c26433

7/30/2019 c26433

1/17

This is the author version published as:

Thisistheacceptedversionofthisarticle.Tobepublishedas:This is the author version published as:

QUT Digital Repository:http://eprints.qut.edu.au/

Rosemann, Michael and Green, Peter and Indulska, Marta and Recker, Jan C.

(2009) Using ontology for the representational analysis of process modellingtechniques. International Journal of Business Process Integration andManagement, 4(4). pp. 251-265.

Copyright 2009 Inderscience Publishers.

7/30/2019 c26433

2/17

Int. J . Business Process Integration and Management, Vol. X, No x, xxxx 1

Copyright 2006 Inderscience Enterprises Ltd.

Using Ontology for theRepresentational Analysis ofProcess Modelling TechniquesMichael Rosemann*Faculty of Science and TechnologyQueensland University of Technology126 Margaret StreetBrisbane QLD 4000, AustraliaE-mail: [email protected]*Corresponding author

Peter Green

UQ Business SchoolUniversity of Queensland11 Salisbury RoadIpswich QLD 4305, AustraliaEmail: [email protected]

Marta IndulskaUQ Business SchoolUniversity of Queensland11 Salisbury RoadIpswich QLD 4305, AustraliaEmail: [email protected]

J an ReckerFaculty of Science and TechnologyQueensland University of Technology126 Margaret StreetBrisbane QLD 4000, AustraliaE-mail: [email protected]

Abstract: Selecting an appropriate business process modelling technique forms animportant task within the methodological challenges of a business process managementproject. While a plethora of available techniques has been developed over the last decades,there is an obvious shortage of well-accepted reference frameworks that can be used toevaluate and compare the capabilities of the different techniques. Academic progress hasbeen made at least in the area of representational analyses that use ontology as a benchmarkfor such evaluations. This paper reflects on the comprehensive experiences with theapplication of a model based on the Bunge ontology in this context. A brief overview of theunderlying research model characterizes the different steps in such a research project. Acomparative summary of previous representational analyses of process modellingtechniques over time gives insights into the relative maturity of selected process modellingtechniques. Based on these experiences suggestions are made as to where ontology-basedrepresentational analyses could be further developed and what limitations are inherent tosuch analyses.

Keywords: Business Process Management, Representational Analysis, Ontology, BusinessProcess Modelling.

Referenceto this paper should be made as follows: Rosemann, M., Green, P., Indulska, M.and Recker, J . (xxxx) Using Ontology for the Representational Analysis of Process

7/30/2019 c26433

3/17

2 M.ROSEMANN,P.GREEN,M.INDULSKA ANDJ .RECKER

Modelling Techniques, Int. J . Business Process Integration and Management, Vol. x, No.x, pp. xxxx.

Biographical notes: Michael Rosemann is a Professor at the School of InformationSystems and Co-Leader of the Business Process Management Group, QueenslandUniversity of Technology. He received his MBA and PhD from the University of Muenster,Germany. His current research interests include business process management, processmodelling, Enterprise Systems and representation theory. Michael is the author/editor of

five books and has published more than 120 refereed papers in academic journals,conferences and books.

Peter F. Green is Professor of Electronic Commerce and Business Information Systems clusterleader in the UQ Business School at the University of Queensland. He has qualifications inComputer Science, Accounting, and a PhD in Commerce (Information Systems) from theUniversity of Queensland. Dr Green is a Chartered Accountant and a Member of the AustralianComputer Society. Dr Green has worked during his career as the Systems Support Manager atthe South-East Queensland Electricity Board (SEQEB), for a Chartered Accountancy firm, anda Queensland government department. Peter has researched, presented, and published widelyon systems analysis and design, conceptual modelling, information systems auditing, andeCommerce. Dr Green's publications have appeared in such internationally refereed

journals as Information Systems, IEEE Transactions on Knowledge & Data Engineering,Data & Knowledge Engineering, Journal of Database Management, and the Australian

Journal of Information Systems.Marta Indulska is a Lecturer at the UQ Business School, The University of Queensland.She obtained her PhD in Computer Science, in the area of Information Systems, at theUniversity of Queensland, in 2004. Martas current main research interests are BusinessProcess Management and ontology. She has published and presented her work at numerousinternational conferences. Her work has also been published by journals such as IEEE

Transactions on Knowledge & Data Engineering, and Data & Knowledge Engineering.Her teaching focuses on topics in Electronic Commerce and Information Systems.

Jan Recker is a Senior Lecturer at the Queensland University of Technology Brisbane,Australia. He received a PhD in Information Systems from Queensland University of

Technology in 2008. His current research interests include business process modelling,process flexibility, representation theory and technology acceptance. Jans work hasappeared in journals such as the Journal of the Association for Information Systems,

Information Systems, the Scandinavian Journal for Information Systems, theCommunications of the Association for Information Systems, the Australasian Journal ofInformation Systems, the Business Process Management Journal, and others.

1 INTRODUCTION

Business Process Management (BPM) has been identified asa top business priority, and building business processcapability is seen as a major challenge for senior executives

in the coming years (Gartner Group, 2006). The interest inBPM has, amongst others, led to an increased popularity ofbusiness process modelling (Davies et al., 2006). Due to thedemand for a more disciplined approach for BusinessProcess Management, many organizations have beenmotivated to make substantial investments in processmodelling initiatives. This situation in turn has triggeredsignificant related academic and commercial work aimingtowards advanced business process modelling solutions.Consequently, a competitive market is providing a plethoraof tools and methods for process modelling initiatives(Mierset al., 2006).

Process modelling techniques range nowadays fromsimple flowcharting techniques (American NationalStandards Institute, 1970) to business modelling approaches

such as Event-driven Process Chains (Keller et al., 1992)and software engineering-driven techniques such as UML(Fowler, 2004) to formalized and academically studiedtechniques such as Petri nets (Petri, 1962) and its dialects.One of the most recent proposals for a new process

modelling technique is the Business Process ModelingNotation (BPMN). BPMN was released in May 2004,adopted by OMG in February 2006 for standardizationpurposes (BPMI.org and OMG, 2006b), and has sincereceived significant attention.The emergence of such new process modelling techniques

leads to the question if there are actual signs of anincreasing maturity within the capabilities of processmodelling techniques. More specifically, there is increasingdemand to shift the academic resources committed to thedevelopment of new and further extensions of existingprocess modelling techniques to the critical evaluation andcomparison of the already available set of modelling

techniques (Moody, 2005; Nelson et al., 2005). This moveis a pre-requisite for an evolving research discipline that

7/30/2019 c26433

4/17

USINGONTOLOGY FOR THE REPRESENTATIONAL ANALY SIS OFPROCESS MODELINGTECHNIQUES 3

builds on the existing body of knowledge, has an awarenessfor the remaining open challenges, and is guided by amethodological procedure in its future research efforts(Kuhn, 1962; Keen, 1980; Weber, 1997). This scenario isparticularly the case in Information Systems (IS) designresearch where the analysis of the strengths and weaknessesof existing methods can be used as the basis for developingnew and improved methods (Bubenko, 1986).The large selection of currently available process

modelling techniques, however, stands in sharp contrast tothe paucity of evaluation frameworks that can be used forthe task of evaluating and comparing those modellingtechniques. A complete analysis of all facets of a processmodelling technique is a significant task and would include,amongst others, its expressive power, the consistency andcorrectness of its meta-model, the perceived intuitiveness ofits notation, the available tool support, the existing referencemodels expressed in this technique (e.g., ITIL, SCOR), andthe like. While there is unfortunately not one single

framework that facilitates such a comprehensive analysis,reasonably mature research has emerged over the lastdecade with a focus on therepresentational capabilitiesandthe expressive power of modelling techniques. Suchframeworks have been developed either inductively fromobservable practice or deductively from applicable theories.

A prominent example of an inductively developedframework is the set of workflow patterns developed by vander Aalst et al. (2003). The development of this frameworkwas triggered by a bottom-up analysis and comparison offifteen workflow management systems during 2000 and2001, with focus on their underlying modelling techniques.

The goal was to bring insights into the expressive power ofthe underlying languages and thereby outline similaritiesand differences between the analyzed systems. Theworkflow patterns research initially was focused on thedevelopment of a minimal set of control flow patterns thatin composition constitute possible control flow logic. Bynow, this research has also explored related perspectivesand a first set of data (Russell et al., 2005a) and resourcepatterns (Russell et al., 2005b) have been proposed. Theworkflow pattern-based evaluation of various processmodelling techniques, such as BPEL4WS (Wohed et al.,2003), UML Activity Diagrams (Wohed et al., 2005), andBPMN (Wohed et al., 2006), is based on the assumptionthat a more complete coverage of these patterns leads totechniques and systems with advanced expressive power.

