Power-Modelling
Toward a More Versatile Approach to Creating and Using Conceptual Models
Ulrich Frank
University of Duisburg-Essen, Essen, Germany
ulrich.frank@uni-due.de
Keywords:
DSML, Interactive Modelling, Multi-level Modelling, Empowerment.
Abstract:
The prospects of conceptual modelling are widely undisputed. Nevertheless the current practice of conceptual
modelling remains unsatisfactory. Usually, modelling languages offer primitive concepts only—with respec-
tive effects on productivity and model quality. The creation of models is restricted to early phases of system
life-cycle. Hence, the benefits of models in later phases are ignored. Furthermore, the creation and use of
conceptual models is still restricted to experts only. In this paper, the outline of a new modelling paradigm,
referred to as power-modelling, is presented. It builds on the potential of domain-specific modelling languages
(DSML), application frameworks and reference models. It regards models as the primary medium to perceive,
interact with and change systems and the environment they are supposed to operate in during the entire sys-
tem life-cycle. For this purpose, power-modelling is built on an extensible set of multi-level DSML that fit
the conceptual perspectives of a wide range of prospective users and a common representation of models and
code, which allows overcoming the notorious problem of synchronizing models and code.
1 INTRODUCTION
If maturity comes with age, a few decades of research
and application should have resulted in a widely per-
fected state of conceptual modelling. There are in-
deed various signs of maturity. Both in Information
Systems and Computer Science it seems undisputed
that conceptual models are a prerequisite to manage
the complexity of large software systems. Further-
more, it is acknowledged that conceptual models are
much better suited than code to involve prospective
users and other stakeholders in the process of devel-
oping software. The benefits of conceptual models for
designing software systems do not come as a surprise.
After all, software systems are linguistic artefacts. On
the one hand, they are realized with some kind of im-
plementation language. On the other hand, as non-
physical artefacts they can be perceived—and used—
by humans only through some kind of linguistic rep-
resentation. At best, this linguistic representation cor-
responds to the language used in the targeted do-
main. Conceptual models are aimed at reconstructing
domain-specific languages in a way that prospective
users perceive them as familiar, while at the same time
they allow for transformations into implementation-
level languages. By structuring and eventually au-
tomating the transformation of models into code, as
it is pursued by approaches to model-driven software
development (Atkinson and K
¨
uhne, 2003; France and
Rumpe, 2007), conceptual modelling is promising to
substantially improve the productivity of developing
software systems.
But conceptual modelling is not restricted to mod-
elling software systems. Exploiting the potential of
information systems often requires re-organizing re-
spective action systems. Consequently, correspond-
ing modelling approaches, such as business process
modelling, turned out to be a good choice for sup-
porting people with analysing and (re-) structuring ac-
tion systems. The insight that corporate action sys-
tems do not only comprise business processes, but
also other subjects, such as goals, resources or or-
ganizational structure, contributed to the emergence
of approaches to enterprise modelling (Scheer, 1992;
Ferstl and Sinz, 2006; Frank, 2013). By integrat-
ing conceptual models of software systems with con-
ceptual models of action systems, enterprise models
promise to provide a foundation for jointly analysing
and designing information systems and the relevant
organizational context. Since more and more orga-
nizations have given up developing software on their
own, the original idea of using conceptual models
for designing software systems does not fit the de-
mands of many organizations anymore. At the same
9
Frank U.
Power-Modelling Toward a More Versatile Approach to Creating and Using Conceptual Models.
DOI: 10.5220/0005423800090019
In Proceedings of the Fourth International Symposium on Business Modeling and Software Design (BMSD 2014), pages 9-19
ISBN: 978-989-758-032-1
Copyright
c
2014 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
time, the ever growing complexity of IT infrastruc-
tures created additional challenges. Again, concep-
tual models of IT infrastructures, including various
representations of their high-level structure or archi-
tecture, have evolved as a remedy. While respective
models are often part of enterprise modelling meth-
ods, they are also featured by approaches to enterprise
architecture, which in general aim mainly at managers
and therefore emphasize a higher level of abstraction
(Lankhorst, 2005; Proper et al., 2010; Buckl et al.,
2010). Therefore, conceptual modelling supports a
wide range of activities related to analysing, design-
ing and managing IT infrastructures and correspond-
ing action systems. A relatively large research com-
munity, both in Information Systems and Computer
Science, may serve as further evidence for the matu-
rity of the field.
