Challenges of Modelling Landscapes

Pragmatics Swept Under the Carpet?

Marija Bjekovi

c and Henderik A. Proper

Public Research Centre Henri Tudor, Luxembourg, Luxembourg

Radboud University Nijmegen, Nijmegen, The Netherlands

EE-Team

, Luxembourg, Luxembourg

{marija.bjekovic, erik.proper}@tudor.lu

Keywords:

Model, Modelling Language, Model Integration.

Abstract:

In enterprise modelling, a wide range of models and languages is used to support different purposes. If

left uncontrolled, this can easily result in a fragmented perspective on the enterprise, its processes and IT

support. On its turn, this negatively affects traceability, the ability to do crosscutting analysis, and the overall

coherence of models. Different strategies are suggested to achieve model integration. They mainly address

syntactic-semantics aspects of models/languages, and only to a limited extent their pragmatics. In actual use,

the ‘standardising’ and ‘integrating’ effects of traditional approaches (e.g. UML, ArchiMate) erodes. This is

typically manifested by the emergence of local ‘dialects’, ‘light weight versions’, as well as extensions of the

standard to cover ‘missing aspects’. This paper aims to create more awareness of the factors that are at play

when creating integrated modelling landscapes. Relying on our ongoing research, we develop a fundamental

understanding of the driving forces and challenges related to modelling and linguitic variety within modelling

landscapes. In particular, the paper discusses the effect of a priori ﬁxed languages in modelling and model

integration efforts, and argues that they bring about the risk of neglecting the pragmatic richness needed across

practical modelling situations.

1 INTRODUCTION

Enterprise models play an important role in the design

and operations of enterprises (Bubenko et al., 2010).

More speciﬁcally, enterprise models can be used to

study the current state of an enterprise, analyse prob-

lems with regard to the current situation, sketch po-

tential future scenarios, design future states of the

enterprise, communicate with stakeholders, manage

change, etc. ((Davies et al., 2006), (Bubenko et al.,

2010), (Anaby-Tavor et al., 2010)).

Next to the fact that enterprise models are cre-

ated for different purposes, it is necessary to do so

from different perspectives, such as business pro-

cesses, value exchanges, products and services, in-

formation systems, etc. In the ﬁeld of information

systems engineering, the use of a multi-perspective

approach has long since been advocated, e.g. (Wood–

Harper et al., 1985), (Zachman, 1987). For enter-

The Enterprise Engineering Team (EE-Team) is a col-

laboration between Public Research Centre Henri Tudor,

Radboud University Nijmegen and HAN University of Ap-

plied Sciences (www.ee-team.eu).

prise modelling, there is even a broader set of per-

spectives to consider ((Frank, 2002), (Winter and Fis-

cher, 2007), (Greefhorst et al., 2006), (Wagter et al.,

2012)).

The collection of models that jointly represent

the different perspectives of one enterprise, are of-

ten expressed using different modelling languages, in-

cluding UML (OMG, 2003), BPMN (OMG, 2008),

ArchiMate (Iacob et al., 2012), i* (Yu and Mylopou-

los, 1996), e3Value (Gordijn and Akkermans, 2003),

SBVR (OMG, 2006), etc. Throughout this paper, we

will use the term enterprise modelling landscape, or

simply modelling landscape, to refer to the variety of

models and corresponding modelling languages used

in a speciﬁc enterprise modelling effort

Since the models included in an enterprise mod-

elling landscape provide different views on the same

enterprise, it is quite natural to expect that the sets of

models form a coherent whole; i.e. linked where rele-

vant and consistent as a whole. A plethora of models

The scope of a particular enterprise modelling effort

can be only one project, cross-project considerations, entire

enterprise etc.

BjekoviÄ

G M. and A. Proper H.

Challenges of Modelling LandscapesPragmatics Swept Under the Carpet?.

DOI: 10.5220/0004773600110022

In Proceedings of the Third International Symposium on Business Modeling and Software Design (BMSD 2013), pages 11-22

ISBN: 978-989-8565-56-3

and modelling languages may easily result in a frag-

mented perspectives on the enterprise, which is likely

to have a negative impact on the traceability across

different models, the ability to do crosscutting anal-

ysis, to manage inconsistency

and to ensure overall

coherence of e.g. the design of the enterprise. The

fact that different models are usually expressed in dif-

ferent modelling languages makes it even harder to

maintain the coherence.

The traditional approach of dealing with fragmen-

tation in modelling landscapes is to create an inte-

grated/uniﬁed modelling language, such as UML and

ArchiMate. However, in actual use, one can observe

how the ‘standardising’ and ‘integrating’ effect of

such languages erodes. This typically manifests itself

in terms of local ‘dialects’, ‘light weight versions’, or

several extensions of an existing standard that are in-

tended to deal with ‘missing aspects’. This point is il-

lustrated by the advent of domain or purpose-speciﬁc

(modelling) languages that allow for the creation of

models that are tuned to the needs of speciﬁc domains

or purposes. At the same time, there exists a number

of approaches that aim to alleviate this fragmenting

effect by assuring the links between the different lan-

guages deﬁnitions, see e.g. ((Frank, 2002), (Vernadat,

2002), (Anaya et al., 2010)). Typically, these links

are deﬁned based on the standardised deﬁnition of the

language, in particular its semantics (as in e.g. (Anaya

et al., 2010)).

As we will discuss throughout this paper, the

drivers of language standardisation are predominantly

of technical nature (Hoppenbrouwers, 2003). There

are clear beneﬁts of language standardisation. For in-

stance, it is generally considered as a necessary con-

dition for CASE tool development. In addition, it is

a ﬁrst step towards automating some model manip-

ulations, e.g. model transformations (including code

generation). Nonetheless, the potential beneﬁts of

standardised languages tend to be quickly generalised

to the entire modelling endeavour. What goes practi-

cally unquestioned, in aiming for language standardi-

sation, is whether ﬁxed languages can be used at all in

different modelling contexts and with different stake-

holder groups. For instance, in ((Kaidalova et al.,

2012), (Bubenko et al., 2010)) it is observed that the

choice of formalism should be related the given mod-

elling task and audience. For example, when the lan-

guage chosen is rather too formal for stakeholders, it

can hinder the modelling process.

We will argue, that standardising/ﬁxing a mod-

elling language leads to a situation in which the

pragmatic richness that is needed across various

Allow inconsistencies between models, e.g. due to dif-

ferent views by differing stakeholders, in a controlled way.

modelling situations in practice is neglected. This

brings about the risk of sweeping pragmatics un-

der the carpet. Indeed, various surveys ((Malavolta

et al., 2012), (Kaidalova et al., 2012)) and empir-

ical studies ((Anaby-Tavor et al., 2010), (Karlsen,

2011), (Briand et al., 1995), (Elahi et al., 2008)) re-

porting on practical experiences with enterprise mod-

elling, point at the need for ﬂexibility in modelling.

