In this position paper, we are concerned with the role of modeling in an enterprise context. In general, the creation, management, and use of models comes at a cost. We content that, especially in an enterprise context, it becomes increasingly important to make explicit trade-ofs between the costs related to modeling and some return in relation to the goals of the enterprise. To better reason about such tradeofs, we propose three concepts: Return on Modeling Efort (RoME), a model's V alue in Action (ViA), and the need to manage the Retention of Modeling Efort (RiME). In doing so, we also suggest some of the avenues we intend to follow in further researching these concepts. By means of this position paper, we also hope to engage more colleagues in this fundamental line of research.
In the context of enterprises, a wide range of models are produced and used. This includes,
among others, enterprise (architecture) models, business process models, ontology models,
enterprise architecture models, information models, value models, business ontologies, as well
as diferent kinds of reference models. We consider each of these kinds of models as being
valued members of the larger family of domain models [
From a general perspective, we would argue that modeling is a natural thing for humans
to do. Whenever there is a need to explicitly reason about, study, or discuss, some part of an
existing/imagined domain, we do so in terms of an abstraction of this domain of interest. When
representing such an abstractions in terms of some artifact, the resulting artifacts are (used
as) domain models as they ‘stand model for’ the domain of interest (in relation to the need at
hand) [
Whatever the domain of interest is, and irrespective of whether it is part of the digital, social,
or physical world, domain models have a potential benefit towards the understanding, assessing,
(re)designing, etc, of the domain of interest. Depending on the needs at hand, models may need
to meet diferent quality criteria [
In an enterprise context, domain models (potentially) have an important role to play. More
specifically, in Software Engineering, Information Systems Engineering, Business Process
Engineering, and Enterprise Engineering in general, a wide range of domain models are produced
and used to meet many diferent purposes. As suggested in [
In general, modeling related eforts, including the creation, administration, and use, of models, require investments in terms of time, money, cognitive efort, etc. We contend that such investments should be met by a (potential) return, especially in an enterprise context where domain modeling eforts are more than ever governed by the laws of economics. Therefore, in our view, a more rigorous underpinning of such cost/benefit trade-ofs is called for.
In our observation, some but not much work has been conducted on balancing the expected
return of a modeling efort in relation to the involved efort. Some authors, indeed, identify
the need to more explicitly identify the purpose for modeling [
In this position paper, we discuss three concepts that are intended to enable us to start reasoning more explicitly about such trade-ofs: the earlier mentioned Return on Modeling Efort (RoME), a model’s V alue in Action (ViA), and the need to manage the Retention of Modeling Efort (RiME). In discussing these concepts, we also identify some of the research questions in further exploring and elaborating these three concepts.
This paper is actually part of a broader joint research efort of the two authors, where we
aim to explore and deepen the foundations of domain modeling (in general), including the
philosophical, ontological, and pragmatic aspects [
In the remainder of this paper, we will start in section 2 with a summary of our current understanding of what domain models are. This sets the scene for the introduction of the concepts of RoME, ViA and RiME in sections 3, 4, and 5 respectively, while also identifying some of the research challenges we see related to the further elaboration of these concepts.
Based on foundational work by e.g. Apostel [23], and Stachowiak [24], more recent work on
the same by diferent authors [
A social artifact that is acknowledged by a collective agent to represent an abstraction of some domain for a particular cognitive purpose.
With domain, we refer to ‘anything’ that one can speak and/or reflect about; i.e. the domain of interest. As such, domain simply refers to ‘that what is being modeled’. In an enterprise and information systems engineering context this includes (but is not limited to) business processes, information structures, business transactions, value exchanges, etc. Furthermore, the domain could be something that already exists in the ‘real world’, something that is desired to exist in the future, or something imagined.
A model is seen as a social artifact in the sense that its role as a model should be recognizable by a collective agent (e.g. people1). In the context of enterprise and information systems engineering, such an artifact typically takes the form of some ‘boxes-and-lines’ diagram. More generally, however, domain models can, depending on the purpose at hand, take other forms as well, including text, mathematical specifications, games, animations, simulations, and physical objects.
The collective agent observes the domain by way of their senses and/or by way of (collective) self-reflection, and, based on this, should acknowledge/accept the artifact as indeed being a model of the domain (for a given purpose).
A model must always be created for some cognitive purpose2; i.e. to express, specify, learn about, or experience, knowledge regarding the modeled domain. As a direct corollary to this, one can conclude that a model, being a social artifact must therefore be a language utterance, as such implying it to be a social-linguistic artifact.