However, our brief assessment of the field’s con-
tributions may be deceptive—and maturity may have
its downsides, too. As we shall see, there are seri-
ous reasons for not being satisfied with the state of
the art in conceptual modelling. There are even rea-
sons to challenge the current paradigm of concep-
tual modelling. In any case, it does not seem ap-
propriate to further follow existing paths of research
by focussing solely on problems within this paradigm
without leaning back once in a while to reflect upon
other ways to conceptualize, develop, maintain and
use models. While I had my doubts for some time
whether we are doing the right thing, I started to in-
tentionally abandon some characteristics of concep-
tual modelling which I had not only taken for granted,
but which I used to preach enthusiastically. I am
still in the—sometimes painful, sometimes exciting—
process of realigning my perspective and my research
agenda. However, a few ideas have emerged that I
regard as promising. One of them is subject of this
paper. I dared calling it “power-modelling”, not only
to express that it might be suited to substantially pro-
mote the power of modelling as a tool to create and
modify systems, but also to empower users by sup-
porting them with sophisticated, but convenient in-
struments to use and modify the systems they work
with.
The paper starts with a critical review of the cur-
rent state of the art. Subsequently, an overview of
approaches that address certain shortcomings of con-
ceptual modelling is given. Against this background,
I will describe the idea of power-modelling by out-
lining the foundational concepts and by illustrating
how it could be implemented. Since the respective
research is in a very early stage and faces serious chal-
lenges, the conclusions will especially focus on future
research.
2 CONCEPTUAL MODELLING: A
CRITICAL REVIEW
There are various reasons, why the current state of
conceptual modelling might not be regarded as satis-
factory (Frank, 2014a). At first, there is the sobering
fact that the dissemination of conceptual modelling
in practice is still rather modest. Many software de-
velopers still regard it as dispensable. Managers per-
ceive it as a cost-driver that does not deliver measur-
able benefit. Finally, prospective users are often not
keen to look into conceptual models, nor are they ca-
pable of designing them on their own. The unsatis-
factory adoption in practice may be contributed to a
lack of respective professional education among to-
day’s workforce. However, I am afraid, we would
take the easy way out, if we were satisfied with this
explanation. There are other, more essential reasons
for questioning the power of current approaches to
conceptual modelling, which also may, in part, hin-
der its acceptance and dissemination in practice. They
relate to the economics of modelling and the psycho-
logical assumptions underlying the construction and
perception of conceptual models.
2.1 Economics of Modelling
With respect to economics of modelling, three as-
pects are of particular relevance. At first, there is
the productivity of creating, analysing and modify-
ing conceptual models, i.e. the time these activities
take for a certain outcome. While modelling produc-
tivity depends on modellers’ skills and experiences, it
is also affected by the available modelling languages
and tools. A modelling language can contribute to
productivity by providing reusable artefacts and by
allowing for abstractions that foster reuse and adapta-
tion of models. Currently, most modelling languages
are restricted to a few semantic primitives that remind
of basic ontologies like the one suggested by (Bunge,
1977) or (Grossmann, 1983). While generic concepts
such as “entity type”, “class”, “attribute”, etc. can
be applied to any domain—which is the purpose of
a general ontology—they require modellers to recon-
struct all concepts of a model from scratch. Hence,
they promote a wide range of reuse, which should im-
prove economies of scale e.g. of modelling tools, but
in a particular case, their contribution to productivity
is very poor. Imagine, you would have to describe a
domain such as accounting and you were restricted to
a language that consisted of generic concepts like the
ones provided by the ERM and the UML! Further-
more, modelling languages provide only a few ab-
straction concepts such as classification, generaliza-
Fourth International Symposium on Business Modeling and Software Design
10
tion or encapsulation. As a consequence, similarities
between a range of models can often not be accounted
for by abstracting on a set of common properties. The
lack of abstraction is especially painful in business
process modelling, where reuse is widely restricted
to copy&paste (Frank, 2012). Second, there is little
protection of investment. The use of conceptual mod-
els is widely restricted to the build time phase. In
later phases, models are used for documentation only,
if they are used at all. Even in those cases, where
models were used to generate code, they will usually
get devalued over time, because during maintenance
changes are directly applied to code and synchroniz-
ing models and code, beside representing a serious
challenge, does not happen. Third, the benefit of mod-
els depends on their quality. However, judging the
quality of conceptual models is a demanding task and
requires experts. Current modelling languages and
tools hardly contribute to model quality, since their
generic concepts allow for almost any kind of absurd
models as long as they are syntactically correct. From
a managerial perspective there is the additional prob-
lem that the economics of modelling is hard to judge.
While it is often not trivial to determine modelling
costs in advance, quantifying the benefits of models
ex ante is almost impossible. As long as managers
are not convinced that modelling is suited to generate
substantial benefit, the lack of legitimate quantifica-
tion methods creates a serious obstacle to conceptual
modelling in practice.
2.2 Cognitive Capabilities
Conceptual modelling is based on basic assumptions
that most members of the modelling community—
including myself—are convinced of. First, there is
the assumption that the analysis and design of com-
plex systems recommends a rational approach, i.e.
an approach that is characterized by the differenti-
ated consideration of pros and cons and that puts em-
phasis on justifying decisions. Second, related to the
first assumption, following a Kantian tradition, a ra-
tional perspective demands for focussing on concepts.