On the other hand, an increasingly popular evaluationframework that has been derived through deductive researchmethods is representational analysis based on foundationalontologies. Unlike the high number of ontologies that havebeen proposed in the AI discipline for various domains, seee.g., (Gruber, 1993; Uschold and Grninger, 1996),foundational (upper) ontologies are well-grounded inthorough philosophical research.

Of those two types of frameworks, this paper is focusedon the latter, i.e., representational analyses of process

modelling techniques based on ontologies. Specifically, weconcentrate on a rapidly emerging body of research that is

grounded in a well-established, mathematically definedontology proposed by Mario Bunge (1977; 1979). Thisontology has been adapted by Weber and Wand (1988) intoa theory of representation that is closer to the demands andterminology of the Information Systems community. Thistheory of representation became widely known as theBunge-Wand-Weber (BWW) representation model.The deployment of the BWW representation model in

studies on the representational capabilities of processmodelling techniques can be justified on at least threepremises. First, unlike many other foundational theoriesbased on ontology, the BWW model has been derived withthe Information Systems discipline in mind (Wand andWeber, 1990a, p. 124). Second, while the BWW model doesnot denote a unique case of IS-specific ontologies, refer, forinstance, to (Guizzardi, 2005), there is an established trackrecord and demonstrated usefulness of representationalanalyses of modelling techniques using the BWWrepresentation model. Third, the BWW model officiates as

an upper ontology for the modelling of InformationSystems, and the foundational character and comprehensivescope allows for wide applicability.The aim of this paper is to explore and discuss the benefits

and limitations of such representational analyses in thecontext of process modelling techniques. Thereby, bothdevelopers and users may be able to pinpoint, scrutinize andcompare potential representational shortcomings of currentspecifications in order to derive more sophisticated, revisedtechniques. To address this objective, we studied previousrepresentational analyses of process modelling techniquesand complemented this body of research with our own

analysis of Petri nets and BPMN in order to achieve areasonably complete set of evaluated process modellingtechniques. Based on the consolidation of these studies andour own experiences, we present a research model that canserve as a guideline for applying representation theory instudies of modelling techniques. We discuss the differentsteps involved in such studies and present some guidelinesfor each of them. We also show how representationalanalysis can be used to draw conclusions on the evolution ofprocess modelling techniques over time, and discuss perilsand limitations related to representational analysis.The remainder of this paper is structured as follows. The

next section gives an overview of the general researchmodel which underlies this type of evaluation. Next, webriefly introduce the selected theoretical model andsummarize in section 3 previous research onrepresentational analyses of process modelling techniques.Section 4 compares the outcomes of these analyses andstudies signs of increasing process modelling capabilitiesover time. Section 5 critically reflects on the limitations ofsuch analyses. We conclude in section 6 with a proposedresearch agenda that might give interested researchersinspirations for related work.

7/30/2019 c26433

5/17


2 A RESEARCH MODEL FOR REPRESENTATIONAL

ANALYSES

Before we provide an overview of the outcomes of previousrepresentational analyses of process modelling techniques, itis necessary to understand and appreciate the overallresearch model that underlies such evaluations.The theoretical foundation and rigor of this type of

research is derived from the selected foundational ontology.Research based on the BWW models can be traced back tothe comprehensive and detailed work by Mario Bunge(1977; 1979) and its accomplishments. The initial andongoing development of such ontologies, and thecomparison of different foundational ontologies, is achallenging task that is located in the discipline ofphilosophy and has its roots in Aristotles foundational workon metaphysics (Aristotle, 1991).The philosophical nature of Bunges ontology, its

terminology and the overall scope, however, are not very

conducive to application in the context of InformationSystems or more specifically Business Process Managementand modelling. Nevertheless, Wand and Weber saw thepotential of Bunges work to the discipline of InformationSystems:

We have chosen to work with Bunges ontology becauseit deals directly with concepts relevant to the informationsystems and computer science domains (e.g. systems,subsystems and couplings). Moreover, Bunges ontology isbetter developed and better formalized than any others wehave encountered. (Wand and Weber, 1995, p. 209)

While this justification in itself bears little merit in terms

of a rigorous motivation for the use of this ontology, itsadoption by Wand and Weber (1990b; 1993; 1995)nevertheless facilitated a wider uptake of this theoreticalmodel within the Information Systems community. TheBunge-Wand-Weber models actually comprise three models(Wand and Weber, 1995; Weber, 1997), viz., therepresentation model, the state-tracking model and thedecomposition model. The representation model is typicallyused for the evaluation of process modelling techniques.

Wand and Webers work based on Bunges theory is notthe only case of ontology-based research. Today, interest in,and applicability of ontology, extends to areas far beyondconceptual modelling. The usefulness of ontology as atheoretical foundation for knowledge representation andnatural language processing is a fervently debated topic inthe artificial intelligence research community (Guarino andWelty, 2002). Holsapple and Joshi (2002) for example,argue the importance of ontology in the emergent era ofknowledge-based organizations and the conduct ofknowledge management in those organizations. Kim (2002)shows how ontology can be engineered to support the firstphase of the evolution of the semantic Web. As theseselected examples show, the theory of ontology hasemerged as a fruitful base from which a wide range oftheoretical contributions in the IS and related disciplines

stem from.

It must be noted, however, that the use of ontology in ISresearch is not undisputed. The BWW model, for instance,has sometimes been subjected to criticism (for acomprehensive discussion refer, for instance to (Kautz etal., 2006). We do not wish to engage in this debate andinstead merely point to the various examples of testing andvalidation that the model has undergone over time (Wandand Weber, 2006). In conclusion, its widespread adoptionand extent of validation have motivated our selection of theBWW model.

Independent from the stream of ontological research, awide variety of process modelling techniques has beendeveloped over the years based on different theoreticalfoundations and for different purposes. The process ofdeveloping a new process modelling technique is lessunderstood than the classical software engineering processand can be regarded as not sufficiently researched. In theprocess of requirements engineering for Process-awareInformation Systems, for example, various stakeholders are

confronted with the need to represent the requirements in aconceptual form (Dumas et al., 2005). Often, however, theyuse modelling techniques that do not possess an underlyingconceptual structure on which to base such models or thathave been proposed with limited consistent theoreticalfoundation underlying their conception or development(Floyd, 1986; van der Aalst, 2003). Moreover, the widerange of purposes with demand for process modellingtechniques (e.g., process improvement, Web services,Enterprise Systems implementations, workflowmanagement, compliance management) and commercialopportunities in this area further contributed to an inflation

of individually developed modelling techniques andcorresponding tools. Clearly, what is needed is a theoreticalbasis to assist as a reference benchmark in the evaluationand comparison of available process modelling techniques.Given the existence of such theory, it would not only bepossible to evaluate these techniques, but also to determineif the discipline of process modelling is building onprevious knowledge, and if new techniques really denote anactual improvement.The process of using the BWW representation model as a

type of reference benchmark for the evaluation of therepresentational capabilities of a modelling technique formsthe core of the research method of representational analysis.In this process step, the constructs of the BWWrepresentation model (e.g., thing, event, transformation) arecompared with the language constructs of the processmodelling technique (e.g. event, activity, actor). The basicassumption is that any deviation from a 1-1 relationshipbetween the corresponding constructs in the representationmodel and the modelling technique leads to a situation ofrepresentational deficiency and/or ambiguity in the use ofthe technique potentially causing confusion to the end users.The set of BWW constructs is well defined in various

languages. Wand and Weber (1990b), for instance, use settheory to formalize the set of constructs, and Rosemann and

Green (2002) developed a semi-formal description of the setof BWW constructs by means of a meta-model using the

7/30/2019 c26433

6/17


Extended Entity-Relationship (EER) modelling notation(Chen, 1976). We do not seek to recite the specification ofthe representation model in this article and instead refer theinterested reader to the extensive formalization described,for instance, in (Wand and Weber, 1990b; 1995; Weber,1997).

While the model itself is well defined, the process ofapplying the representation model constructs as part of arepresentational analysis is less specified. It was onlyrecently that more advanced procedural models have beenproposed (Rosemann et al., 2005) that guide researchersthrough the process of comparing the representationalmodel with the selected modelling technique. Again, for thepurpose of this paper we do not wish to formally specify theprocess of analysis and merely note that this work has beencarried out, e.g., (Rosemann et al., 2004; Gehlert andEsswein, 2007).