At the same time they observe a lack in ﬂexibil-

ity of tools and the underlying (ﬁxed) languages

to aptly ﬁt the needs of speciﬁc modelling situa-

tions. This often leads to different levels of disci-

pline in which the standard language is obeyed to,

e.g. resulting in dialect-like variations of the orig-

inal language ((Bubenko et al., 2010), (Malavolta

et al., 2012), (Briand et al., 1995), (Elahi et al.,

2008), (Karlsen, 2011)), or workarounds (e.g. us-

ing ad hoc notes and annotations) to compensate

for the missing elements in the language/tool (De-

len et al., 2005). This may even go as far as

the use of home-grown, organisation-speciﬁc semi-

structured models/languages instead of the standard

notation ((Anaby-Tavor et al., 2010), (Malavolta

et al., 2012), (Karlsen, 2011)).

In our view, this indicates a lack of fundamental

understanding of the role of language in modelling,

and more speciﬁcally, the place of ﬁxed language in

attempts to integrate models. In this paper, based on

our ongoing research, we intend to shed more light on

this topic. We will therefore start in Section 2 with a

discussion on the potential beneﬁts of standardising

modelling languages. In Sections 3 and 4 we then

explore the effects of their use in modelling and inte-

gration respectively, also identifying more explicitly

the risk of sweeping pragmatics under the carpet of

the modelling landscape. We then continue in Sec-

tion 5 with a fundamental discussion of models and

modelling. This understanding is used then to discuss

in Section 6 the role of modelling languages, and their

potential standardisation. In the conclusion, we syn-

thesise the insights and suggest the direction to ex-

plore more realistic strategies for creating and man-

aging modelling landscapes.

2 STANDARDISATION

In our ﬁeld, we typically deal with linguistic mod-

els (Karagiannis and H

offerer, 2006), i.e. models ex-

pressed in a modelling language. A further distinc-

tion can be made between textual and graphical lan-

guages (Harel and Rumpe, 2004). Given their signiﬁ-

cant usage in enterprise modelling efforts, our discus-

sion focusses on graphical languages. Traditionally, a

Third International Symposium on Business Modeling and Software Design

modelling language is deﬁned in terms of an abstract

syntax, a concrete syntax and semantics.

The abstract syntax deﬁnes the basic elements and

rules for creating models. The abstract syntax of

graphical modelling language is commonly given by

means of the meta-model. The meta-model actually

represents the conceptual foundation of the modelling

language, i.e. a speciﬁc classiﬁcation of concepts to

be used in discourse about the ‘world’ (Falkenberg

et al., 1998). As such, the meta-model provides a par-

ticular ontological position ﬁltering the view on the

‘world’ that one chooses to take (Falkenberg et al.,

1998). It is also argued in (Falkenberg et al., 1998)

that all other aspects of the modelling language de-

pend on the concepts contained in the meta-model.

The concrete syntax or notation deals with the (vi-

sual) representation of the modelling language on the

medium, by deﬁning the visual symbols and rules for

their combination (and their correspondence to the ab-

stract syntax of the language. The medium itself can

for example be restricted to a speciﬁc form, such as

graphical, textual, or video, but the notation in gen-

eral can also be restricted in terms of fonts, icons and

layout rules. See e.g. (Moody, 2009).

The semantics deals with the meaning of a mod-

elling language. The conventional way of deﬁning

semantics is in terms of a semantic domain and a se-

mantic mapping (Harel and Rumpe, 2004). Accord-

ing to (Harel and Rumpe, 2004), the semantic do-

main captures the “decisions about the kinds of things

language should express” (Harel and Rumpe, 2004,

p. 68). The semantic mapping, in turn, establishes

the correspondence from syntactic elements to the se-

mantic domain. However, this approach to deﬁning

semantics is required for the mechanical manipula-

tion of models, e.g. by computer tools, since they can

only manipulate semantics in terms of syntactic rep-

resentations (Harel and Rumpe, 2004). We propose

to label this aspect of semantics as syntactic seman-

tics. In the realm of human use of language, meaning

is approached differently by taking into account the

entire context in which the linguistic communication

is embedded and the function of linguistic utterances

in that context

. We propose to label this aspect of

semantics as pragmatic semantics. When stakehold-

ers use modelling language in modelling, they address

the semantics from this perspective. We will discuss

this topic further in Section 6.

As illustrated in Figure 1, the ﬁxed i.e. standard-

ised deﬁnition of the modelling language a priori

identiﬁes and restricts the intended sets of models the

This way of addressing meaning is inherent to the func-

tional perspective or action tradition on language. We will

elaborate this in the Section 6

ML Pragmatics

MODELLING

LANGUAGE

SPECIFICATION

INTENDED

SITUATIONS

OF USE

ACTUAL

SITUATIONS

OF USE

INTENDED

MODEL SETS

ACTUAL

MODEL SETS

Figure 1: Intended and actual use of the modelling lan-

guage.

language allows to express. This also limits the sets

of intended modelling situations in which, by using

a particular ﬁxed language, a satisfactory model of

the domain can be produced. Therefore, when freez-

ing languages (Hoppenbrouwers, 2003), the design-

ers of the language, implicitly or explicitly, restrict

the intended use of a language. The more models a

language can ‘produce’, the more expressive the lan-

guage is. However, the actual suitability of the lan-

guage is dependent on the particular modelling situa-

tion in which language is used. It refers to whether

the modelling language allows to create the model

of the domain such that it satisﬁes the needs of ac-

tual modelling situation (e.g. the level of detail in do-

main, coverage of speciﬁc aspects, speciﬁc form etc.).

This is the area of modelling pragmatics. According

to (Thalheim, 2012), it studies the use of languages

in a particular modelling situation depending on the

purposes and goals of models within a community of

practice.

In most cases, graphical modelling languages are

deﬁned semi-formally, i.e. with explicit (and more or

less formal) deﬁnitions of the (abstract and concrete)

syntax. The conceptual foundation of the language

may be deﬁned at different levels of genericity (i.e.

involving more or less generic concepts). This inﬂu-

ences the intended model sets supported by the lan-

guage. Also, different syntactic restrictions may be

included in the language deﬁnition, further restricting

intended model sets. The semantics is, however, usu-

ally given in an informal manner in the language spec-

iﬁcation, i.e. using natural language. The latter does

not lend itself that easily to machine interpretation.

Formally deﬁned semantics is required for making the

language speciﬁcation (fully) machine readable. The

standard and precise deﬁnition of modelling language

syntax and semantics is indeed pre-requisite for au-

tomation. This makes possible e.g. model transforma-

tions, interoperability, computer-aided analysis tech-

niques, simulation, and (developing tools for) various

other manipulations of models. These are some clear

beneﬁts of ﬁxed languages, and one possible strategy

Challenges of Modelling Landscapes - Pragmatics Swept Under the Carpet?

to ensure a return on modelling effort.

The predominant factor for a priori ﬁxing the

modelling language is therefore the technology (Hop-

penbrouwers, 2003), i.e. the fact that mechanical ma-

nipulation of models requires ﬁxed representations. It

is also assumed that by having standard (and precise)

deﬁnition of the modelling language, all the meta-

discussions on concepts can be avoided. This would

contribute to the certainty and efﬁciency of commu-

nication (Hoppenbrouwers, 2003), and shared under-

standing of models would be easier to reach.

However, as the next two Sections will discuss,

standardisation comes at a cost, in particular for prag-

matics, which ends up being swept under the carpet.