Conceptual modelling is typically aimed at the recon-
struction of terms used in the domain of interest. Re-
ferring to existing terminology is supposed to make
conceptual models accessible by people working in
the domain. At the same time, a reconstruction is
required in order to create concepts that fit the spe-
cific modelling purpose and that facilitate mapping
to implementation languages. As a consequence, it
it common to specify concepts according to the def-
initions of types or classes in implementation lan-
guages. While these basic ideas make perfect sense,
they are based on an idealization that is seriously chal-
lenged by research in cognitive psychology. There is
a plethora of work that shows the rather limited abil-
ity of most humans to precise and consistent think-
ing, for an overview see (Kahneman et al., 1982).
Furthermore, the way people acquire and use con-
cepts is often in clear contrast with the definitions we
use in conceptual modelling (Lakoff, 1990). Among
the most relevant insights are the following: Con-
cepts are often not associated with (extensional) defi-
nitions, but rather with a few typical examples (proto-
types”). Conceptual categories are often perceived as
having no clear boundaries (membership gradience”,
(Lakoff, 1990), p. 12). The last example is of es-
pecial relevance for our investigation: Most cognitive
models are embodied with respect to use. ((Lakoff,
1990), p. 12) As a consequence, we cannot expect
prospective users to be much interested in and capable
of thinking in concepts as we use them for modelling
purposes.
2.3 Preliminary Conclusion
Our brief analysis of the current state of conceptual
modelling resulted in a number of problems. That
does not mean, however, to question the idea of con-
ceptual modelling in general. In order to cope with
complexity and change there is no alternative for us
other than to somehow develop models. Without ab-
straction not only our understanding of systems and
of the world in general will remain poor, but so would
be our ability to design and implement systems. Nev-
ertheless, those problems should make us think. In
particular, there are three areas that demand for more
attention. There is need to promote modelling pro-
ductivity. For this purpose, reuse has to be fostered
by providing modelling constructs that incorporate
domain-specific semantics. To promote model qual-
ity, the reusable artefacts offered to users should be
thoroughly developed and tested. At the same time,
economies of scale are crucial. For this purpose, an
artefact needs to be reused in a wide range of cases,
which may create a conflict to modelling productiv-
ity. To improve the involvement of prospective users,
models, new forms of presenting concepts and mod-
els to users are required. Since there is evidence that
people imagine or learn concepts by associating them
with contexts of use, it might be advisable to develop
modelling environments that focus on the use of con-
cepts. The use of models during later phases of the
system life-cycle demands for developing scenarios
that illustrate the respective benefit. In addition, tech-
nical challenges, such as solving the notorious prob-
lem of synchronizing code and model have to be ad-
Power-Modelling - Toward a More Versatile Approach to Creating and Using Conceptual Models
11
dressed.
3 SELECTED APPROACHES
There have been various approaches that address
some of the problems elucidated above. Some of them
are aimed at improving modelling productivity and
model quality, while others focus on implementation
issues or on user involvement.
3.1 Focus on Productivity and Quality
Reference models are based on a convincing idea. By
developing models that serve as a reference for cer-
tain domains, the costs for realizing large models in
respective domains could be reduced substantially. At
the same time, reference models should be developed
with outstanding care and expertise. Therefore, they
should effectively improve model quality. Further-
more, reference models do not only stress a descrip-
tive intention, i.e. representing a domain as it is, but
also a prescriptive intention, i.e. representing concep-
tualisations that seem especially favourable for the fu-
ture development of organizations in the targeted do-
main. Despite their convincing foundation, the dis-
semination of reference models in practice remained
extremely modest. Domain-specific modelling lan-
guages (DSMLs) are build on a similar idea. Different
from general-purpose modelling languages (GPMLs),
they comprise domain-specific concepts, thus free-
ing modellers from the need to specify these con-
cepts from scratch. At the same time, DSMLs fea-
ture usually, but not necessarily, a concrete syntax
that is adapted to representations known in the do-
main they cover. In addition to promoting productiv-
ity through improved reuse and ergonomics, DSML
also contribute to model quality, since their syntax
and semantics facilitate preventing models that are all
too strange. Modelling is more convenient, and the
range of possible models is clearly restricted com-
pared to a GPML.
While DSMLs are certainly suited to address the
productivity challenge, they are not the silver bullet of
conceptual modelling, since they face a fundamental
conflict of system design. On the one hand, increasing
productivity demands for concepts that are specific to
a particular domain, i.e. that incorporate a high de-
gree of domain-specific semantics. On the other hand,
economies of scale demand for DSMLs that cover a
wider domain, i.e. that include a smaller degree of
domain-specific semantics. This fundamental conflict
of designing DSMLs is illustrated in fig. 1
Level of (domain-specific) Semantics
Potential Productivity Gain
Scale of Reuse
Figure 1: The fundamental conflict of designing DSMLs.