In general terms, representational analysis focuses onmapping relationships that are not 1:1. Such cases are

classified as theoretical, i.e., potential, representationalshortcomings. These undesirable situations can be furthercategorized into the following four types, as shown inFigure 1 (Weber, 1997):

construct overload describes a situation in which aconstruct in the process modelling technique representstwo or more representation model constructs(m:1 relationship),

construct redundancy is the opposite case, i.e., oneconstruct in the representation model is depicted by twoor more constructs in the process modelling technique

(1:m relationship), construct excess is the case in which at least one

construct in the process modelling technique does notmap to any construct in the representation model(0:1 relationship), and

construct deficitdescribes the case in which at least oneconstruct in the representation model does not map toany construct in the process modelling technique(1:0 relationship).

**********************************Figure 1 approximately here**********************************

Based on these four types of undesirable situation, it ispossible to make predictions as to the representationalcapabilities of a process modelling technique for providingcomplete and clear representations of the domain beingmodelled (Weber, 1997). In particular, if a processmodelling technique provides constructs for each element ofthe representation model, i.e., construct deficit is notpresent, it is regarded as ontologically complete. In turn, theontological clarity of a process modelling technique can bemeasured by the degrees of construct overload, constructredundancy, and construct excess.

At this stage of the research progress, theserepresentational issues are of theoretical nature. Moreexplicitly, these theoretical findings denote potential issuesfor modelling users working with these process modellingtechniques. Most of the related research is exclusivelyfocused on this step within the research model. Theidentified potential issues, however, require furtherempirical testing. For this purpose they have to betransformed into propositions and testable hypotheses. Asan example, a recent representational analysis (Recker et al.,2006) identified construct deficit in the BPMN modellingtechnique with regards to representing aspects and conceptsof states (e.g., stable states, state laws, conceivable statespaces, etc.). This deficit has led to the proposition thatmodellers using BPMN would have difficulty capturing thebusiness rules of a given situation, as business rules demandrigorously specified state and transformation laws. Such aproposition can be converted into a testable hypothesis thatin turn can be confirmed or falsified via empirical testing.

Empirically validating the theoretically identifiedrepresentational issues requires access to sources ofevidence. In most cases, this requirement will meaninterviews and occasionally focus groups, or experiments,with business analysts or experienced students as proxieswhen access to practitioners is too difficult to organize.Surveys as a way of collecting related data are anotheroption, but in reality researchers often struggle to identifythe required number of participants for such a study. In ourexperience, predominantly semi-structured interviews havebeen used as an empirical research method in the process ofrepresentational analysis (Green and Rosemann, 2001;

2002; Davieset al., 2004; Recker et al., 2006).The design of such interviews should follow a definedprotocol that explores the significance of the identifiedissues. As Figure 2 indicates, five different levels of severityof an issue can be differentiated, with only level Vrepresenting a serious issue. Further follow-up questionswould explore how the interview partner addresses thiscritical problem (e.g. free form annotations of the model,use of complementary techniques, etc.) and, in general,allow for extended reasoning and the exploration of furtherfactors that may have remained invisible from thetheoretical analysis.


Gathered responses can be further codified based oncollected demographic data. Further investigations intopossible correlations will provide insights into the extent towhich a perceived problem corresponds, for example, withyears of experience in business process modelling. Previousresearch (Rosemann and Green, 2000; Davies et al., 2004)has shown that besides the characteristics of the person whouses the process modelling technique (e.g., self-efficacy) thepurpose of process modelling (e.g., for workflowengineering, Enterprise Systems implementation,

7/30/2019 c26433

7/17


compliance management) is a primary contextual factorwith impact on the actual perceived criticality of theidentified issue. The need for including these contextualfactors in a representational analysis stems from theobservation that different modelling objectives, purposes,cognitive abilities as well as other factors impact on the wayprocess modelling technique users conceptualize theirrelevant real-world domain.

Overall, the subset of theoretical representational issuesthat is evaluated as critical in light of the empirical databecomes the set of empirically validated representationalshortcomings and is further classified with contextualfactors (e.g. person, purpose). The earlier stated proposition,for example, was confirmed using semi-structuredinterviews. In this particular case, 75 percent ofinterviewees that had a need to model business rules statedthat they were unable to do so with BPMN. Follow-upquestions further revealed that practitioners employedworkarounds, such as narrative descriptions of business

rules, spreadsheets, tables, and even UML state diagrams, inorder to compensate for this deficit (Recker et al., 2005;2006).

Such identified and validated issues that stem fromrepresentational analysis form an important input for furtherrevisions and improvements of existing process modellingtechniques. In this phase, it is important to communicate thetheoretical and empirical research outcomes back to thedevelopers of the modelling technique or tool providers.The assumption of this type of research is that theoreticallyidentified, empirically validated and communicated issuesrelated to a process modelling technique have the potential

to guide revisions of these techniques and ultimately lead toan increased quality of process models. The availability ofimproved models provides the involved stakeholders withbetter opportunities to achieve the goals underlying theprocess modelling initiative. From a research viewpoint, thisis the phase in which methods for the evaluations of thequality of a process model can be utilized (e.g., based on thesemiotic framework (Lindland et al., 1994; Krogstie et al.,2006)). Furthermore, improved model quality has to beplaced in the context of the critical success factors ofprocess modelling in order to have a realistic understandingof the impact of an improved modelling technique on theoverall success of process modelling. In this context, theimpact of the factor modelling language to the overallsuccess of process modelling initiatives was found to besignificant (Bandara et al., 2005; Bandara and Rosemann,2005). Therefore, an improvement in the modellingtechnique will contribute ultimately to the success of theoverall process modelling effort.

Figure 3 summarizes the phases of this type of researchproject.


Figure 3 also shows how we have attempted to follow aKuhnian approach to scientific method in our work.According to Kuhn (1962), scientific method is the processby which scientists, collectively and over time, endeavour toconstruct an accurate (that is, reliable, consistent and non-arbitrary) representation of real-world phenomena.Ultimately, the real-world phenomenon that we areinterested in is the quality of model produced, in this case,the quality of the business process model produced. In otherwords, the ultimate dependent variable of interest to ourwork is quality of the model produced. The items throughwhich we measure quality include, inter alia, level ofcompleteness, ambiguity, and understandability. Ourunderlying assumption is that, if the modelling technique(s)used to produce the models is representationally incomplete,representationally ambiguous, and/or provides symbols thatare not clearly understandable, then, ceteris paribus, themodels produced using those techniques will suffer thesame problems. Our theoretical foundation, the BWW

models, informs us that a significant intervening variablepotentially influencing our dependent variable is therepresentational capability of the modelling technique. Wemeasure this intervening variable through the levels ofrepresentational completeness and clarity of a modellingtechnique. The levels of these measures are predictedthrough the representational analysis of the modellingtechnique and are then confirmed through empirical testingwith users of that particular technique. We also realize thatthere are significant moderating variables that will impactthe optimal level of representational capacity required in amodelling technique, viz., the role of the person doing the

modelling, and the purpose for which they are doing themodelling. Finally, suggestions to improve the modellingcapacity of the technique can be made based on theempirically confirmed measures. Resultant business processmodels produced using the revised modelling technique canthen be assessed in terms of their improved quality. Indeed,this aspect of the research program leads us to a veryexciting phase whereby we hope to be able to test theimproved business process model quality resulting fromrevisions to the modelling technique provided by the resultsof our evaluation of the representational capacity of theoriginal modelling technique. In our work with the BusinessProcess Modelling Notation (BPMN), for example, we havebeen in contact with the developers of that business processmodelling technique and we have communicated with themregarding the results of our empirical testing of propositionsresulting from our analysis of that technique. It will be aunique opportunity for testing improvements in businessprocess model quality ultimately if we see future versions ofBPMN reflecting the results of our original analytical work.

7/30/2019 c26433

8/17


3 THE BWW REPRESENTATION MODEL & RELATED

WORK

To fully appreciate the overall research model underlyingrepresentational analyses, it is necessary to gain anunderstanding of the underlying theory.The development of the representation model that is

known as the Bunge-Wand-Weber model emerged from theobservation that, in their essence, Information Systems arerepresentations of real world systems. Real world systems,in turn, can be explained and described using ontology - thestudy of the nature of the world and attempt to organize anddescribe what exists in reality, in terms of the properties of,the structure of, and the interactions between real-worldthings (Bunge, 1977, pp. 3-6; Shanks et al., 2003).