3 USE OF FIXED LANGUAGES

A key problem in the use of ﬁxed languages in mod-

elling is rooted in the lack of suitability of a language

for an actual modelling situation. It is indeed often

the case that the choice of the modelling language is

imposed from the “outside” onto the modelling situa-

tion. Typically, the existing modelling infrastructures

within the enterprise, the expertise of the modelling

team, etc. constrain the choice of the language.

As reported in several surveys on the practice of

modelling, see e.g. (Davies et al., 2006) (Anaby-

Tavor et al., 2010) (Malavolta et al., 2012), general-

purpose modelling languages are the most widely

used modelling languages. Nonetheless, these sur-

veys also indicate that ‘variants’ of these languages

are in place. For instance, several experience reports

of the use of i* in speciﬁc situations (Briand et al.,

1995) (Elahi et al., 2008) explain in detail why and

how the language was extended to be able to make

models that satisfy the needs of the given modelling

situation. In the case of ArchiMate, this has e.g.

resulted in the suggestion to distinguish between a

‘sketching’ and a ‘designing’ (Lankhorst et al., 2005)

variation of the notation (using more sketchy lines and

more informal looking fonts). This variation can even

be combined in one model to differentiate between

the status of different parts of the model. On the same

line, (Malavolta et al., 2012) indicate the need for in-

formal ‘variants’ of (software and enterprise) archi-

tecture models for their communication to different

stakeholders.

In our view, these ‘variants’ are essentially

purpose-speciﬁc variations of the same original

generic modelling language, differing only in their

syntactic and semantic restrictions, i.e. purpose-

speciﬁc modelling languages (Bjekovi

c et al., 2012).

They emerge from the need to make the language suit-

able for the communicative task in the actual mod-

elling situation at hand. When the modelling lan-

guage used is not suitable enough, variations will

emerge to compensate for this lack of suitability.

Pragmatics re-emerging from under the carpet.

An extreme case of adapting the language to the

actual modelling situation is the use of ‘home-grown’

notations (Anaby-Tavor et al., 2010) and/or emergent

modelling languages, i.e. the languages that are be-

ing constructed along the modelling process. For

instance, in (Anaby-Tavor et al., 2010) business ar-

chitects express a clear preference for home-grown,

semi-structured models, since they offer ﬂexibility in

terms of re-factoring, delayed commitment to syn-

tax, and closer ﬁt to the inherent way of thinking in

these phases. These semi-structured models emerged

through the repeated use in similar modelling situa-

tions, whereby the sets of concepts and their mean-

ings, and (right level of) restrictions gradually yielded

a new language.

Collaborative modelling situations also demon-

strate the challenge of adequate modelling support.

For instance, in situations whose primary goal is col-

lective knowledge creation, e.g. developing vision and

strategy, scoping the problem, and high-level busi-

ness design, the need for simple and intuitive mod-

elling notations, as well as unconstrained medium

(e.g. plastic walls, whiteboards) prevails (Bubenko

et al., 2010). As most stakeholders do not have mod-

elling expertise, the language and tools have to ac-

commodate this, and thus are required to be simple,

intuitive, and corresponding to the natural interaction

that occurs in such situations (Barjis, 2009).

Depending on the nature of a modelling situation,

the modelling language is, to a greater or lesser de-

gree, able to support the formulation of the desired

models. We have discussed a number of different

strategies used in practice to compensate for the lack

of language suitability. These strategies in one way

or the other act on the language speciﬁcation, aiming

to ‘extend’ the actual sets of models which a given

language can express. In doing so, there is inevitably

the risk of violating the intended pragmatics of the

ﬁxed language. However, such a practice may well

be an indication that the pragmatic richness of mod-

elling situations to be supported by the language has

not been adequately taken into account when the lan-

guage was designed. In our view, answering this

dilemma requires a more fundamental understanding

of the role of language in modelling.

Third International Symposium on Business Modeling and Software Design

4 MODEL INTEGRATION

Since enterprises are modelled using different mod-

els/views, it is desirable to maintain their coherence.

The use of a wide range of models and modelling

languages can easily lead to a fragmentation of the

modelling landscape; i.e. a break up of coherence. To

avoid or deal with such a situation, different strate-

gies are employed, e.g. ((Frank, 2002), (Iacob et al.,

2012), (Anaya et al., 2010)). We classify them into

language uniﬁcation and language federation strat-

egy. They both address the integration challenge at

the level of ﬁxed language deﬁnition. In this section,

we discuss some of the challenges that such an ap-

proach raises.

The strategy of uniﬁcation of enterprise modelling

language(s) has the ambition to deﬁne a standard set

of constructs in which all the models (i.e. perspec-

tives) can be expressed. The uniﬁed language is in-

tended to be used instead of the different languages

that partially cover the domain of interest. We can ob-

serve this logic in the deﬁnition of UML, ArchiMate,

as well as language of EKD method (Stirna and Pers-

son, 2012). This approach boils down to preventing

the fragmentation from occurring in the ﬁrst place.

The uniﬁed language offers a ﬁxed, but integrated,

view on some domain of interest. Besides a priori ﬁx-

ing the set of constructs, the standardising effect of

the unifying language lies also in the fact that it a

priori ﬁxes the perspectives for modelling some do-

main. The integration between the different perspec-

tives, i.e. models, is easier to ensure, given that con-

sistency and coherence rules can be embedded in the

language deﬁnition. The CASE tool can then auto-

matically check these properties.

Regretfully, however, it is nearly impossible to a

priori identify which perspectives should be part of an

integrated language. The challenge lies in the fact that

the relevance of different perspectives (and its related

modelling concepts) is highly context-dependent. For

instance, different perspectives may be relevant for

different (types of) enterprises, or even in different

transformation projects of the same enterprise. More-

over, over time, new perspectives may become rele-

vant for a particular enterprise.

To cater for context-dependency, standard mod-

elling languages like UML and ArchiMate offer

means for their extension. For instance, UML has the

well-known stereotyping mechanism (whose prob-

lems are also well-known). In the case of Archi-

Mate, the very design of the language (Lankhorst

et al., 2010) provides different possibilities for ex-

tension. However, these mechanisms are of limited

scope, they mostly allow for reﬁnement of concepts,

not for adding the entire domains which were not en-

visaged by the original language deﬁnition.

At the same time, one can observe how there is

a drive for the ArchiMate language as a standard to

be extended with additional domains. The move from

the ArchiMate 1.0 standard to the ArchiMate 2.0 stan-

dard included two additional domains, namely mo-

tivation and migration. Further integration between

TOGAF and ArchiMate is likely to lead to even more

extensions. Moreover, the extensions with e.g. busi-

ness policies and rules, are also considered (Iacob

et al., 2012). Where will it stop?

The fact that such uniﬁed languages are typically

very generic is already an indicator that the perspec-

tives and concepts that are speciﬁc to different con-

texts cannot be covered. Therefore, in the actual use

of a uniﬁed modelling language in the speciﬁc con-

text, the need to extend the language with ‘missing

domains’ is likely to emerge. On the other hand,

deﬁning the comprehensive and overly detailed lan-

guage covering all the potentially relevant perspec-

tives and related concepts would most likely result in

the overly complex language that would be costly to

use.