3.2 Focus on Implementation
If one regards a model mainly as an instrument for de-
veloping software, the transformation of models into
code is a major aspect of increasing the value of mod-
els. Some time ago, Wiederhold et al. proposed an ap-
proach, called “megaprogramming” to substantially
promote the productivity of programming large ap-
plication systems (Wiederhold et al., 1992). Even
though they do not explicitly speak of modelling,
there are clear relations to conceptual modelling in
their work. They suggest to compose large software
systems from “megamodules”. Megamodules can be
thought of as domain-specific abstraction: megamod-
ules “capture the functionality of services provided
by large organizations like banks, airline reservation
systems, and city transportation systems. ... The
concepts, terminology, and interpretation paradigm of
a megamodule is called its ontology. (Wiederhold
et al., 1992), p. 89. However, the representation of
megamodules as well as their composition happens on
the code level only. Therefore, they are hardly suited
for being used by non-experts. It is, however, con-
ceivable to combine the idea of megamodules with a
representation that corresponds more clearly to a ter-
minology users are familiar with. Even though meg-
amodules, like DSML, promote domain-specific arte-
facts, their use is based on a bottom-up approach in
the sense that a system is created from some kind
of building block without providing a blueprint for
the overall design. Application frameworks stress a
top-down approach. An application framework rep-
resents an architecture and the partial implementa-
tion of a class of application systems. “Black box”
frameworks can be adapted only through interfaces,
while “white box” frameworks allow for modifying
their code. Frameworks can substantially boost the
productivity of application development. However,
while the adaptability of black box frameworks is
Fourth International Symposium on Business Modeling and Software Design
12
limited, using white box frameworks efficiently re-
quires considerable effort. Similar to megamodules
the representation of frameworks is usually restricted
to code. Approaches to model-driven software de-
velopment take advantage of conceptual models for
designing systems. At the same time, they promote
implementation productivity by aiming at generating
software systems from models (France and Rumpe,
2007), (Stahl and V
¨
olter, 2006). To cope with a mul-
titude of platforms and programming languages, spe-
cific effort has been put on generating platform- and
language-independent representations that allow for
a straightforward transformation into particular im-
plementations (Mellor, 2004), (Pastor and Molina,
2007). Even though model-driven software develop-
ment is suited to improve development productivity
and software quality, it remains unsatisfactory with
respect to the evolution of software systems: Usually
it is not possible to generate an entire software system
from models, i.e. there is need for manual extensions.
As a consequence, the evolution of code and models
has to be synchronized. While there are a few ap-
proaches that address the challenge of synchronizing
models and code (e.g. (Balz et al., 2010), (Agrawal,
2003)), they are not satisfactory, because they cannot
always ensure consistency.
3.3 Focus on User Involvement
To some extent, the development of DSML and corre-
sponding editors is aimed at making modelling more
convenient for users who are not trained in concep-
tual modelling. The idea of using models at run time
(Blair et al., 2009), while also contributing to the
protection of investments, may be suited to motivate
more people to use models: as a conceptual represen-
tation of the systems they interact with and may want
to modify. However, recent research in this area is
mainly focused on software engineering aspects such
as synchronisation of models and systems, e.g. (Song
et al., 2011), or self-adapting systems, e.g. (Amoui
et al., 2012), (Morin et al., 2009). Krogstie proposes
using models during the entire life-cycle of a system
to emphasize user empowerment (Krogstie, 2007).
For this purpose, modelling should not longer be re-
stricted to system development. Instead, for mod-
elling to have a “larger effect”, he proposes “to en-
able all knowledge workers to be active modelers.
(p. 305). Enterprise software systems should be pre-
sented to their users as “interactive models”, the use
of which is “about discovering, externalizing, captur-
ing, expressing, representing, sharing and managing
enterprise knowledge. (p. 306). In addition to known
approaches to enterprise modelling, Krogstie stresses
the need for more intuitive representations of models,
such as “visual scenes for pro-action learning” and
descriptions of the relevant context that focus on ac-
tions (p. 308). Using models as objects and objectiva-
tion of organizational knowledge work and of individ-
ual learning is appealing. However, Krogstie remains
vague about the realization of his vision. He suggests
a “model-generated workplace (MGWP)” that “is a
working environment for the business users involved
in running the business operations” (p. 312), but he
does not provide details of how to accomplish “inte-
grated modelling and execution platforms” (p. 308).
4 OUTLINE OF
POWER-MODELLING
The idea presented in this section was inspired by var-
ious streams of work, some of which are described
above. It is also a result of our long-standing-work
on enterprise modelling that confronted us with some
serious problem we were not able to solve as long
as we were still bound to the traditional principles of
conceptual modelling and of implementing modelling
tools. Hence, the outline of power-modelling is also
an attempt to suggest a new paradigm of creating and
using conceptual models.
4.1 Objectives and Challenges
Against the background of our previous discussion,
the following objectives mark desirable features of a
future conception and realization of modelling.