Concerned that the lack of theoretical foundationsunderlying modelling method development would result inthe development of Information Systems that were unable tocompletely capture important aspects of real world systems,

Wand and Weber (1989; 1990b; 1993; 1995) developed andrefined a set of models for the evaluation of modellingtechniques and the scripts prepared using such techniques.The BWW representation model is one of three theoreticalmodels defined by Wand and Weber (1995) that make upthe BWW models. The application of the representationmodel to Information Systems foundations has been referredto by a number of researchers (Green and Rosemann, 2004)and the model is now often referred to as simply the BWWmodel. Some minor alterations have been performed overthe years by Wand and Weber (1993; 1995) and Weber(1997), but the current key constructs of the BWW model

can be grouped into the following clusters: things includingproperties and types of things; states assumed by things;events and transformations occurring on things; and systemsstructured around things (Green and Rosemann, 2005;Rosemann et al., 2006). For a complete definition of theconstructs refer to, for example, (Weber, 1997).The BWW model has over the years reached a significant

level of maturity, adoption and dissemination, and has beenused in over thirty research projects for the evaluation ofdifferent modelling techniques, including data models(Wand and Weber, 1993), schema models (Weber andZhang, 1996), object-oriented models (Opdahl andHenderson-Sellers, 2001; 2002) and reference models(Fettke and Loos, 2007). It also has a strong track record inthe area of process modelling, with contributions comingfrom various researchers (Rosemann et al., 2006). In thissection, we briefly summarize those BWW related studiesthat focus specifically on process modelling techniques.

Keen and Lakos (1996) determined essential features for aprocess modelling scheme by evaluating six processmodelling techniques in a historical sequence by using theBWW representation model. Among the modellingtechniques evaluated were: ANSI flowcharts (AmericanNational Standards Institute, 1970), Data Flow Diagrams(DFD) (Gane and Sarson, 1979) and the IDEF Method 3

Process Description Capture Method (Mayer et al., 1995)and the Language for Object-Oriented Petri nets

(LOOPN++) (Keen and Lakos, 1994). The evaluation isrestricted to the assessment of the ontological completenessof each technique and the authors did not empirically verifytheir findings on the features of process modelling schemes.From their analysis, Keen and Lakos concluded that, ingeneral, the BWW representation model facilitates theinterpretation and comparison of process modellingtechniques. They propose the BWW constructs of system,system composition, system structure, system environment,transformation, and coupling to be essential processmodelling technique requirements. However, we found thatthese findings are not entirely reflected in the leadingprocess modelling techniques (Rosemannet al., 2006).

Green and Rosemann (2000) analyzed the Event-drivenProcess Chain (EPC) notation (Keller et al., 1992; Scheer,2000) with the help of the BWW model, assessing bothontological completeness and clarity. Their findings havebeen empirically validated through interviews and surveys(Green and Rosemann, 2001; 2002; Davies et al., 2004).

Confirmed shortcomings were found in the EPC notationwith regard to the representation of real world objects andbusiness rules, and in the thorough demarcation of theanalyzed process.

Green et al. (2005) examined the Electronic Businessusing eXtensible Markup Language Business ProcessSpecification Schema (ebXML BPSS) v1.01 (OASIS, 2001)in terms of ontological completeness and clarity. While theempirical validation of results has not yet been performed,the analysis shows a relatively high degree of ontologicalcompleteness of ebXML.

Green et al. (2007) also compared different modelling

standards for enterprise system interoperability, includingBusiness Process Execution Language for Web Servicesv1.1 (BPEL4WS) (Andrews et al., 2003), Business ProcessModelling Language v1.0 (BPML) (Arkin, 2002), WebService Choreography Interface v1.0 (WSCI) (Arkin et al.,2002), and ebXML. These standards, which proclaim toallow for specification of intra- and inter-organizationalbusiness processes, have been analyzed in terms of theirontological completeness. The study found that ebXMLprovides a wider range of language constructs forspecification requirements than other techniques, indicatedthrough its comparatively high degree of ontologicalcompleteness. In addition, a minimal ontological overlap(MOO) analysis (Wand and Weber, 1995; Weber, 1997)was conducted in order to determine the set of modellingstandards with a minimum number of overlappingconstructs but with maximal ontological completeness(MOC), i.e., maximum expressiveness. The study identifiedtwo sets of standards that together allow for the mostexpressive power with the least overlap of constructs, viz.,ebXML and BPEL4WS, and, ebXML and WSCI. At thepresent point in time, this analysis too, has not yet beenempirically validated.

Recker et al. (2005) used representational analysis toidentify shortcomings in the Business Process Modeling

Notation (BPMN) v1.0 (BPMI.org and OMG, 2006b) fromthe viewpoint of both clarity and completeness. Theoretical

7/30/2019 c26433

9/17


findings were tested with nineteen practitioners from sixAustralian organizations, resulting in the finding that therewould appear to be representational shortcomings in BPMN,for example, in the modelling of business rules, and theusage of the Lane and Pool constructs (Recker et al., 2006).

Recker and Indulska (2007) examined Petri nets (Petri,1962) from the viewpoint of ontological completeness andclarity. The focus on Petri nets in their study stemmed fromthe observation that, while originally being a specificationdevoted to system dynamics, Petri nets have over timefound adoption in regular process modelling communities,which in turn raises a need for evaluating theirrepresentational capacity. The study found that,surprisingly, the notation that consists of a mere sevenconstructs, provides a relatively high degree of ontologicalcompleteness. Deficits were for instance found in thesupport for the modelling of systems structured aroundthings. The study also found that the same flexibility thataffords Petri nets a higher ontological completeness, also

results in extensive construct overload (for instance a placeconstruct in a Petri net can be used to represent a thing,class, or state). The results from this analysis have not yetbeen tested with practitioners.

Overall, most of the conducted research has been of apurely theoretical nature. Most of the evaluations lack, atthe time of writing, empirical verification of the theoreticalfindings. While this may in general be seen as problematic,the available theoretical studies nevertheless allow us tocompare the results of these theoretical analyses over timein order to understand if process modelling techniques areactually improving in terms of their representational

capabilities.

4 THE PROCESS MODELING DISCIPLINE SIGNS OF

MATURITY?

While representational analyses of process modellingtechniques per se provide means for exploring strengths andweaknesses of these techniques, they can also be used as ameans for assessing the evolution of technique developmentover time. As the process modelling discipline evolved onlyrecently as a dedicated research field, we were curious

whether this emerging research field would follow theoverall guideline of establishing and building on acumulative tradition (Keen, 1980). Our motivation then wasto study the development of process modelling techniquesover time, using representational capabilities as ameasurement of increasing maturity.

We assessed the outcomes of twelve previous ontologicalanalyses of process modelling techniques and put them in ahistorical sequence, starting with the analysis of Petri nets inits original and most basic form (Petri, 1962). Petri netswere chosen as we perceive it to be the intellectualbirthplace of more rigorous and disciplined processmodelling.

For comparison purposes the focus of this study wasontological completeness only. The notion of ontological

completeness of a particular process modelling techniqueserves as an indication of its representational capabilities,being the extent to which the techniques are able to providecomplete descriptions of a real-world domain. See (Reckeret al., 2009) for more details on this study.The consolidation of the previous representational

analyses leads to several interesting results. A longitudinalstudy of the ontological completeness shows an obviousincrease in the coverage of BWW constructs. This findingcan be interpreted as a sign of increasing representationaldevelopment over time. Figure 4 visualizes this trend overtime, as measured by the number of BWW constructscovered by each technique.


We can see that, while the original Petri net specificationdid not provide exceptionally good representationalcoverage (41%) as defined by the BWW representationmodel, it still performed better than more recent techniquessuch as DFD (28%) or IDEF3 (38%). A noticeable spike inFigure 4 depicts the high level of development (in terms ofontological completeness) of the ebXML standard (76%). Itis interesting to note that ebXML is specified in UML(OASIS, 2001), with a semi-formal construct definition anddescription, whereas BPEL4WS, WSCI, and BPMN, forexample, have textual specifications supplemented bydiagrams of examples. As such, the ebXML specification isless subjective in its possible interpretations. BPMN, too,

appears to perform very well (66%) and hence appears to bequite mature in terms of representation capabilities. Thishigher level of ontological completeness can perhaps partlybe explained by the fact that previous approaches, includingEPC and Petri nets, influenced the development of BPMN(BPMI.org and OMG, 2006b).

It appears in general that techniques that focus ondescribing process flow from a business perspective (forinstance DFD and IDEF3) are less ontologically completethan those that have to cater for more syntactical rigor dueto their focus on executability or translatability intoexecutable techniques (such as BPEL4WS or ebXML).

However, the DFD and IDEF3 techniques were developedsome time ago. Overall there is an upward trend in terms ofthe representational capability of the analyzed techniques.Indeed, it is clear that new techniques are building on thecapabilities of the previous techniques. BPMN specificallyhas been designed by its authors with consideration ofprevious techniques and their advantages (BPMI.org andOMG, 2006b, p. 1).