According to (Egyedi, 2007), this tension is in-

herent to any standard deﬁnition process, including

modelling standards. The authors argue that deﬁning

the context-independent standards (e.g. enterprise- or

application-independent, etc.) typically leads to very

comprehensive and/or generic standards, therefore

also difﬁcult or too expensive to use. Even more,

this tension is recognised as a fundamental dilemma

in developing standards, which is very difﬁcult to re-

solve (Egyedi, 2007).

Moreover there is another issue with a uniﬁed lan-

guage approach. A common belief is that by means of

a priori imposing standardised vocabulary, frequent

meta-discussions could be avoided, and the knowl-

edge transfer could be facilitated. However, as well as

with perspectives, the languages used in enterprises

are also context-dependent, depending on e.g. pro-

fessional background and education of stakeholders.

These factors exert an inﬂuence on the default way

different people conceptualise (Linden et al., 2012).

In that sense, imposing another language, which is

‘outside’ their area of practice, straight from the be-

ginning is likely to cause, and not resolve, conceptual

misunderstandings.

A language such as ArchiMate is designed to deal

with this issue by enabling users to deﬁne their own

viewpoints, i.e. to have their model (the view) derived

from the integrated model. However, this viewpoint

is still to be deﬁned from the unique ﬁxed ‘footprint’,

i.e. uniﬁed metamodel.

Challenges of Modelling Landscapes - Pragmatics Swept Under the Carpet?

Conversely to the key drive of language uniﬁca-

tion, the rationale of the strategy of federating lan-

guages is not to prevent, but rather to allow and man-

age the variety of modelling languages within the

modelling landscapes. To avoid fragmentation, it

aims to ensure the links between the modelling lan-

guages. A number of different approaches exists to

establish these links. We can classify them on several

dimensions, for the goal of this discussion.

One classiﬁcation dimension regards the way

in which bridges are constructed between the lan-

guages. One approach is to directly connect the lan-

guages on a point-to-point basis, e.g. (B

ezivin et al.,

2006) (Zivkovic et al., 2007) (Fabro and Valduriez,

2009). The links can also be established with the help

of a ‘mediator’, as in Uniﬁed Enterprise Modelling

Language (UEML) (Anaya et al., 2010). The medi-

ator’s role is to serve as the basis to which all the

modelling languages are mapped, and to ensure the

consistency between the models. In UEML, the uni-

ﬁed ‘ontology’ (Opdahl et al., 2012) plays this role.

However, it requires the languages to be (re)deﬁned

in accordance with the speciﬁc grammar, as well as in

full formal precision, given that the approach targets

the semantic interoperability of tools and associated

languages.

We may also distinguish the approaches based on

the moment (in the language’s lifetime) when the

links are established, i.e. based on the temporality.

In the case of point-to-point bridges, the links are es-

tablished between already existing languages. How-

ever, the links may also be deﬁned at the moment

of language design. In that case, the new language

is deﬁned extending or specialising an already ex-

isting (more generic) language, the approach akin to

the older work on the so-called meta-model hierar-

chies (Falkenberg and Oei, 1994) (Falkenberg et al.,

1998). For instance, (Vallecillo, 2010) discuses in

detail different techniques for combining (domain-

speciﬁc) modelling languages based on this hierarchy

logic.

Another dimension of classiﬁcation may be

whether the links are established based on syntac-

tic correspondences only, or whether they also in-

volve semantic correspondences. MDE approaches to

model transformations and model weaving (Fabro and

Valduriez, 2009) generally only consider the syntacti-

cal level. In turn, semantic integration is explored in

order to enhance the quality and reduce the complex-

ity of a priori establishing language bridges (Kara-

giannis and H

offerer, 2006). It is however question-

able what kind of semantics of language constructs

can be captured a priori. This is especially pro-

nounced for generic languages such as i* (Yu and

Mylopoulos, 1996) or UML, but also relevant for

domain-speciﬁc languages, as discussed in (Frank,

2011). Without taking the context of language use,

the abstract concepts (underlying generic languages)

may allow for various interpretations. The precise,

i.e. contextualised interpretation is, however, neces-

sary for the bridges to be meaningful. For instance,

a modelling language such as i* (Yu and Mylopou-

los, 1996) can be used for modelling strategic goals

of actors in relation to the system, but also to express

information systems requirements. In each of these

contexts, the inherent semantics of e.g. the modelling

construct actor will vary: when modelling strategic

goals of enterprise, actor can only be a human actor,

while in the context of modelling software require-

ments, a machine may be an actor as well. Without

taking this into account, the a priori mappings be-

tween i* and ArchiMate (based on language deﬁni-

tion and not use) might be meaningless in some of the

contexts of i* use.

This context-dependent semantic variation can

only be determined by taking into account the con-

text of language use, not a priori. Indeed, in dis-

cussing the meaning of models, (Thalheim, 2012)

distinguishes its two complementary aspects: refer-

ential and functional meaning. While the referential

meaning is well investigated, the functional meaning

relates model elements with the context in which they

are used. The key to understanding the functional

meaning is in modelling pragmatics.

In addition, as we have seen in the reports of the

use of modelling languages, there is an indication

that, in various usage contexts, the syntactical vari-

ation with respect to the original language speciﬁca-

tion also takes place. Therefore, the value of building

the language bridges a priori and based on a priori

ﬁxed languages, i.e. out of the context of its actual

use, might be questioned.

5 WHAT IS MODELLING?

The remainder of the paper discusses our fundamen-

tal understanding of the driving forces and challenges

related to modelling and linguistic variety within en-

terprise modelling landscapes. In our ongoing re-

search, we develop a theory to explain why and how

enterprise modelling landscapes emerge and evolve,

where we focus on the use of modelling languages.

Based on such a theory, our ambition is to revisit

the integration strategies and propose their realis-

tic variants that better caters for the pragmatics of

models/languages. Our view on models and mod-

elling is rooted in semiotics, linguistics and cogni-

Third International Symposium on Business Modeling and Software Design

tive science. This view is inspired by different re-

lated research tackling the fundamental modelling

aspects such as ((Stachowiak, 1973), (Rothenberg,

1989), (Falkenberg et al., 1998), (Hoppenbrouwers

et al., 2005), (Proper et al., 2005), (Thalheim, 2012)).

We look at the models as being essentially a means of

communication about some domain of interest, and

the process of modelling as a communication-driven

process led by a pragmatic focus (Hoppenbrouwers

and Wilmont, 2010).

Though different views on models and modelling

exist, as well as many different deﬁnitions, we will

elaborate the reasons for which we propose the fol-

lowing (general) deﬁnition of model (based on ((Sta-

chowiak, 1973), (Rothenberg, 1989), (Falkenberg

et al., 1998), (Thalheim, 2011))):

A model is an artefact acknowledged by an

observer as representing some domain for a

particular purpose.

By stating that a model is an artefact, we exclude

from the deﬁnition the conceptions (Falkenberg et al.,

1998) or so-called “mental models” (mental spaces

in (Fauconnier, 2010)). Nonetheless, we do consider

conceptions as important within the modelling pro-

cess. Later in this section, we elaborate on their role

and importance, especially in the case of collaborative

modelling.

With observer we refer to the group of people cre-

ating (i.e. model creators) and using the model (i.e.

model audience). On one extreme, it can refer to the

entire society, on the other extreme, to the individ-

ual. Though it may not be the general rule, in en-

terprise modelling context, it is very often the case

that model creators are at the same time its audience.