Objective O1: A powerful approach to conceptual
modelling should enable the use of models during the
entire life-cycle of a system. Rationale: The com-
plexity of enterprise systems demand for a represen-
tation that users are able to understand, not only to ob-
tain a better comprehension of a software system, but
also of the the context, since an increasing part of an
enterprise is represented by its software systems. At
the same time, people need (cognitive) models any-
way to make sense of software systems and of organi-
zations. Therefore, explicit models that fit the cogni-
tive capabilities of users should be suited to increase
organizational transparency. Objective O2: Concep-
tual models that represent software systems and the
context they operate in should be interactive. Ratio-
nale: In order to serve as a universal interface to en-
terprise systems that fosters user empowerment, mod-
els need to allow for navigation, for searching and for
modifying their states—not only during build time,
but during run time. Objective O3: Conceptual mod-
els should provide access to their conceptual foun-
Power-Modelling - Toward a More Versatile Approach to Creating and Using Conceptual Models
13
dation, i.e. to the modelling language they are de-
fined with, i.e. respective modelling tools should be
self-reflective. Rationale: Users may want to get a
better understanding of the concepts the models they
interact with are based on. Furthermore, advanced
users may even want to change those concepts. Ob-
jective O4: Conceptual models should be constructed
from concepts that are used in professional discourse
in the relevant domain. Rationale: Domain-specific
concepts promote modelling productivity and make
the use of modelling concepts more intuitive. Objec-
tive O5: Conceptual models of an enterprise should
cover multiple perspectives and foster their integra-
tion. Rationale: The complexity of many organiza-
tions goes along with specialization which in turn re-
sults in different professional perspectives and lan-
guages. In order to satisfy the demand for provid-
ing users with representations they are familiar with,
models of an enterprise need to account for different
perspectives. On the one hand, that relates to pro-
viding modelling concepts which correspond to the
technical language that is characteristic for a certain
perspective. On the other hand, it should also be
possible to present a model using different notations,
both graphical and textual, in order to satisfy differ-
ent cognitive styles. Objective O6: To increase the
economics of modelling economies of scale have to
be increased by promoting reuse. Rationale: Today,
the effort it takes to develop elaborate models is of-
ten still prohibitively high. The above objectives cre-
ate considerable challenges. Among those, three are
hard to overcome within the current paradigm. Aim-
ing at both, modelling concepts that reflect particu-
lar, domain-specific technical languages, and that pro-
mote economies of scale faces an obvious conflict:
A language can either be more domain-oriented or
built for serving more general purpose. Using mod-
els at run time that allow interacting with and eventu-
ally changing a software system demands a tight in-
tegration of models and code in order to keep their
changes in synch. However, this is almost impossi-
ble due to limitations of prevalent programming lan-
guages. The elements of a model on M
1
are repre-
sented as objects on M
0
, even though they are con-
ceptually located on M
1
. This is the case, too, for
the representation of metaclasses in meta model edi-
tors. As a consequence, there is need to generate code
(objects on M
0
cannot be further instantiated), which
creates the notorious problem of synchronizing mod-
els and code. The lack of abstraction concepts in pro-
cess modelling languages creates a serious obstacle
for reuse. While generalisation/specialisation may be
regarded as a suitable approach, it cannot be applied
to processes in a straightforward way: To satisfy the
substitutability constraint (Liskov and Wing, 1994),
specialisation has to be monotonic, which is impossi-
ble to achieve for process models (Frank, 2012)
4.2 Cornerstones of Power-Modelling
For the vision of power-modelling to become real, the
current paradigm is not sufficient. There is need to
change our perspective on conceptual modelling and
to aim at a different linguistic foundation—both of
modelling and implementation languages. The fol-
lowing aspects mark cornerstones of a conception of
power-modelling.
Emphasis on DSML: DSMLs that are recon-
structed from existing technical terminologies (objec-
tive O4) promise clear advantages with respect to pro-
ductivity, model quality and comprehensibility. For
this purpose, we can build on an existing integrated
set of DSMLs for enterprise modelling (Frank, 2013).
Multi-Level Modelling: In order to overcome the
fundamental conflict of designing DSMLs (objectives
O4 and O6), a multi-level language architecture is
proposes. It is inspired by the definition and refine-
ment of technical languages in practice. On a more
generic, textbook” level, concepts are introduced that
are supposed to fit a wide range of more specific do-
mains. For that purpose, they remain intentionally
abstract and underspecified. On more specific lev-
els that could cover, for example, particular indus-
tries, concepts are refined and added. This process
of stepwise refinement may in the end lead to nu-
merous language levels, from more generic, over re-
gional” to local” DSMLs. Multi-level modelling de-
scribes a language architecture that allows for an ar-
bitrary number of classification levels. This allows
to achieve both, a wide range of reuse, i.e. benefi-
cial economies of scale, on higher levels, and a high
productivity that is enabled by more specific DSML
that reuse concepts of more generic DSML. Since all
DSMLs are integrated in one language architectures,
users can navigate all classification levels they are in-
terested in (objective O3). Fig. 2 illustrates a multi-
level language architecture and corresponding editors.