7/30/2019 c26433

10/17


5 REPRESENTATIONAL ANALYSES: FACTORS FOR

CONSIDERATION

A representational analysis of process modelling techniqueshas two facets. On the one hand, it provides insights intopotential issues with a modelling technique. On the otherhand, it can also contribute to the further development of theselected theoretical model. In fact, our experiences with theapplication of the BWW models and our findings from thelongitudinal analysis of process modelling techniques alignwith some of the previous criticisms of representationalanalyses (Rosemann et al., 2004). In the following sub-sections, we discuss selected limitations and pitfalls ofrepresentational analyses and some approaches to avoidingthem.

5.1 Relevance of the underlying representation model

The fact that even the most mature process modelling

technique (ebXML) supports only 76% of the BWWconstructs may perhaps suggest that the selected theorymight be too demanding. Being an upper-level theory ofrepresentation, different application domains may induce aneed for refining the set of constructs contained in themodel to more relevant subtypes specific to the givendomain. Regarding this potential lack of relevance of theBWW models, we suggest the development of a weightedscoring model that takes into consideration the specificneeds of the BPM domain and assigns relative importanceto constructs accordingly. All BWW-based analyses so farhave been performed based on the assumption that

representational capability for each of the BWW constructsis equally important. However, each domain might havedifferent needs regarding the expressive power of modellingtechniques, and therefore differing levels of importance forrepresentation of various situations. For example,transformations are critical in the BPM domain, see also(Soffer and Wand, 2005), and by far outweigh theimportance of being able to represent the history of statechanges. For empirical evidence supporting this propositionplease refer, for example, to (Recker et al., 2006). Suchdifferences should be recognized in the relative weighting ofthe BWW constructs for a given domain. With a developedand tested scoring model, it would then be possible to arriveat a final numeric score that measures the goodness of theanalyzed technique or the significance of the identifiedissues. This score may then be easily compared with thescores of other notations, thereby providing an easy andobjective way to choose a technique or notation that excelsin areas of representation that are deemed important in agiven domain.The current BWW representation model also needs to be

investigated in order to determine if there are constructs thatmay require further specialization. The need for suchspecialization has been noted by Weber (1997, p. 99)himself and has in some domains already been carried out,

see, for instance, (Guizzardi, 2005, pp. 135 ff.).

For example, our empirical studies suggest that events andtransformations occurring on things may require furtherspecialization. BPMN, for example, distinguishes betweennine event types, representing a differentiation scheme thatis not covered by the BWW constructs of event and itssubtypes. The same can be seen in standards such asebXML, BPEL4WS, BPML, and WSCI. A similar situationholds for the transformation construct that we often found tobe susceptible to construct redundancy. For example, inBPML there are ten language constructs representingdifferent types of transformations. A similar situation existsin standards like BPEL4WS and ebXML. This situationimplies that, just as properties in the BWW representationmodel are specialized, perhaps transformations should be aswell for the domain of BPM.

Also, it is interesting to note that throughout the analysesof process modelling techniques, control flow mechanismssuch as logical connectors, selectors, gateways and the likeare repeatedly regarded as construct excess as they do not

map to any construct of the BWW representation model.However, these constructs are agreed to be essential to theBPM domain (for empirical evidence supporting thisproposition refer to (Recker et al., 2006)). While one mightargue that an extension of the existing BWW model mightbe perceived as being required to deal with such issues inthe BPM domain, this would contradict the soundphilosophical nature of the theory. As such, extensionsshould be avoided and instead complementary analyses,e.g., using workflow patterns (van der Aalst et al., 2003),should be conducted.

5.2 Subjectivity in the process of analysis

Another problem is related to the anecdotal criticism of thesubjective nature of the analysis in which propositions aredeveloped principally from the perceptions and experienceof the researcher in the use of the BWW models in theanalysis. This concern is conceded also by Weber (1997, p.94) who contends that two individuals might look at thesame grammar and see two different sets of grammaticalconstructs. Moreover, one persons perception of amapping between an ontological construct and agrammatical construct might not be the same as anotherpersons perception (Weber, 1997, p. 94).

Accordingly, the representation mapping step should beconducted in multiple iterations between variousresearchers. In a first step, the mapping should be performedby each involved researcher. The researchers should thenmeet to discuss and defend their respective mappings,leading to a second joint mapping draft. This joint draftshould then be presented, discussed and defended in front ofthe complete research team or even experts in the field ofrepresentational analyses, thereby leading to a final, agreeddraft. In previous studies, we were able to increase theextent of agreed mappings, i.e., construct mappings that areagreed upon by every involved researcher, over three

iterations. This process in turn served as an indication of theobjectivity of the mapping results. While our previous

7/30/2019 c26433

11/17


studies have used a percentage agreement calculation,Cohens Kappa (Cohen, 1960) should be used in order tomeasure the level of agreement in research teams thatconsist of multiple individual coders.

Subjectivity in such analyses can further be reduced withthe use of meta-models. If a meta-model exists for thechosen representation model, and if the target technique ornotation is specified in the same language as the meta-model, then it is possible to perform a more objectiveassessment that is based on pattern matching between themeta-models of the representation model and the technique.As previously stated, an EER (Extended Entity Relationshipmodel) based meta-model for the BWW representationmodel has already been developed (Rosemann and Green,2002). This model can be used to more easily andobjectively analyze techniques and notations specified inEER or ER. There is a need to translate this model intoUML, as this notation tends to be the standard used inspecifications currently. There is also a need for the authors

of techniques and notations to realize the importance of amore formal specification, rather than a text-based one.While ebXML, for example, has a semi-formal, UML-basedspecification (OASIS, 2001), and could therefore be moreobjectively analyzed with a UML-based BWW meta-model,many other notations (e.g., BPMN) are informally specified.The need for meta-model-based specification, however, isincreasingly recognized by technique developers. TheBPMN development team, for example, is currentlyworking on a UML-based meta-model (BPMI.org andOMG, 2006a).

5.3 Multi-level analysis of modelling capabilities

The majority of representational analyses of modellingtechniques is based on the (textual, semi-formal or formal)specification of the modelling technique, e.g., in form of ameta-model. The actual application of a modellingtechnique, however, has to be also seen from the viewpointof the supporting modelling tool and the companysindividual modelling conventions. Thus, potentially, threelevels for a representational analysis can be differentiated:a) the (classical) level of the modelling techniqueb) the way a certain modelling tool supports a modelling

techniquec) the company-specific way in which a tool-supported

technique is constrained by internal conventions.

The current focus on the first level of analysis providesimportant insights into potential and actual issues with amodelling technique. However, we have found that many ofthese issues are overcome by a modelling tool that mightaddress these issues (Recker et al., 2006). For example,some process modelling techniques do not provideconstructs for decomposing or structuring the modelledsystem or process. While this denotes a theoreticalrepresentational shortcoming, we have found that in practiceit is not perceived as such due to complementary toolsupport, for example, by means of storing and linking inter-related process models in a repository. Individual modelling

conventions provide another opportunity to addresspotential sources of ambiguity. An example of this is the useof colour-coding schemes to differentiate languageconstructs (e.g., coloured activity symbols to differentiateautomated from manual tasks) in order to overcome thetheoretical shortcoming of construct overload.Thus, a representational analysis that only focuses on the

level of a generally specified modelling technique does notconsider these two additional levels, which are relevant inthe practical application of process modelling.

5.4 Terminology employed in theory and practice

Empirical testing of representational analyses can in somecases be problematic due to the differences in terminologyused by researchers and BPM practitioners. Theterminology of the BWW models is concerned with abstractand formal specifications of high-level constructs that maynot be easily, or correctly, understood without prior

exposure to the model and its specification. The design ofcomplementary empirical research strategies that seek totest theoretical propositions thus has to take into account theproblem of translating the terminology of the BWW modelsto the appropriate domain and the users within. As anexample, our representational analysis of BPMN revealed alack of constructs for representing concepts such asconceivable state spaces and lawful event spaces. Whendeveloping an accordant interview protocol for testing thepropositions derived from these findings, we had to placeemphasis on communicating and explaining the detecteddeficiencies in a terminology close to the one used by

modelling practitioners instead of asking users, for example,Can you model conceivable state spaces?

