on Modelling Effort (RoME)? This is what drives us in this theme, pushing us to look at the different roles models may play:

• Provide insight.
• Provide guidance.
• Act as a steering/regulative instrument.
• Act as a common frame of reference.
• Be executable.

and study ways to improve the ability for models to play these roles:

• Develop strategies to analyze, mine, animate (gaming!) and visualize models in order to gain insight.
• Develop techniques to use models to guide people/actors in their operational work.
• Develop techniques to be able to execute/animate models (as soon as possible) in the system development life-cycle.

Again, just as the quality of models is enabled/disabled by the modelling language used, the ability to provide returns on a modelling effort is also enabled/disabled by the language used. To be able to do a formal analysis of properties of a system portrayed by a model, the modelling language used must have a high enough formality to allow for such an interpretation of the model.

Within the ArchiMate project, experiments have been conducted into model based analysis of enterprises. Within the MBSD department several other experiences exist with model analysis as well as the execution of models. Some of these results may be translated to an enterprise engineering context. Currently we are also bridging the gap to operations research.

The act of modelling

Given the quality requirements on a desired model, and its creation process, as well as the potential return on modelling effort, the next challenge is to create the model as effectively and efficiently as possible. In other words, how to shape and guide modelling processes in an optimally effective way?

In practice, models are not created by people in isolation, but rather in the context of groups of people. The creation of models often involves a collaborative process in which different actors will play a role, where each actor will bring their own stakes, goals and abilities to the table. What happens when people create models? What cognitiveand interactive processes are at play? How can models be created in a collaborative setting? How can a shared understanding, agreement and commitment be warranted? Here we also observe and study how people model naturally; in other words, before adding interventions aimed to improve their effectiveness.

In view of an already growing demand for lightweight formal models (for a large part due to advances in AI and EE), enabling domain experts (usually, people from business management and operations) to create models with minimal involvement of expert modellers. Breaking the knowledge acquisition bottleneck is another driving theme in our Act of Modelling reasearch.

We base our current work concerning the above issues on exploratory work we did some years ago, which led us to view the act of modelling as a rational, grounded conversation. This implies a string emphasis on communication, language, and intersubjective interaction in modelling.

Resulting publications:

1. act of modelling as a grounded, rational conversation [references]
2. making modelling goals and entailing strategies explicit [references]
3. modelling strategies for collaborative modelling in policy making [josephine references]
4. studying and designing modelling methods as if they are interactive, collaborative games [refs stijn]
5. creating metrics and an evaluation system for operational modeling sessions [refs Denis]

Reference models as theory

Some specific application domains, such as information retrieval, knowledge representation, and information & knowledge markets, have such a high complexity that they warrant the development of conceptual models that actually are a theory in themselves. Having such reference models available provides an accelerator for modelling activities in specific situations as they can be used as a base for situation specific models. In the past, the TEE group was part of the IRIS department. Within the context of that department, work has been conducted on reference models (as theories) underpinning:

1. Information filtering [Paul de Vrieze's work]
2. Information markets [Bas van Gils' work]
3. Knowledge markets [Sietse Overbeek's work]

Application domains

Though most of our group's focus is on applying the above themes in the context of enterprise engineering, the questions raised clearly apply to conceptual modelling in a more general sense. Application to other fields is therefore quite possible. In fact, we have begun exploring a link with the related field of knowledge acquisition (a sub-field of Artificial Intelligence), essentially an application of conceptual modelling to the domain of knowledge intensive systems. From this central field, there are opportunities to branch off onto fields like AI and law and AI and health care.