For a detailed description of multi-level modelling see
(Frank, 2014b).
Common representation of models and code: Us-
ing models during the entire life-cycle of a systems
create the challenge of synchronizing models and
code, which is caused by the fact that current imple-
mentation languages allow for one or two classifica-
tion levels only. There are, however, a few languages,
which are based on the recursive golden braid” archi-
tecture that facilitate systems with more classification
levels. Among these languages, XMF (Clark et al.,
Fourth International Symposium on Business Modeling and Software Design
14
Reference DSML
Language Designer
Organisational Unit
Position
Specific DSML
(Local „Dialect“)
Organisation Analyst
Department
Team Market Analyst
Particular Organisation
Model
Manager
Marketing Department
Quality Circle
Product Group PG 1
Market
Analyst MA2
Market Research
Team
Committee
Quality Circle
Meta Modeling Language
Meta Language Designer
MetaEntity
Association
MetaAttribute
Metamodel Editor
Organisation Schema Editor
Specific Organisation Editor
X
specifies
X
specifies
X
specifies
foundation of X
creates X
creates X
Figure 2: Illustration of multi-level language architecture.
2008b), (Clark et al., 2008a) is particularly suited for
supporting power-modelling. First, it enables an ar-
bitrary number of classification levels. Second, it
is accompanied by a (meta) modelling environment,
Xmodeler, that features a common representation of
models and code, hence, it eliminates the need for
synchronizing models and code and facilitates the use
of models during the entire life-cycle of a system.
In order to develop a suitable foundation of multi-
level modelling and multi-level enterprise systems,
we modified the metamodel of XMF (Frank, 2014b).
By overcoming the separation of model and code,
every system can be seen as a collection of models
which can be used interactively (objective O2) with
multiple representations that can be exchanged based
on an implementation of the model-view-controller
pattern (for details see (Clark et al., 2008a)).
Wider conception of conceptual modelling: In the
existing paradigm, a conceptual model is created from
the concepts of a modelling language, typically ar-
ranged in diagrams that represent some kind of a
graph. However, modelling can also be thought of
as an act of configuration that makes use of existing,
more generic models and of DSML. In that case, mod-
elling is more interactive, where a respective mod-
elling system suggests alternative modelling options
and requests user preferences. Since many users make
sense out of action context rather than of singular con-
cepts, interactive modelling should also go along with
the representation of the context a model is supposed
to address. A prototypical context could be provided
by an enterprise model that integrates various per-
spectives on the enterprise (Frank, 2013) (objective
O5). Furthermore, such a wider conception would in-
clude to not restrict modelling to drawing diagrams.
Instead, models would be accessible through the ele-
ments of a graphical user interface, through tables or
even through plain text. While it makes sense to re-
Power-Modelling - Toward a More Versatile Approach to Creating and Using Conceptual Models
15
Figure 3: Definition of a DSML for modelling organizational structures of universities.
strict models to types or classes, a wider conception
of modelling would also include models that comprise
objects on M
0
using similar (graphical) representa-
tions to enable their interactive use. For example:
The representation of a business process type could
be supplemented by similar representations of busi-
ness process instances (for an example in the context
of so called “self-referential enterprise systems” see
(Frank and Strecker, 2009), p. 14).
Ubiquitous use of (multi-perspective) models:
Models are proposed as the primary medium to per-
ceive, interact with and change systems—and the en-
vironment they are supposed to operate in during the
entire system life-cycle. To serve this purpose, mod-
els have to be prepared for various perspectives to sat-
isfy different skills and needs. Then they would no
longer be a tool for system analysts and designers,
but for everybody acting responsibly in an environ-
ment that is penetrated by information systems. As
a consequence, power-modelling would contribute to
empowering users by giving them (guided) access to
a system’s conceptual foundation and by supporting
communication with other stakeholders.
4.3 Illustration
The following scenarios highlight selected aspects of
power-modelling.
Refinement of DSML: Developing a DSML for
modeling organizational structures may appear as a
fairly trivial undertaking. However, this is not the
case. First, a remarkable concept variance has to be
accounted for. For example: A term like “depart-
ment” may represent clearly different kinds if orga-
nizational units in different environments. In a uni-
versity, a department consists of institutes and chairs.
In some industrial enterprises, a department is part of
a head department, in others not. Applying the idea
of multi-level modelling to this domain would sug-
gest developing a reference DSML for organization
modelling. Such a reference DSML would include
a descriptive graphical notation. A respective edi-
tor could be used by organization analysts to create
an organizational schema for a certain domain, e.g.
a particular organization or a range of organizations
of a certain kind. Finally, this more specific DMSL
would be used by local managers to build a particular
organizational structure that corresponds to the previ-
ously defined schema. Fig. 3 illustrates the definition
of a DSML for modelling organizational structures of
universities by using a more general DSML that rep-
resents a textbook-level terminology.