5.5 The true nature of a 1:1 relationship

As outlined in earlier sections, the idea of a representationalanalysis is based on the assumption that any relationshipbetween corresponding constructs of the representationalbenchmark and the selected modelling technique, which isnot a 1:1 relationship, leads to potential ambiguity orincompleteness. However, the notion of a 1:1 relationship isnot precise and the notion of the entirety of one constructrequires further specification. Instead of asking, Is there asymbol in the modelling technique, which supports arepresentational construct? a more detailed study couldexplore the nature of the representational construct. Forexample, just because the BWW construct thing issupported in the Event-driven Process Chain by anOrganizational Unit (Green and Rosemann, 2000) doesnot mean that the EPC is in this regard ontologicallycomplete. Instead, it is necessary to identify all relevantfacets of a thing, and then it becomes obvious that an EPCcannot model, for example, individual instances ofdocuments which are transformed within a process. In fact,the construct Organizational Unit represents a subtype of

a thing, more specifically an indirect subtype thatspecializes the notion of Agentive Physical Entity (Masolo

7/30/2019 c26433

12/17


et al., 2003). As a consequence, there is a need not only tofurther empirically test the actual significance of the non-1:1-relationships, but at the same time evaluate if a 1:1relationship truly represents a corresponds withrelationship or if it is maybe more a is-a relationship,which supports only one specialization of the construct.This latter case again hints at a potential need to specializecertain constructs in the underlying representation model.

5.6 Involving designers of the modelling technique

Only a limited number of research projects that evaluatedthe representational capabilities of modelling techniques,has applied empirical analysis in order to verify theidentified issues. In these few cases, either businessanalysts, e.g., (Davies et al., 2004; Recker et al., 2006), orexperienced coursework students, e.g., (Green andRosemann, 2001), have been used.

What so far has been missing in all considered

representational analyses is the discussion of the confirmedanalysis findings with the designers of the respectivemodelling technique. In fact, it is a general trend in ISresearch that assessment usually stops as soon asweaknesses in the theory or the developed artefact havebeen identified. Hevner et al. (2004, p. 80) note hereto thatthe refinement and reassessment process is typicallydescribed in future research directions. However, whatgood can findings from extensive evaluation strategies do ifthey are not incorporated in the re-design, extension orimprovement of the evaluated artefact? In order tocounteract this limitation we sought to involve the designers

of the BPMN modelling technique in our study of itsrepresentational capabilities and conducted telephoneinterviews with the development team in which we advisedthem and discussed with them the identified representationalissues. The main aim of these communications was to focuson how future revisions of the technique specification mayovercome the identified representational shortcomings (seealso the second last step in the research model in Figure 3).

6 CONTRIBUTIONS & CONCLUSIONS

This paper discussed the use of an ontology-based theory ofrepresentation for the study of process modellingtechniques. We gave insights into the process ofrepresentational analyses, explored some related perils andpitfalls and also showed benefits that can be obtained fromthese studies. For example, we discussed how the outcomesof representational analyses can be of interest to thedevelopers and users of process modelling techniques.Developers of process modelling techniques should bemotivated to examine representational analyses of currentlyused process modelling techniques in order to build uponthese and counteract any weaknesses in newly developedtechniques or technique extensions. On the other hand, users

of process modelling techniques might be motivated to useontological completeness and clarity as potential evaluation

criteria for the selection of a most appropriate modellingtechnique.

We also presented a longitudinal study of representationalanalyses of process modelling techniques based on thereview of previous studies in the field. The findings clearlyshow signs of a maturing modelling discipline, as measuredby an increased ontological completeness of processmodelling techniques over time. The results also identify thecommon core constructs of process modelling techniques(for example, transformation, properties, events) as well astheir key differentiators (for example, subsystem, systemenvironment, lawful state space). Furthermore, the findingsprovide valuable insights for the future application of theBWW model as a benchmark for such analyses ofmodelling techniques. We hope that such research mightalso motivate other researchers to conduct a similar studyfor data or object-oriented modelling techniques.

Forthcoming from the review of related studies and ourown experiences, we foresee significant opportunities for

further research in the area of representational analysis ofprocess modelling techniques. Accordingly, in theconclusion of this paper we would like to share somethoughts on future research challenges in this area.There is evidence, based on our experience, that there is a

need for a specialization of certain constructs of the BWWmodel and for the development of a weighted scoring modelfor the domain of BPM. While the BWW model can be seenas one way of conceptualizing the things and systems in theworld (and the interactions between those things) that arerelevant to Information System domains, a question remainswhether the BPM domain itself has a need to do the same.

In our analyses we have found that there are someconstructs of the BWW model that are supported by onlyone technique of the chosen twelve, for example the BWWconstructs kind and history. While this might indicate anarea of improvement of the representation power of processmodelling techniques, it might also indicate that, perhaps,the particular BWW construct is not necessary formodelling in the domain of BPM. Such issues requirefurther empirical testing in order to determine whether theconstructs are indeed required. Based on such furthertesting, it will be possible to develop a weighted scoringmodel that assigns higher weights to the critical constructs,lower weights to constructs that are not detrimental tomodelling business processes, and zero weights toconstructs that are not deemed to be necessary for the BPMdomain. At the same time, the BWW model should befurther studied in light of the empirical testing, to determinewhether there is a need to specialize any constructs intomore differentiated sub-types (e.g., transformation as ourstudy would indicate).

Furthermore, we also see strong benefit in performing astudy of ontological clarity of process modelling techniquesover time, as a complement to ontological completeness.This should be followed by experiments that aim to assessthe impact of completeness and clarity on the intuitiveness

of the model, similar to (Gemino and Wand, 2003; 2005),and a study of the impact of representational capabilities on

7/30/2019 c26433

13/17


the uptake and acceptance of process modelling techniquesin modelling initiatives (Recker, 2008). We anticipate thatthese types of study will contribute to developing acomprehensive understanding of the quality of a processmodel produced, incorporating syntactical (for example,consistency of the language meta-model), semantic (forexample, representational shortcomings in a technique) andpragmatic aspects (for example, the impact of ontologicalclarity on the perceived intuitiveness of the modelproduced).

As a concluding remark we would like to add that we havefound the research method of representational analysis veryfruitful in understanding and exploring the challenges ofprocess modelling and we expect this type of research tocontinue to give stimulating input to both academic andpractical work in the area of process modelling in the future.

REFERENCES

American National Standards Institute (1970)ANSI StandardFlowchart Symbols and their Use in Information

Processing (X3.5),American National StandardsInstitute, New York, New York.

Andrews, T., Curbera, F., Dholakia, H., Goland, Y ., Klein, J .,Leymann, F., Liu, K., Roller, D., Smith, D., Thatte, S.,

Trickovic, I. and Weerawarana, S. (2003) BusinessProcess Execution Language for Web Services. Version1.1, URL http://xml.coverpages.org/BPELv11-May052003Final.pdf, accessed February 10, 2006.

Aristotle (1991)The Metaphysics. Translated by J ohn H.

McMahon, Prometheus Books, Buffalo, New York.Arkin, A. (2002) Business Process Modeling Language, URLhttp://www.bpmi.org/, accessed January 16, 2006.

Arkin, A., Askary, S., Fordin, S., Jekeli, W., Kawaguchi, K .,Orchard, D., Pogliani, S., Riemer, K., Struble, S.,

Takacsi-Nagy, P., Trickovic, I. and Zimek, S. (2002)Web Service Choreography Interface (WSCI) 1.0, URLhttp://www.w3.org/TR/wsci/, accessed January 16, 2006.

Bandara, W., Gable, G.G. and Rosemann, M. (2005) Factors andMeasures of Business Process Modelling: ModelBuilding Through a Multiple Case Study,European

Journal of Information Systems,14, 347-360.Bandara, W. and Rosemann, M. (2005) What Are the Secrets of

Successful Process Modeling? Insights From anAustralian Case Study,Systmes d'Information etManagement,10, 47-68.

BPMI.org and OMG (2006a) BPMNModel UML Documentation,URLhttp://www.bpmn.org/Documents/BPMNMetaModel.zip, accessed March 10, 2006.

BPMI.org and OMG (2006b) Business Process Modeling NotationSpecification. Final Adopted Specification, URLhttp://www.bpmn.org, accessed February 20, 2006.

Bubenko, J.A. (1986) "Information Systems Methodologies - AResearch View" in T. W. Olle, H. G. Sol and A. A.

Verrijn-Stuart (eds.) Information Systems Design

Methodologies: Improving the Practice, North-Holland,Amsterdam, The Netherlands, pp. 289-318.

Bunge, M.A. (1977)Treatise on Basic Philosophy Volume 3:Ontology I - The Furniture of the World,KluwerAcademic Publishers, Dordrecht, The Netherlands.

Bunge, M.A. (1979)Treatise on Basic Philosophy Volume 4:Ontology II - A World of Systems,Kluwer AcademicPublishers, Dordrecht, The Netherlands.

Chen, P.P.-S. (1976) The Entity Relationship Model - Toward aUnified View of Data, ACM Transactions on DatabaseSystems, 1, 9-36.

Cohen, J. (1960) A Coefficient of Agreement for Nominal Scales,Educational and Psychological Measurement, 20, 37-46.

Davies, I., Green, P., Rosemann, M., Indulska, M. and Gallo, S.(2006) How do Practitioners Use Conceptual Modelingin Practice?, Data & Knowledge Engineering, 58, 358-380.