Finally, a model is the representation of an abstraction. This implies that, in line with the cognitive purpose of the model, some (if not most) ‘details’ of the domain are consciously filtered out.
In the context of enterprise and information systems engineering, a specific class of domain models has grown to play an important role, namely, conceptual models. In the traditional information systems engineering view [31], a conceptual model captures the essential structures of some universe of discourse. In this context, conceptual models are used to express the 1The pre-noun collective does suggest that it it would require to involve multiple people. We do, indeed, acknowledge the use of domain models by an individual person as well, but prefer to treat this as a special case concerning a ‘self-shared’ model. 2In earlier work, we did not include the explicit focus on cognitive purpose, but rather spoke about some purpose in general. In retrospect, we think this was an omission. Adding cognitive clarifies the role of models as a way to express, specify, learn about, or experience, knowledge regarding the modeled domain. We would like to thank Jan Schoonderbeek for making us aware of this omission. concepts, and their (allowed) relations, of the universe of discourse (while avoiding the inclusion of implementation/storage details).
In our current understanding (based on [
1. the purpose of the model is dominated by the ambition to remain as-trueas-possible to the conceptualization of the domain by the collective agent, while 2. there is an explicit mapping from the elements in the model to the latter domain conceptualization.
The domain conceptualization identifies the fundamental concepts in terms of which the collective agent create(s) their conception of the world. This mapping specifies the real-world semantics of that model and characterizes its ontological commitment [30].
A conceptual model, therefore, provides an explicit – human understandable – representation of a theory about the entities and their ties that are assumed to exist in a given domain of interest (the ontological commitment), as such explicitly capturing descriptive and/or prescriptive selected aspects of the modeled domain. As a result, conceptual models enable us to explicitly clarify the things we talk and reason about (at a chosen level of abstraction and from a desired perspective).
In an information systems engineering context, the ambition for a conceptual model to remain as-true-as-possible to the conceptualization of the domain by the collective agent, has a direct correspondence to the conceptualization principle as put forward in the well known ISO report on the design of information systems [31].
The field of information systems engineering, indeed, provides a fruitful application area
for conceptual modeling. At the same time, however, we suggest to avoid ‘framing’ our
understanding of what a conceptual model is to this application area only. The ambition to
as-true-as-possible to the conceptualization of the domain by the collective agent is not only
applicable in the context of information systems engineering. For instance, Guarino et al. [
At a more generalized level, we also observe that in many diferent endeavors in which we (as humans) aim to understand the workings of some domain and/or aim to express/study design alternatives, we actually do so in terms of (purpose and situation specific) domain models. This includes many examples across science and engineering at large. We also argue that in these cases, a deepening of our understanding of the essential mechanisms leads to a natural drive to create domain models that remain as-true-as-possible to the original domain (and our conceptualization thereof), i.e. conceptual models.
Identifying conceptual models as a specific class of domain models, does raise the question
regarding the role of ‘other’ domain models that are ‘not conceptual’. In [
An interesting analogy, which certainly needs further investigation, is the notion of surrogate modeling in the context of simulation [34] of real-world systems3. The level at which a simulation model reflects all (relevant) properties of a (planned/existing) real-world system is referred to as the fidelity of the simulation model: “Fidelity in the modeling context refers to the degree of the realism of a simulation model” [34]. Likewise, one can speak of being as-true-as-possible-to a given domain as a sort of conceptual fidelity . Conceptual fidelity, also frequently called domain appropriateness, represents the level of homomorphism between a given representation and the underlying domain conceptualization it commits to. In the ideal case, this representation artifact is not only isomorphic to the structure of that conceptualization (i.e., it represents in a univocal and non-redundant way all its constituting concepts and only them) but it also only allows for interpretations that represent state of afairs deemed acceptable by that conceptualization [ 35]. As, in the case of simulation models of real-world systems, the involved high fidelity models may be too computationally intensive to simulate as a whole, one uses so-called surrogate models [34] that are computationally more eficient, while approximating the high fidelity model good enough to meet the (optimization) purpose at hand.
It is important to note that we do not argue that non-conceptual models would be bad; far from it. However, it needs to be clear what the ‘conceptual deviation’ are of a non-conceptual model in relation to the conceptual model of the same domain, and what the benefit are of these deviations in terms of e.g. computational eficiency. As such, it might quite well be the case that one conceptual model has diferent associated non-conceptual models catering for diferent needs [36].
As mentioned in the introduction, we take the view that a more rigorous underpinning is needed of the costs and benefits involved in domain modeling, as well as the associated trade-ofs; i.e. the Return on Modeling Efort (RoME). A more explicit thinking in terms of modeling efort, and the potential return on these eforts, is likely to already help to guide (scope, time-box, select an appropriate level of formality, etc) modeling eforts in an enterprise context.