The resulting DSML could then be used for cre-
ating an editor that facilitates modelling the organiza-
tional structure of a particular university. The result-
ing model is located on M
0
. Even though it does not
satisfy the abstraction usually demanded for by con-
ceptual modelling, it is still beneficial to use a graphi-
cal notation that corresponds to that of the DSML de-
fined on M
1
. Such a model would enable interactive
access to particular organizational units—and also al-
low for navigating to the specifications of respective
concepts on higher classification levels.
Modelling as Interactive Configuration: Instead
Fourth International Symposium on Business Modeling and Software Design
16
CO 1
CO 2
CO 3
CR 1
CR 2
CR 3
S 1
S 2
Figure 4: Configuration through stepwise selection of intended model properties.
of creating a model bottom-up from the concepts pro-
vided by a modelling language, the following sce-
nario shows how a model can be created by combin-
ing reference models, application frameworks, multi-
level modelling and a common representation of mod-
els and code to not only create models more conve-
niently but to realize a corresponding implementation
at the same time. The scenario is based on the exis-
tence of a repository of various reference models that
were created with a set of integrated DSMLs. It is
aimed a modelling and thereby implementing an or-
der management system for food retailing. Fig. 4 il-
lustrates the use of a tool that facilitates the config-
uration of particular models from an existing set of
reference models. At first, the user would specify the
targeted domain by selecting options from lists pre-
sented by the modelling tool. Based on that selection,
a set of possibly fitting business process types would
be presented to the user. After selecting one business
process type, the user could refine the process model
by selecting from properties that are offered by the
system. In case, the control flow requires further mod-
ification, the user could modify it using a respective
DSML for business process modelling.
5 CONCLUSIONS AND FUTURE
RESEARCH
The presented outline of a new approach to creating
and using conceptual models has a twofold motiva-
tion. On the one hand, it stresses the need for lean-
ing back once in a while and questioning assumptions
that we tend to take for granted. On the other hand,
it presents the cornerstones of a more versatile and
powerful approach to conceptual modelling. The pre-
sentation of the vision is focussed on illustrating the
idea in order to inspire a discussion about its benefits
and about potential enhancements. Nevertheless, the
required foundation, especially a multi-level language
architecture and a common representation of models
and code, is available. It has been developed during
the recent years (Frank, 2014b), (Frank and Strecker,
2009) and builds on a mature meta-programming lan-
guage (Clark et al., 2008b). While it required to give
up certain assumptions that we had taken for granted,
such as the rigid dichotomy of instantiation and spe-
cialization, it opens new ways of designing and imple-
menting systems that are represented to users as mod-
els on different levels of abstraction. This paradigm
shift requires not only giving up “standard” language
architectures like MOF (Object Management Group,
2006), but also replacing existing implementation lan-
guages. Therefore, it may be regarded by some as not
realistic. However, if we, at least in academia, do not
Power-Modelling - Toward a More Versatile Approach to Creating and Using Conceptual Models
17
give up the widespread fixation on mainstream tech-
nology, progress will hardly be possible. There is still
a long way to go. Future research needs to address
the development of reference models on different lev-
els of abstraction that are specified with respective
DSML. There is also need to develop advanced pat-
terns of interaction that support the configuration and
modification of models. Last but not least, it is re-
quired to overcome the limitations of current imple-
mentation languages by moving to more flexible lan-
guages architecture like the one XMF is based on.
REFERENCES
Agrawal, A. (2003). Metamodel based model transfor-
mation language to facilitate domain specific model
driven architecture. In Crocker, R. and Steele, G.
L. J., editors, Companion of the 18th Annual ACM
SIGPLAN Conference on Object-Oriented Program-
ming, Systems, Languages, and Applications (OOP-
SLA 2003), pages 118–119, New York. ACM.
Amoui, M., Derakhshanmanesh, M., Ebert, J., and Tahvil-
dari, L. (2012). Achieving dynamic adaptation via
management and interpretation of runtime models.
Journal of Systems and Software, 85(12):2720–2737.
Atkinson, C. and K
¨
uhne, T. (2003). Model-driven devel-
opment: a metamodeling foundation. IEEE Software,
20(5):36–41.
Balz, M., Striewe, M., and Goedicke, M. (2010). Continu-
ous maintenance of multiple abstraction levels in pro-
gram code. In Proceedings of the 2nd International
Workshop on Future Trends of Model-Driven Devel-
opment (FTMDD 2010).
Blair, G., Bencomo, N., and France, R. B. (2009). Models@
run.time: Computer. Computer, 42(10):22–27.
Buckl, S., Matthes, F., Roth, S., Schulz, C., and Schweda,
C. (2010). A conceptual framework for enterprise ar-
chitecture design. In Proper, E., Lankhorst, M. M.,
Sch
¨
onherr, M., Barjis, J., and Overbeek, S., edi-
tors, Trends in Enterprise Architecture Research, vol-
ume 70 of Lecture Notes in Business Information Pro-
cessing, pages 44–56. Springer, Berlin and Heidelberg
and New York.