Davies, I., Rosemann, M. and Green, P. (2004) "ExploringProposed Ontological Issues of ARIS with Different

Categories of Modellers" in15th AustralasianConference on Information Systems, AustralianComputer Society, Hobart, Australia.

Dumas, M., van der Aalst, W.M.P. and ter Hofstede, A.H.M.(Eds.) (2005) Process Aware Information Systems:Bridging People and Software Through Process

Technology,John Wiley & Sons, Hoboken, New Jersey.Fettke, P. and Loos, P. (2007) Ontological Evaluation of Scheers

Reference Model for Production Planning and ControlSystems,J ournal of Interoperability in BusinessInformation Systems,2, 9-28.

Floyd, C. (1986) "A Comparative Evaluation of System

Development Methods" in T. W. Olle, H. G. Sol and A.A. Verrijn-Stuart (eds.) Information SystemDesignMethodologies: Improving the Practice, North-Holland,Amsterdam, The Netherlands, pp. 19-54.

Fowler, M. (2004) UML Distilled: A Brief Guide To The StandardObject Modelling Language (3rd edition), Addison-Wesley Longman, Boston, Massachusetts.

Gane, C. and Sarson, T. (1979)Structured Systems Analysis: Toolsand Techniques, Prentice-Hall, Englewood Cliffs,California.

Gartner Group (2006) Growing IT's Contribution: The 2006 CIOAgenda (EXP Premier Report, No. January2006),Gartner, Inc, Stamford, Connecticut.

Gehlert, A. and Esswein, W. (2007) Toward a Formal ResearchFramework for Ontological AnalysesAdvancedEngineering Informatics,21, 119-131.

Gemino, A. and Wand, Y. (2003) Evaluating ModelingTechniques based on Models of Learning,Communications of the ACM, 46, 79-84.

Gemino, A. and Wand, Y. (2005) Complexity and Clarity inConceptual Modeling: Comparison of Mandatory andOptional Properties, Data & Knowledge Engineering,55, 301-326.

Green, P. and Rosemann, M. (2000) Integrated Process Modeling.An Ontological Evaluation, Information Systems,25, 73-

87.

7/30/2019 c26433

14/17


Green, P. and Rosemann, M. (2001) Ontological Analysis ofIntegrated Process Models: Testing Hypotheses,Australasian Journal of Information Systems,9, 30-38.

Green, P. and Rosemann, M. (2002) "Perceived OntologicalWeaknesses of Process Modelling Techniques: FurtherEvidence" in S. Wrycza (ed.)10th European Conferenceon Information Systems, Association for InformationSystems, Gdansk, Poland, pp. 312-321.

Green, P. and Rosemann, M. (2004) Applying Ontologies toBusiness and Systems Modeling Techniques andPerspectives: Lessons Learned,J ournal of DatabaseManagement,15, 105-117.

Green, P. and Rosemann, M. (2005) "Ontological Analysis ofBusiness Systems Analysis Techniques: Experiences andProposals for an Enhanced Methodology" in P. Greenand M. Rosemann (eds.) Business Systems Analysis withOntologies, Idea Group, Hershey, Pennsylvania, pp. 1-27.

Green, P., Rosemann, M. and Indulska, M. (2005) Ontological

Evaluation of Enterprise Systems Interoperability UsingebXML, IEEE Transactions on Knowledge and DataEngineering,17, 713-725.

Green, P., Rosemann, M., Indulska, M. and Manning, C. (2007)Candidate Interoperability Standards: An OntologicalOverlap Analysis,Data & Knowledge Engineering, 62,274-291.

Gruber, T.R. (1993) A Translation Approach to Portable OntologySpecifications,Knowledge Acquisition, 5, 199-220.

Guarino, N. and Welty, C. (2002) Evaluating OntologicalDecisions with Ontoclean,Communications of the ACM,45, 61-65.

Guizzardi, G. (2005) Ontological Foundations for StructuralConceptual Models,Telematica Instituut, Enschede, TheNetherlands.

Hevner, A.R., March, S.T., Park, J. and Ram, S. (2004) DesignScience in Information Systems Research,MISQuarterly, 28, 75-105.

Holsapple, C.W. and Joshi, K.D. (2002) A Collaborative Approachto Ontology Design, Communications of the ACM, 45,42-47.

Kautz, K., Wyssusek, B., Lyytinen, K., Milton, S., Kazmierczak,E., Opdahl, A.L., Krogstie, J., Guarino, N., Guizzardi,G., Wand, Y . and Weber, R. (2006) Debate Section: "OnOntological Foundations of Conceptual Modelling",Scandinavian Journal of Information Systems,18, 61-152.

Keen, C.D. and Lakos, C. (1994) "Information Systems Modellingusing LOOPN++, an Object Petri Net Scheme" in H. G.Sol, A. Verbraeck and P. W. G. Bots (eds.) 4thInternational Working Conference on Dynamic

Modelling and Information Systems, Delft UniversityPress, Noordwijkerhout, The Netherlands, pp. 31-52.

Keen, C.D. and Lakos, C. (1996) "Analysis of the DesignConstructs Required in Process Modelling" in M. Purvis(ed.) International Conference on Software Engineering:Education and Practice, IEEE Computer Society,

Dunedin, Ireland, pp. 434-441.

Keen, P.G.W. (1980) "MIS Research: Reference Disciplines and aCumulative Tradition" in E. R. McLean (ed.)1stInternational Conference on Information Systems, ACMPress, Philadelphia, Pennsylvania, pp. 9-18.

Keller, G., Nttgens, M. and Scheer, A.-W. (1992)SemantischeProzessmodellierung auf der Grundlage

"Ereignisgesteuerter Prozessketten (EPK)" (WorkingPaper, No. 89), Institut fr Wirtschaftsinformatik,Universitt Saarbrcken (in German), Saarbrcken,Germany.

Kim, H.M. (2002) Predicting How Ontologies for the SemanticWeb Will Evolve, Communications of the ACM, 45, 48-54.

Krogstie, J., Sindre, G. and Jrgensen, H.D. (2006) ProcessModels Representing Knowledge for Action: a RevisedQuality Framework, European J ournal of InformationSystems, 15, 91-102.

Kuhn, T.S. (1962)The Structure of Scientific Revolutions, ChicagoUniversity Press, Chicago, Illinois.

Lindland, O.I., Sindre, G. and Solvberg, A. (1994) UnderstandingQuality in Conceptual Modeling, IEEE Software, 11, 42-49.

Masolo, C., Borgo, S., Gangemi, A., Guarino, N. and Oltramari, A.(2003) Ontology Library (final) (WonderWebDeliverable, No. Del 18), Laboratory For AppliedOntology, Trento, Italy.

Mayer, R.J ., Menzel, C.P., Painter, M.K., de Witte, P.S., Blinn, T.and Perakath, B. (1995) Information Integration ForConcurrent Engineering (IICE) IDEF3 Process

Description Capture Method Report (Interim TechnicalReport, No. AL-TR-1995-XXXX), Logistics Research

Division, College Station, Texas.Miers, D., Harmon, P. and Hall, C. (2006)The 2006 BPM SuitesReport,BPTrends.com.

Moody, D.L. (2005) Theoretical and Practical Issues in Evaluatingthe Quality of Conceptual Models: Current State andFuture Directions,Data & Knowledge Engineering, 15,243-276.

Nelson, H.J., Poels, G., Genero, M. and Piattini, M. (2005) Qualityin Conceptual Modeling: Five Examples of the State ofthe Art,Data & Knowledge Engineering, 55, 237-242.

OASIS (2001) ebXML Business Process Specification SchemaVersion 1.01, URLhttp://www.ebxml.org/specs/ebBPSS.pdf, accessed

March 12, 2005.Opdahl, A.L. and Henderson-Sellers, B. (2001) Grounding the

OML Metamodel in Ontology,J ournal of Systems andSoftware, 57, 119-143.

Opdahl, A.L. and Henderson-Sellers, B. (2002) OntologicalEvaluation of the UML Using the Bunge-Wand-WeberModel,Software and Systems Modeling, 1, 43-67.

Petri, C.A. (1962) "Fundamentals of a Theory of AsynchronousInformation Flow" in C. M. Popplewell (ed.) IFIPCongress 62: Information Processing, North-Holland,Munich, Germany, pp. 386-390.

Recker, J . (2008) "Understanding Process Modelling Grammar

Continuance: A Study of the Consequences ofRepresentational Capabilities" in Faculty of Information

7/30/2019 c26433

15/17


Technology, Queensland University of Technology,Brisbane, Australia.