Returning briefly to the distinction between conceptual models and non-conceptual models, we postulate that the RoME of a conceptual model is at least as high as the sum of the RoME of each of the non-conceptual models that have been derived from it.
The ambition to further elaborate, and underpin, the concept of RoME results in the following main research questions, which will guide us in our future research.
• What are the factors that define modeling efort; i.e. the efort needed to create, manage, and use models. How to measure these? 3We would like to thank Carlos Kavka, from the Luxembourg Institute of Science and Technology, for pointing us in the direction of the notion of surrogate models.
• What are ways to potential limit/reduce modeling efort, while retaining the same (potential) returns? Information technologies can potentially help in reducing the modeling efort. For instance, in terms of hybrid mixes of human and artificial intelligence [ 37, 38]) to help derive/infer domain models from diferent sources of data pertaining to the domain of interest. Process mining [39] can be regarded as specific illustration of this point. • What are the factors that define the value of models. How to measure these?
The discussion in section 4 regarding ViA already tries to take first steps in understanding the value of models. • What are ways to increase the (potential) value of models; i.e. increase their return.
Model-driven software engineering (including low-code), as well as the use of models as operational/executable artifacts in, e.g., rule engines, process engines, and even gaming engines, are illustrations of ways in which information technology can be used to increase the value of models. An example of the use of operational models to increase the value of legacy data without the need to migrate databases to diferent formats and platform is the ODBC (Ontology-Based Data Access) (also called Virtual Knowledge Graph) strategy [40, 41]. • How to make trade-ofs between modeling efort and its (potential) return in relation to specific contexts and purposes?
To answer these questions, a combination of theory-driven and practice-driven work is needed. The desire to gather empirical data from real-world situations, in order to help answer the above research questions, was one of the drivers for the Models-at-Work initiative4. The idea of this initiative is to gather a library of cases in which domain models have played an important role, while also documenting their RoME. The aim is to, for each of these cases, gather insights into questions such as: • Purpose & requirements – What was the intended purpose (and audience) of the model and/or its creation? What were specific requirements on the model? • Context & challenge – What was the social and/or technical context in which the model was created and/or used. What was specifically challenging? What were uncertainties? Were there any social and/or technical complexities? • Activities & efort – What were the activities involved in creating the model? How much efort (time, budget, people/roles involved, etc) was needed to create the model? What tools and methods were used? How was the validity (in relation to the goal & requirements) of the model managed and assessed? • Resulting model – What kind of model was produced? Did the model have to include ‘quality compromises’ for strategic/political reasons? Was the developed model a refinement of a standard or published/known model? Was a specific modeling language and/or tool used? • Return on modeling efort – In line with the intended purpose, what was the expected return on modeling efort? What was the materialized return on modeling efort? Which stakeholder(s) made the investment in modeling, and which stakeholder(s) reaped the benefits?
When considering the return on modeling efort, it is necessary to consider both models and modeling, from a value oriented perspective. In doing so, we take the value proposition ontology as defined in [ 21] as a base. In particular the notions of value bearer (that what potentially has value), value ascription (the act of assigning value to the value bearer), value beholder (a role of the actor who ascribes value) and value beneficiary (a role of the actor who reaps the benefits of the value bearer). In actual trade-ofs between the expected return on the investments in the creation and use of a model, it will be necessary to distinguish between the ex-ante expected value and the ex-post realized value.
In the case of models and modeling, there seem to be three potential value bearers:
Such a process may e.g. result in the added value that those who are involved in the modeling process5 develop a deeper and/or more consistent (joint) understanding of the modeled domain, and also have the chance of building shared terminology based on that understanding.
In this case, the value beneficiary (and value beholder) can pertain to those actor(s) who are directly involved in the (co-)creation process, and/or those who stand to benefit from an increased (joint) understanding of the latter actor(s). 2. Value in use – The operational usage of the model (in line with its purpose6).
This may, e.g., involve the use of the model to support decision making, give prescriptive/ descriptive guidance towards development processes and/or operational processes, etc. In this case, the value beneficiary is the user of the model, while the value beholder may indeed be the same user of the model, but may also be the actor who has a more overall role/interest (such as the transfer of design knowledge from requirements engineering, via design, to implementation). 3. Value in transaction – The (ownership of the) model itself.
This pertains to e.g. reference models, etc, capturing generalized knowledge that can potentially be re-applied in diferent situations.