The observer is the key element in modelling, as it

is only through the appreciation of the observer that

some artefact is acknowledged as being the model.

Similarly to (Falkenberg et al., 1998), we deﬁne

domain as any part “part” or “aspect” of the “world”

considered relevant by the observer in the given mod-

elling context. The “world” here may refer not only

to the “real” world, but also to hypothetical or imag-

ined worlds. Even more, the domain of a model can

be another model as well.

A model always has a purpose. This purposeful-

ness dimension is explicitly present in most of the

model deﬁnitions, e.g. ((Stachowiak, 1973), (Rothen-

berg, 1989), (Thalheim, 2011)). Although acknowl-

edged as an essential dimension of models, the con-

cept of purpose is rarely deﬁned and its role in the

entire modelling process is scantly discussed.

We see the purpose of the model as a combination

of the following dimensions: (1) the domain which

the model should pertain to and (2) the intended us-

age of the model (e.g. analysis, sketching, contract-

ing, execution, etc.) by its intended audience. In line

with (Rothenberg, 1989), (Thalheim, 2011), we argue

that, although usually implicitly present in modelling,

the purpose should be explicit within the modelling

process; i.e. the model creator should be aware of the

intended usage and audience of the model. This is

quite important, since the ﬁtness-for-purpose directly

determines/inﬂuences the degree to which the mod-

els satisﬁes the conception of the domain within the

actual modelling situation. This subsequently inﬂu-

ences the value of the model for its intended usage.

As illustrated in Figure 2, when modelling, the ob-

server O decides

what the relevant aspects of the

“world” under consideration are in the given mod-

elling situation. This results in the conception of the

domain, c

. This process of abstracting away from

certain aspects of the “world” which are not rele-

vant should be driven by the purpose p of the model

m ((Rothenberg, 1989), (Thalheim, 2011)) (depicted

as inﬂuence 1 of the purpose p on the relation concep-

tion of, see Figure 2).

I – view on modelling (high-level)

Single-observer modelling situation.

conception of

Audience

Discusses the particular

modelling situation

(p, d, cd, md)

Still does not introduce

the languages discussion.

Figure 2: The process of modelling.

The observer subsequently tries to shape an arte-

fact (i.e. the model-to-be) in such a way that it ad-

equately represents, for the purpose p, his/her con-

ception of the domain c

. The purpose p is a con-

ception as well, i.e. the conception of the purpose of

the model-to-be c

. Even more, the observer O also

has the conception of the model-to-be, c

. The mod-

elling process actually consists in the observer’s grad-

ual alignment of these three conceptions (illustrated in

Figure 3). This process usually takes place in paral-

lel with the very shaping of the model artefact. When

their mutual alignment is achieved, the artefact is ac-

knowledged as the representation of the (conception

of the) domain d for the purpose p. In other words,

the observer O acknowledges that the artefact m is a

model of the domain d for the purpose p.

Obviously, the observer’s judgement may be inﬂuenced

by many different factors, e.g. observer’s intentions, expe-

rience, previous knowledge, etc. We exclude from our con-

sideration the potential political intentions of the observer.

Challenges of Modelling Landscapes - Pragmatics Swept Under the Carpet?

I – view on modelling (detailed)

Single observer. Zoom at conceptions.

cd d

conception of

Audience

Discusses the particular

modelling situation

(p, d, cd, md) and zooms

into the conceptions

world.

Still does not introduce

the languages discussion.

Figure 3: Aligning conceptions in modelling.

II – Collaborative modelling

cd1

cd2

cdn

cm1

cm2

cmn

cp1= cp2 = … = cpn ?

cd1= cd2 = … = cdn ?

cm1= cm2 = … = cmn ?

cp1

cp2

cpn

Figure 4: Co-aligning conceptions in modelling.

The previous explanation holds for a single ob-

server modelling process. But what does happen in

a collaborative modelling process? In such a situa-

tion, a group of n human actors is involved in the

process of modelling, and is supposed to jointly ob-

serve some domain and come up with its model, for

some purpose. The great challenge in the collabo-

rative modelling consists in the fact that each partic-

ipant has its own conception of the domain, of the

model taking shape and of the purpose of that model.

This is illustrated in Figure 4. In addition, these in-

dividual conceptions are inﬂuenced by the individ-

ual pre-conceptions (Proper et al., 2005), brought by

the particular social, cultural, educational and profes-

sional background. So, in order to reach shared un-

derstanding and appreciation of the artefact m as a

common model of the domain d for the purpose p,

the co-alignment of the n×3 conceptions has to take

place. Indeed, this co-alignment of conceptions is

generally considered as a critical step in the collabora-

tive modelling, where all the discussions, negotiations

and agreement about the model has to take place.

We have seen how, in the modelling process, mod-

els gradually come into being by (co-)aligning the dif-

ferent conceptions in the observer’s mind(s), and by

their subsequent externalisation. To externalise the

(aligned) conception of the domain c

into m, the ob-

server O uses some system of symbols/signs. Essen-

tially, the observer establishes the mapping between

the conception into some system of signs, whereby

signs used come to represent the observer’s concep-

tion of the domain c

. The models externalised us-

ing some system of symbols are usually referred to as

symbolic models.

We will now look more closely at the role of the

modelling language, as well as the factors contribut-

ing to the modelling and linguistic variety in the mod-

elling landscapes.

6 MODELLING LANGUAGES

6.1 Role of Modelling Languages

A language in use may be regarded as a medium sys-

tem (Hoppenbrouwers, 2003), involving both a lan-

guage and a medium. The medium refers to the physi-

cal means to achieve communication (Hoppenbrouw-

ers, 2003), e.g. audio, video, writing, etc. We enter-

tain this view as it allows us to discuss, from a funda-

mental perspective, different and often conﬂicting re-

quirements put on modelling languages. We thus pro-

pose to consider that a modelling language has two

primary roles, the role as a language to be used by

humans, and the role as a medium, i.e. carrier of sets

of models aimed at mechanical manipulation. Fun-

damentally, we regard language as an instrument of

human activity, primary in support of reﬂection and

communication. This is in line with a functional per-

spective (Cruse, 2011) and the action tradition on lan-

guage (Clark, 1993). We are thus primarily interested

in how ﬁxed modelling languages play that role.

In its role as a language, it should serve as a sup-

port of activities taking place in the modelling pro-

cess. The central issue therefore is to which extent an

a priori ﬁxed language can act as an effective means

of human communication and knowledge creation in

the actual modelling situation. Let us look closely at

its adequacy for creating conceptions. As shown in

Figure 5, if model m is expressed in a modelling lan-

guage L

, what is the language L

in which the con-

ception c

is constructed in the observer’s mind: is it

the modelling language L

, or some other language,

e.g. the observer’s native language?

Obviously, the deﬁnitive answer to this question is

not easy to provide. Nevertheless, we believe that it is

important to create awareness of the potential gap be-

tween L

and L

and its consequences. As previously

discussed, a ﬁxed modelling language comes with an

a priori embedded ﬁlter on the ‘world’. In the mod-

elling process, it thus tends to constrain, or at least

inﬂuence, the conception of a domain. Depending on

the actual modelling situation, this pre-conceived ﬁl-

Third International Symposium on Business Modeling and Software Design

I – view on modelling (high-level)

Single-observer modelling situation.

conception of

Audience

Discusses the particular

modelling situation

(p, d, cd, md)

Languages focus!