Recker, J. and Indulska, M. (2007) An Ontology-Based Evaluationof Process Modeling with Petri Nets,J ournal ofInteroperability in Business Information Systems,2, 45-64.

Recker, J., Indulska, M., Rosemann, M. and Green, P. (2005) "DoProcess Modelling Techniques Get Better? AComparative Ontological Analysis of BPMN" in B.Campbell, J . Underwood and D. Bunker (eds.)16thAustralasian Conference on Information Systems,Australasian Chapter of the Association for InformationSystems, Sydney, Australia.

Recker, J., Indulska, M., Rosemann, M. and Green, P. (2006)"How Good is BPMN Really? Insights from Theory andPractice" in J. Ljungberg and M. Andersson (eds.)14thEuropean Conference on Information Systems,Association for Information Systems, Goeteborg,Sweden, pp. 1582-1593.

Recker, J., Rosemann, M., Indulska, M. and Green, P. (2009)Business Process Modeling: A Comparative Analysis,

J ournal of the Association for Information Systems, 10,333-363.

Rosemann, M. and Green, P. (2000) "Integrating Multi-PerspectiveViews Into Ontological Analysis" in W. J . Orlikowski, P.Weill, S. Ang and H. C. Krcmar (eds.) 21st InternationalConference on Information Systems, Association forInformation Systems, Brisbane, Australia, pp. 618-627.

Rosemann, M. and Green, P. (2002) Developing a Meta Model forthe Bunge-Wand-Weber Ontological Constructs,Information Systems,27, 75-91.

Rosemann, M., Green, P. and Indulska, M. (2004) "A ReferenceMethodology for Conducting Ontological Analyses" inH. Lu, W. Chu, P. Atzeni, S. Zhou and T. W. Ling (eds.)Conceptual Modeling ER 2004, Springer, Shanghai,China, pp. 110-121.

Rosemann, M., Green, P. and Indulska, M. (2005) "A ProceduralModel for Ontological Analyses" in D. Hart and S.Gregor (eds.) Information Systems Foundations:Constructing and Criticising, ANU E Press, Canberra,Australia, pp. 153-163.

Rosemann, M., Recker, J ., Indulska, M. and Green, P. (2006) "AStudy of the Evolution of the RepresentationalCapabilities of Process Modeling Grammars" in E.

Dubois and K. Pohl (eds.) Advanced InformationSystems Engineering - CAiSE 2006, Springer,Luxembourg, Grand-Duchy of Luxembourg, pp. 447-461.

Russell, N., ter Hofstede, A.H.M., Edmond, D. and van der Aalst,W.M.P. (2005a) "Workflow Data Patterns:Identification, Representation and Tool Support" in L.M. L. Delcambre, C. Kop, H. C. Mayr, J. Mylopoulosand . Pastor (eds.) Conceptual Modeling - ER 2005,Springer, Klagenfurt, Austria, pp. 353-368.

Russell, N., van der Aalst, W.M.P., ter Hofstede, A.H.M. andEdmond, D. (2005b) "Workflow Resource Patterns:

Identification, Representation and Tool Support" in .Pastor and J. Falco e Cunha (eds.)Advanced

Information Systems Engineering - CAiSE 2005,Springer, Porto, Portugal, pp. 216-232.

Scheer, A.-W. (2000) ARIS - Business Process Modeling (3rdedition), Springer, Berlin, Germany.

Shanks, G., Tansley, E. and Weber, R. (2003) Using Ontology toValidate Conceptual Models, Communications of theACM, 46, 85-89.

Soffer, P. and Wand, Y . (2005) On the Notion of Soft-Goals inBusiness Process Modeling,Business ProcessManagement Journal, 11, 663-679.

Uschold, M. and Grninger, M. (1996) Ontologies: Principles,Methods and Applications,The Knowledge EngineeringReview,11, 93-136.

van der Aalst, W.M.P. (2003) Dont Go with the Flow: WebServices Composition Standards Exposed, IEEEIntelligent Systems,18, 72-76.

van der Aalst, W.M.P., ter Hofstede, A.H.M., Kiepuszewski, B.and Barros, A.P. (2003) Workflow Patterns, Distributedand Parallel Databases, 14, 5-51.

Wand, Y. and Weber, R. (1988) "An Ontological Analysis ofSome Fundamental Information Systems Concepts" in J.I. De Gross and M. H. Olson (eds.) 9th InternationalConference on Information Systems, Association forInformation Systems, Minneapolis, Minnesota, pp. 213-226.

Wand, Y. and Weber, R. (1989) "An Ontological Evaluation ofSystems Analysis and Design Methods" in E. D.Falkenberg and P. L indgreen (eds.) Information SystemConcepts: An In-depth Analysis. Proceedings of the IFIP

TC 8/WG 8.1 Working Conference on Information

SystemConcepts, North Holland, Amsterdam, The

Netherlands, pp. 79-107.Wand, Y . and Weber, R. (1990a) "Mario Bunge's Ontology as aFormal Foundation for Information Systems Concepts"in P. Weingartner and G. J. W. Dorn (eds.)Studies onMario Bunge's Treatise, Rodopi, Amsterdam, TheNetherlands, pp. 123-149.

Wand, Y. and Weber, R. (1990b) An Ontological Model of anInformation System, IEEE Transactions on SoftwareEngineering,16, 1282-1292.

Wand, Y . and Weber, R. (1993) On the OntologicalExpressiveness of Information Systems Analysis andDesign Grammars,J ournal of Information Systems, 3,217-237.

Wand, Y. and Weber, R. (1995) On the Deep Structure ofInformation Systems, Information Systems Journal, 5,203-223.

Wand, Y . and Weber, R. (2006) On Ontological Foundations ofConceptual Modeling: A Response to Wyssusek,Scandinavian Journal of Information Systems,18, 127-138.

Weber, R. (1997) Ontological Foundations of InformationSystems, Coopers & Lybrand and the AccountingAssociation of Australia and New Zealand, Melbourne,Australia.

Weber, R. and Zhang, Y . (1996) An Analytical Evaluation of

NIAM's Grammar for Conceptual Schema Diagrams,Information Systems Journal, 6, 147-170.

7/30/2019 c26433

16/17


Wohed, P., van der Aalst, W.M.P., Dumas, M. and ter Hofstede,A.H.M. (2003) "Analysis of Web Services CompositionLanguages: The Case of BPEL4WS" in I.-Y . Song, S.W. Liddle, T. W. Ling and P. Scheuermann (eds.)Conceptual Modeling - ER 2003, Springer, Chicago,Illinois, pp. 200-215.

Wohed, P., van der Aalst, W.M.P., Dumas, M., ter Hofstede,A.H.M. and Russell, N. (2005) "Pattern-Based Analysisof the Control-Flow Perspective of UML ActivityDiagrams" in L. M. L. Delcambre, C. Kop, H. C. Mayr,

J . Mylopoulos and . Pastor (eds.)Conceptual Modeling- ER 2005, Springer, Klagenfurt, Austria, pp. 63-78.

Wohed, P., van der Aalst, W.M.P., Dumas, M., ter Hofstede,A.H.M. and Russell, N. (2006) "On the Suitability ofBPMN for Business Process Modelling" in S. Dustdar, J.L. Fiadeiro and A. Sheth (eds.)Business ProcessManagement - BPM 2006, Springer, Vienna, Austria, pp.161-176.

Key

Set of constructs described in the BWW modelBWW

BWW

MT

Set of constructs comprising theModellingTechniqueMT

1:0

Construct described in the BWW model

Modelling technique construct

1:m

m:1

0:1

Figure 1 Types of potential representational shortcomings

Do you need this concept?

yes

no yes

Can you directly model thisconcept?

no

Do you perceive it as aproblem?

yes

Is it a critical

problem?

yes no

IIIIIIIVV

no

Figure 2 Questionnaire structure and response classification

7/30/2019 c26433

17/17


Modelling techniquedevelopment

Theory development and adoption

Theoreticalrepresentational

issues

Person

Purpose

Emp. validatedrepresentational

issues

Analytical evaluation Empirical testing

Revisedtechnique

Modellingtechniques

BWW

representationmodel

Bunges

ontology

Model quality

Modelevaluation

Modelling techniquere-engineering

Figure 3 The main phases of representational analysis of modelling techniques

0

5

10

15

20

25

Petrinet(1

962)

ANSI

Flow

charts(1

970)

DFD(1

979)

ISO/TC

97(1982)

MERISE(1

991)

EPC(1

992)

IDEF3(1

995)

ebXM

L1.01(

2001)

BPML

1.0(2002)

WSC

I1.0(2

002)

BPEL4W

S1.1(2

003)

BPMN

1.0(2004)

TimeNumberofcorresponding

BWWc

onstructs

Figure 4 Comparison of representation mapping analyses