Bunge, M. (1977). Treatise on Basic Philosophy: Volume
3: Ontology I: The Furniture of the World. Reidel,
Dordrecht.
Clark, T., Sammut, P., and Willans, J. (2008a). Applied
metamodelling: a foundation for language driven de-
velopment.
Clark, T., Sammut, P., and Willans, J. (2008b). Superlan-
guages: developing languages and applications with
XMF. Ceteva.
Ferstl, O. K. and Sinz, E. J. (2006). Modeling of business
systems using som. In Bernus, P., Mertins, K., and
Schmidt, G., editors, Handbook on Architectures of
Information Systems, International Handbooks on In-
formation Systems, pages 347–367. Springer, Berlin
and Heidelberg and New York.
France, R. B. and Rumpe, B. (2007). Model-driven devel-
opment of complex software: A research roadmap. In
Briand, L. C. and Wolf, A. L., editors, Workshop on
the Future of Software Engineering (FOSE ’07), pages
37–54. IEEE CS Press.
Frank, U. (2012). Specialisation in business process mod-
elling: Motivation, approaches and limitations. icb re-
search report, no. 51, university of duisburg-essen.
Frank, U. (2013). Multi-perspective enterprise modeling:
Foundational concepts, prospects and future research
challenges. online first: http://link.springer.com/
article/10.1007/s10270-012-0273-9. Software and
Systems Modeling.
Frank, U. (2014a). Enterprise modelling: The next steps.
Enterprise Modelling and Information Systems Archi-
tectures, 9(1):24–40.
Frank, U. (2014b). Multilevel modeling: Toward a new
paradigm of conceptual modeling and information
systems design. Business and Information Systems
Engineering, 6:accepted for publication.
Frank, U. and Strecker, S. (2009). Beyond erp systems: An
outline of self-referential enterprise systems: Require-
ments, conceptual foundation and design options. icb
research report, no. 31, university of duisburg-essen.
Grossmann, R. (1983). The Categorical Structure of the
World. Indiana University Press, Bloomington.
Kahneman, D., Slovic, P., and Tversky, A., editors (1982).
Judgment under uncertainty: Heuristics and biases.
Cambridge University Press, Cambridge and New
York.
Krogstie, J. (2007). Modelling of the people, by the peo-
ple, for the people. In Krogstie, J., Opdahl, A.,
and Brinkkemper, S., editors, Conceptual Modelling
in Information Systems Engineering, pages 305–318.
Springer Berlin Heidelberg.
Lakoff, G. (1990). Women, fire and dangerous things: What
categories reveal about the mind. Univ. of Chicago
Press, Chicago, 1 edition.
Lankhorst, M. M. (2005). Enterprise architecture at work:
Modelling, communication, and analysis. Springer,
Berlin and Heidelberg and New York.
Liskov, B. H. and Wing, J. M. (1994). A behavioral no-
tion of subtyping. ACM Transactions on Program-
ming Languages and Systems, 16:1811–1841.
Mellor, S. J. (2004). MDA distilled: Principles of model-
driven architecture. Addison-Wesley object technol-
ogy series. Addison-Wesley, Boston.
Morin, B., Barais, O., J
´
ez
´
equel, J.-M., Fleurey, F., and Sol-
berg, A. (2009). Models@run.time to support dy-
namic adaptation. IEEE Computer, 42(10):46–53.
Object Management Group (2006). Meta object facility
(mof) core specification: Version 2.0.
Pastor, O. and Molina, J. C. (2007). Model-Driven Architec-
ture in Practice: A Software Production Environment
Based on Conceptual Modeling. Springer, Berlin and
New York.
Proper, E., Lankhorst, M. M., Sch
¨
onherr, M., Barjis, J., and
Overbeek, S., editors (2010). Trends in Enterprise
Architecture Research: 5th Workshop, TEAR 2010,
Fourth International Symposium on Business Modeling and Software Design
18
Delft, The Netherlands, November 12, 2010, Proceed-
ings, volume 70 of Lecture Notes in Business Infor-
mation Processing. Springer, Berlin and Heidelberg
and New York.
Scheer, A.-W. (1992). Architecture of Integrated Informa-
tion Systems: Foundations of Enterprise Modelling.
Springer, Berlin and New York.
Song, H., Huang, G., Chauvel, F., Xiong, Y., Hu, Z.,
Sun, Y., and Mei, H. (2011). Supporting runtime
software architecture: A bidirectional-transformation-
based approach. Journal of Systems and Software,
84(5):711–723.
Stahl, T. and V
¨
olter, M. (2006). Model-driven software
development: Technology, engineering, management.
John Wiley, Chichester and England and Hoboken and
NJ.
Wiederhold, G., Wegner, P., and Ceri, S. (1992). Toward
megaprogramming. Commun. ACM, 35(11):89–99.
Power-Modelling - Toward a More Versatile Approach to Creating and Using Conceptual Models
19