Lc ?

Figure 5: The role of language in the modelling process.

ter may prove to be inadequate for the particular ob-

server and for the particular modelling problem.

For instance, research from cognitive linguistics

shows that the entire social, cultural, educational and

professional background of an observer plays a role

in shaping their ‘linguistic personality’ (Novodra-

nova, 2009). This includes their ‘world view’; i.e.

their natural way of conceptualising phenomena in

the ‘world’ (Schmid, 2010). Likewise, (Linden et al.,

2012) shows that different people have different de-

fault interpretations of the abstract concepts underly-

ing enterprise modelling languages. It is reasonable

to assume that, in a particular modelling situation,

(at least) non-modelling-experts form conceptions in

a language signiﬁcantly different from L

. This is

likely to increase the L

−L

gap and negatively im-

pact the general suitability of L

Another factor potentially effecting the L

−L

gap involves the nature of the modelling task in

the particular modelling situation. While the re-

lation between the nature of the modelling prob-

lem and the suitability of the modelling language

needs further study, the empirical data indicates rather

negative inﬂuence of overly restrictive (in terms of

syntactic-semantic restrictions embedded in) ﬁxed

language on the creativity in modelling. For in-

stance, (Anaby-Tavor et al., 2010) observe the inad-

equacy of ﬁxed modelling languages to support the

exploration phases where things are unclear and am-

biguous, and models are used to organise informa-

tion, gain insight, envision alternative possible futures

etc. (Anaby-Tavor et al., 2010). Similarly, (Bubenko

et al., 2010) observe that in highly creative and collab-

orative situations such as vision and strategy devel-

opment, rather informal and intuitive notations (and

mediums) seem to be of better support.

In its role as a medium, a modelling language

should accommodate the formulation of models,

while allowing their mechanical manipulation. The

potential added value of the modelling language,

from this perspective, lies primarily in its re-usability

across different modelling problems, and the extent to

which the language speciﬁcation is machine readable.

As discussed in Section 2, the reusability of a lan-

guage relates to its expressiveness. Obviously, this

makes sense for the development of tools and auto-

mated model management. However, while general-

purpose modelling languages are usually more ex-

pressive, they are less suitable than domain or

purpose-speciﬁc languages for speciﬁc problem do-

mains and modelling situations. These languages in-

corporate in their deﬁnition concepts that are tuned to

the modelling of particular domains. The overall aim

is to foster modelling productivity, facilitate the un-

derstanding of the models by the domain stakeholders

and increase the overall quality of resulting models, in

particular semantic and pragmatic quality (Krogstie

et al., 2006). While domain and purpose-speciﬁc lan-

guages seem to correspond to the natural, i.e. hu-

man, need for suitability of the modelling language,

they are not easy to reuse across different situations.

General-purpose modelling languages, on the other

hand, are easier to reuse for modelling different do-

mains. Nevertheless, the interpretation challenge of

the models expressed in them is more pronounced.

It certainly does not make sense to have situation-

speciﬁc modelling language emerge from scratch in

each new modelling situation. However, it does make

sense to embed in the (generic) meta-model the ele-

ments that are repeatedly discussed in similar mod-

elling contexts. Therefore, there is a need to carefully

balance potential sets of models one would like to ex-

press in a language, and potential sets of modelling

situations one would like to support.

The second added value of a language as a

medium is machine readability. This is driven by

the need for automated manipulation of models. This

is achieved by formal, i.e. precise and unambiguous,

deﬁnition of both the syntax and semantics of the lan-

guage, usually in a mathematical language. This es-

sentially boils down to expressing the semantics of

the model/language in terms of another syntactic rep-

resentation, i.e. expressing the syntactic semantics.

Though necessary for the machine’s correct interpre-

tation of the model, this kind of a priori ﬁxed se-

mantics does not tell anything about what the model

means to the observer. In particular, it allows by no

means to a priori precisely capture the meaning of a

model as it occurs in the actual modelling situation,

i.e. the model’s pragmatic semantics.

6.2 Language Variety

We are now able to suggest two primary drivers for

the variety within enterprise modelling landscapes. In

our view, these drivers relate to:

Challenges of Modelling Landscapes - Pragmatics Swept Under the Carpet?

Abstraction variety – Abstraction is at the heart of

modelling. As we have seen, it boils down to pur-

posefully neglecting irrelevant details of the observed

“world”. The need for differing levels of abstrac-

tion in dealing with an enterprise is related to its

complexity/multifaceted-ness. The abstraction vari-

ety thus leads to the increase in number of needed

perspectives, i.e. models, in modelling.

Manifestation variety – The manifestation here

refers to the way the model is represented on the

medium. Fundamentally, the need for this kind of va-

riety is rooted in the complexity and heterogeneity of

social structures underlying modelling. The more het-

erogeneity in stakeholder groups, the more likely is

the need for different manifestations of arguably the

“same” model. This invites an increase in linguistic

variety on top of the abstraction variety.

These types of variety are illustrated in Figure 6.

Throughout the paper, we have seen that these factors

are to a large extent situational, i.e. they depend on

the particular enterprise and enterprise modelling ef-

fort, involved stakeholders (observer), the purpose for

which models are created in this effort, etc. This leads

us to the conclusion that modelling landscapes should

be situated. Indeed, the ‘standard eroding’ effects dis-

cussed in Sections 3 and 4, might be seen as the man-

ifestation of the need to make these landscapes better

situated, driven by the pragmatic needs of the wider

organisational context that the landscapes cover.

V – Modelling landscape

Sources of diversity

C D

conception of

Abstraction

variety

Manifestation

variety

Figure 6: Sources of variety in the modelling landscapes.

Evidently, both of these drivers stem from the lan-

guage role of the modelling language. They also fun-

damentally conﬂict with the drivers for the a priori

standardisation of the language, which stem from its

medium role. We believe that this natural polarity

deserves careful management, rather than denial. It

should by no means be swept under the carpet.

7 CONCLUSION & OUTLOOK

Based on the discussions so far, we posit that, at the

heart of the challenge of creating integrated modelling

landscapes, lies the question: What can be a priori

ﬁxed in a modelling language? To create more bal-

anced strategies that cater for the pragmatic needs of

modelling landscapes, it is necessary to carefully ex-

amine which aspects may be feasible to a priori ﬁx in

the modelling language. In this ﬁnal section, and rely-

ing on the presented fundamental view on modelling

and language, we draw some initial conclusions as a

tentative (though partial) answer to this question.

First of all, the semantics. Given that (the concep-

tion of) the domain actually does not ‘exist’

a priori

but emerges in the very process of its modelling, the

semantics, in the pragmatic semantics sense, also can-

not be captured a priori. One would expect to be able

to at least ﬁx the grammar (abstract syntax) and sym-

bols used for its representation on the medium (con-

crete syntax). To the extent to which the grammar is

tuned to the needs of the intended set of modelling

situations, it can be a priori ﬁxed. One can also start

by only a priori ﬁxing the core grammar, and allow-

ing its further reﬁnement a posteriori during the use

of the language. For instance, it can be possible to

start the modelling process with lightly constrained

vocabulary adapted to the domain/purpose (e.g. based

on some historical heuristics), and then gradually in-

clude more formalisation, to the extent necessary for

the intended usage and audience of the model. Need-

less to say, this would necessitate modelling infras-

tructures to be more ﬂexible. A growing interest in

this subject can indeed be observed, e.g. ((Cho et al.,

2011), (Ossher et al., 2009)).

The point we aim to make is that, although hav-

ing an a priori ﬁxed representation to a large extent

facilitates the development of tools and automation of

model manipulations, ﬁxing the language that is to be

used in human communication may seriously damage

its capacity to adequately express thoughts, i.e. con-

ceptions of domains in the given modelling situation.

Even if carefully deﬁned, the standardised enterprise

modelling language will inevitably demonstrate the

need to evolve, to adapt to the dynamically changing

The term exist is used here in the sense of Heidegger’s

notion of breaking down, discussed in (Winograd and Flo-

res, 1986). Indeed, “Heidegger insists that it is meaningless

to talk about the existence of objects and their properties

in the absence of concernful activity, with its potential for

breaking down. What really is is not deﬁned by an objec-

tive omniscient observer, nor is it deﬁned by an individual

– the writer or computer designer – but rather by a space

of potential for human concern and action” (Winograd and

Flores, 1986, p.37).

Third International Symposium on Business Modeling and Software Design

‘reality’ of enterprises and their environments, and

thus to the human sense-making of that ‘reality’.

As our next step, we aim to further explore the

theoretical and practical considerations presented in

this paper. We aim to start by analysing the available

instruments for modelling language design, adapta-

tion and combination, and the potential of their im-

proving or combining in order to support purpose-

speciﬁc language adaptations. In addition, we aim

to extend these instruments to allow for explicit

modelling of the modelling pragmatics. For in-

stance, megamodel (Favre and Nguyen, 2004), view-

point (ISO, 2011), metamodel hierarchies (Falken-

berg et al., 1998), metamodel inference (Ossher et al.,

2009) etc. are some of the instruments of our particu-

lar interest.

ACKNOWLEDGEMENTS

This work has been partially sponsored by the

Fonds National de la Recherche Luxembourg

(www.fnr.lu), via the PEARL programme.

REFERENCES

Anaby-Tavor, A., Amid, D., Fisher, A., Bercovici, A., Os-

sher, H., Callery, M., Desmond, M., Krasikov, S., and

Simmonds, I. (2010). Insights into enterprise concep-

tual modeling. Data Knowl. Eng., 69(12):1302–1318.

Anaya, V., Berio, G., Harzallah, M., Heymans, P., Matulevi-

cius, R., Opdahl, A. L., Panetto, H., and Verdecho, M.

(2010). The Uniﬁed Enterprise Modelling Language -

Overview and Further Work. Computers in Industry,

61:99–111.

Barjis, J. (2009). Collaborative, Participative and Interac-

tive Enterprise Modeling. In ICEIS, pages 651–662.

ezivin, J., Bouzitouna, S., Fabro, M. D. D., Gervais, M.-P.,

Jouault, F., Kolovos, D. S., Kurtev, I., and Paige, R. F.

(2006). A canonical scheme for model composition.

In ECMDA-FA, pages 346–360.

Bjekovi

c, M., Proper, H. A., and Sottet, J.-S. (2012). To-

wards a coherent enterprise modelling landscape. In

PoEM (Short Papers). CEUR-WS.org.

Briand, L. C., Melo, W. L., Seaman, C. B., and Basili, V. R.

(1995). Characterizing and Assessing a Large-Scale

Software Maintenance Organization. In ICSE, pages

133–143.

Bubenko, J. A. j., Persson, A., and Stirna, J. (2010). An In-

tentional Perspective on Enterprise Modeling. In Sali-

nesi, C., Nurcan, S., Souveyet, C., and Ralyt

e, J., edi-

tors, Intentional Perspectives on Information Systems

Engineering, pages 215–237. Springer.

Cho, H., Sun, Y., Gray, J., and White, J. (2011). Key chal-

lenges for modeling language creation by demonstra-

tion. In ICSE 2011 Workshop on Flexible Modeling

Tools, FLEXITOOLS.

Clark, H. (1993). Arenas of Language Use. University of

Chicago Press.

Cruse, A. (2011). Meaning in Language: An Introduction

to Semantics and Pragmatics. Oxford Textbooks in

Linguistics. Oxford University Press.

Davies, I., Green, P. F., Rosemann, M., Indulska, M., and

Gallo, S. (2006). How do practitioners use conceptual

modeling in practice? Data Knowl. Eng., 58(3):358–

380.

Delen, D., Dalal, N. P., and Benjamin, P. C. (2005). Inte-

grated modeling: the key to holistic understanding of

the enterprise. Commun. ACM, 48(4):107–112.

Egyedi, T. M. (2007). Standard-compliant, but incompati-

ble?! Computer Standards & Interfaces, 29(6):605–

613.

Elahi, G., Yu, E. S. K., and Annosi, M. C. (2008). Modeling

Knowledge Transfer in a Software Maintenance Orga-

nization - An Experience Report and Critical Analy-

sis. In PoEM 2008, pages 15–29.

Fabro, M. D. D. and Valduriez, P. (2009). Towards the ef-

ﬁcient development of model transformations using

model weaving and matching transformations. Soft-

ware and System Modeling, 8(3):305–324.

Falkenberg, E. and Oei, J. (1994). Meta Model Hierar-

chies from an Object–Role Modelling Perspective. In

Proceedings of the First International Conference on

Object–Role Modelling (ORM–1), pages 218–227.

Falkenberg, E. D., Hesse, W., Lindgreen, P., Nilsson, B. E.,

Oei, J., Rolland, C., Stamper, R. K., Assche, F. J. V.,

Verrijn-Stuart, A. A., and Voss, K. (1998). FRISCO -

A Framework of Information System Concepts - The

FRISCO Report. Technical report, IFIP WG 8.1 Task

Group FRISCO.

Fauconnier, G. (2010). Mental Spaces. In Geeraerts, D.

and Cuyckens, H., editor, The Oxford Handbook of

Cognitive Linguisics, pages 351–376. Ofxord Univer-

sity Press.

Favre, J. and Nguyen, T. (2004). Towards a megamodel to

model software evolution through transformation. In

SETRA workshop.

Frank, U. (2002). Multi-perspective Enterprise Model-

ing (MEMO) - Conceptual Framework and Modeling

Languages. In Proceedings of HICSS’02, volume 3.

IEEE.

Frank, U. (2011). Some Guidelines for the Conception of

Domain-Speciﬁc Modelling Languages. In EMISA,

pages 93–106.

Gordijn, J. and Akkermans, H. (2003). Value based require-

ments engineering: Exploring innovative e-commerce

ideas. Requirements Engineering Journal, 8(2):114–

134.

Greefhorst, D., Koning, H., and Van Vliet, H. (2006). The

many faces of architectural descriptions. Information

Systems Frontiers, 8(2):103–113.

Harel, D. and Rumpe, B. (2004). Meaningful Modeling:

What’s the Semantics of ”Semantics”? IEEE Com-

puter, 37(10):64–72.

Hoppenbrouwers, S. (2003). Freezing Language; Concep-

tualisation processes in ICT supported organisations.

Challenges of Modelling Landscapes - Pragmatics Swept Under the Carpet?

PhD thesis, University of Nijmegen, Nijmegen, The

Netherlands.

Hoppenbrouwers, S., Proper, H., and Weide, T. v. d. (2005).

A Fundamental View on the Process of Conceptual

Modeling. In Conceptual Modeling - ER 2005, vol-

ume 3716, pages 128–143. Springer.

Hoppenbrouwers, S. J. B. A. and Wilmont, I. (2010). Fo-

cused Conceptualisation: Framing Questioning and

Answering in Model-Oriented Dialogue Games. In

PoEM 2010, pages 190–204.

Iacob, M.-E., Jonkers, H., Lankhorst, M., Proper, H., and

Quartel, D. (2012). ArchiMate 2.0 Speciﬁcation. The

Open Group.

ISO (2011). ISO/IEC/IEEE 40210:2011 Systems and soft-

ware engineering – Architecture description.

Kaidalova, J., Seigerroth, U., Kaczmarek, T., and Shilov, N.

(2012). Practical Challenges of Enterprise Modeling

in the Light of Business and IT Alignment. In PoEM

2012, pages 31–45.

Karagiannis, D. and H

offerer, P. (2006). Metamodels in

Action: An overview. In ICSOFT (1). INSTICC Press.

Karlsen, A. (2011). Enterprise Modeling Practice in ICT-

Enabled Process Change. In PoEM 2011, pages 208–

222.

Krogstie, J., Sindre, G., and Jørgensen, H. D. (2006). Pro-

cess models representing knowledge for action: a re-

vised quality framework. EJIS, 15(1):91–102.

Lankhorst, M., editor (2005). Enterprise Architecture

at Work: Modelling, Communication and Analysis.

Springer, Berlin, Germany.

Lankhorst, M., Torre, L. v. d., Proper, H., Arbab, F., and

Steen, M. (2005). Viewpoints and Visualisation. In

(Lankhorst, 2005), pages 147–190.

Lankhorst, M. M., Proper, H. A., and Jonkers, H. (2010).

The Anatomy of the ArchiMate Language. IJISMD,

1(1):1–32.

Linden, D. J. T. v. d., Hoppenbrouwers, S., Lartseva, A., and

Molnar, W. (2012). Using Psychometrics to Explicate

Personal Ontologies in Enterprise Modeling. Applied

Ontology, 7:1–15.

Malavolta, I., Lago, P., Muccini, H., Pellicone, P., and Tang,

A. (2012). What industry needs from architectural lan-

guages: an industrial study. Technical report, Univer-

sity of L’Aquilla.

Moody, D. (2009). The “Physics” of Notations: Toward

a Scientiﬁc Basis for Constructing Visual Notations

in Software Engineering. IEEE Transactions on Soft-

ware Engineering, 35(6):756–779.

Novodranova, V. F. (2009). Representation of the Language

Personality in LSP. In Reconceptualizing LSP , Online

proceedings of the XVII European LSP Symposium.

OMG (2003). UML 2.0 Superstructure Speciﬁcation – Final

Adopted Speciﬁcation. Technical Report ptc/03–08–

02.

OMG (2006). Semantics of Business Vocabulary and Rules

(SBVR). Technical Report dtc/06–03–02, Object

Management Group, Needham, Massachusetts.

OMG (2008). Business Process Modeling Notation, V1.1.

OMG Available Speciﬁcation OMG Document Num-

ber: formal/2008-01-17.

Opdahl, A., Berio, G., Harzallah, M., and Matulevicius, R.

(2012). Ontology for Enterprise and Information Sys-

tems Modelling. Applied Ontology, 7(1):49–92.

Ossher, H., Bellamy, R. K. E., Amid, D., Anaby-Tavor,

A., Callery, M., Desmond, M., de Vries, J., Fisher,

A., Frauenhofer, T., Krasikov, S., Simmonds, I., and

Swart, C. (2009). Business insight toolkit: Flexi-

ble pre-requirements modeling. In ICSE Companion,

pages 423–424. IEEE.

Proper, H. A., Verrijn-Stuart, A. A., and Hoppenbrouwers,

S. (2005). On Utility-based Selection of Architecture-

Modelling Concepts. In APCCM 2005, pages 25–34.

Rothenberg, J. (1989). The Nature of Modeling. In Arti-

ﬁcial intelligence, simulation & modeling, pages 75–

92. John Wiley & Sons, Inc., USA.

Schmid, H.-J. (2010). Entrenchment, Salience, and Basic

Levels. In Geeraerts, D. and Cuyckens, H., editor,

The Oxford Handbook of Cognitive Linguistics, pages

117–138. Oxford University Press.

Stachowiak, H. (1973). Allgemeine Modelltheorie.

Springer, Berlin, Germany.

Stirna, J. and Persson, A. (2012). Evolution of an Enterprise

Modeling Method - Next Generation Improvements of

EKD. In PoEM 2012, pages 1–15.

Thalheim, B. (2011). The Theory of Conceptual Models,

the Theory of Conceptual Modelling and Foundations

of Conceptual Modelling. In Handbook of Conceptual

Modeling, pages 543–577. Springer.

Thalheim, B. (2012). Syntax, Semantics and Pragmat-

ics of Conceptual Modelling. In NLDB, pages 1–10.

Springer.

Vallecillo, A. (2010). On the Combination of Domain Spe-

ciﬁc Modeling Languages. In ECMFA, pages 305–

320.

Vernadat, F. (2002). UEML: Towards a Uniﬁed Enterprise

Modelling Language. International Journal of Pro-

duction Research, 40(17):4309–4321.

Wagter, R., Proper, H., and Witte, D. (2012). A Practice-

Based Framework for Enterprise Coherence. In Pro-

ceedings of PRET 2012, volume 120 of LNBIP.

Springer. To appear.

Winograd, T. and Flores, F. (1986). Understanding Com-

puters and Cognition - A New Foundation for Design.

Ablex Publishing Corporation.

Winter, R. and Fischer, R. (2007). Essential Layers, Ar-

tifacts, and Dependencies of Enterprise Architecture.

Journal Of Enterprise Architecture, 3(2):7–18.

Wood–Harper, A., Antill, L., and Avison, D. (1985). Infor-

mation Systems Deﬁnition: The Multiview Approach.

Blackwell, Oxford, United Kingdom.

Yu, E. and Mylopoulos, J. (1996). Using goals, rules,

and methods to support reasoning in business pro-

cess reengineering. International Journal of Intel-

ligent Systems in Accounting, Finance and Manage-

ment, 5(1):1—13.

Zachman, J. (1987). A framework for information systems

architecture. IBM Systems Journal, 26(3).

Zivkovic, S., K

uhn, H., and Karagiannis, D. (2007). Fa-

cilitate Modelling Using Method Integration: An Ap-

proach Using Mappings and Integration Rules. In

ECIS, pages 2038–2049.

Third International Symposium on Business Modeling and Software Design