In this case, the direct value beneficiaries are the actor who uses the model to support them in their own activities, while the indirect value beneficiaries are the actors who have the original owners of the model. The latter actors are indirect value beneficiaries who receive the value of the model via a transaction (e.g. transfer of ownership of the model, or the right to apply the model). In line with this, the direct value beholders are the actors ascribe value to the usage/application of the model in a concrete situation, while the indirect value beholders ascribe value to the model by way of its potential value in (future) transactions. More generally, in [14], we have discussed the potential of ‘models acts’ as complex language acts (in the sense of speech act theory [42]). ‘Creative speech acts’ (speech acts with a double direction of fit) can bring about the existence of things in the world when uttered in a certain context. Some models can then describe the propositional 5Which could be a group of actors, but can also be a single actor expressing their thoughts about an existing/future domain. 6One could indeed also gain value from a model by (ab)using it beyond its intended purpose. content of these creative model acts, which bring value by bestowing the model owners with rights associated with the propositional content that said model describes. At a more fundamental level, we would argue that ultimately value in creation and value in use are the root/direct value bearers of models. The value in transaction is derived from the potential of a model’s future value in use. The combination of value in creation and value in use is what we refer to as the Value in Action (ViA) of models.
In the work we reported in [14], we provided a goal structure in terms of taxonomy of modeling related goals. This taxonomy distinguishes between models with a prescriptive purpose (intervening, planning, coordinating), a creative purpose (bringing about changes in reality), and a descriptive purpose (understanding, problem-solving, communicating, and documenting), each time involving models that receive their value in action.
In future research, we aim to further operationalize the diferent dimensions for model value. To some extent, this can be based on expert interviews, and theoretical analysis. However, empirical data based on case reports in line with the aforementioned Models-at-Work initiative.
Based on the premise that models provide value in action, models as artifacts embody a potential value in the future; i.e. when they are taken ‘into action’. This potential value may need to be ‘safeguarded’ in terms of e.g. the digital/physical integrity of the model (as an artifact), as well as its security in the sense of undesired (reading/writing) access by third parties.
Next to that, models may be subject to a ‘shelf life’ in the sense that they may lose their potential value. For instance, if the gap between its creation and use (model in action) becomes too large, the purpose for which the model was intended may have lost its relevance. Furthermore, when a model pertains to a domain that may change of time (such as business processes, application portfolios, IT infrastructures, etc.) the model may lose actuality with regard to the state of afairs of the modeled domain 7.
As such, we argue that the retention of the potential value of a model needs explicit management. The activities involved in the retention of this potential value are likely to also add extra efort to the modeling efort as a whole, and thus need to be balanced against the (remaining) potential value. Further elaboration of the factors involved in the retention of modeling efort can be based partially on expert interviews, and theoretical analysis, but also needs more in depth analysis of real world cases.
In this position paper, we zoomed in on the role of domain modeling in an enterprise context from a value perspective. When modeling in the context of enterprise, it becomes more important to 7The need to create modeling management processes and infrastructures that secure that models remain actual w.r.t. to their referent appears in the literature of enterprise modeling since at least the 90’s as Living Enterprise Models[43], somewhat also reflect in conceptual modeling in the notion of Active Models and, more recently, also in the notion of Digital Twins[44] relate the costs involved in modeling activities to some return in relation to the goals of the enterprise. This requires explicit trade-ofs between the costs and benefits of modeling.
To better reason about such trade-ofs, we proposed the connected concepts of RoME, ViA and RiME. We think that these concepts would already be beneficial in creating awareness for the needed trade-ofs in practice. At the same time, as also discussed, we realize that much more research is needed towards the further elaboration of these concepts. As such, the discussion in this paper also provide a starting point for (our own) further research, where we hope to engage colleagues to join us on this journey; Let’s go to RoME ViA RiME. framework for value co-creation, in: J. Gordijn, H. A. Proper (Eds.), Proceedings of the 12th International Workshop on Value Modeling and Business Ontologies, VMBO 2018, Amsterdam, The Netherlands, February 26th – 27th, 2018, volume 2239 of CEUR Workshop Proceedings, CEUR-WS.org, 2018, pp. 122–132. URL: http://ceur-ws.org/Vol-2239/article_ 13.pdf. [23] L. Apostel, Towards the Formal Study of Models in the Non-Formal Sciences, Synthese 12 (1960) 125–161. [24] H. Stachowiak, Allgemeine Modelltheorie, Springer, Heidelberg, Germany, 1973. doi:10.
1007/978-3-7091-8327-4. [25] D. Harel, B. Rumpe, Meaningful Modeling: What’s the Semantics of “Semantics”?, IEEE
Computer 37 (2004) 64–72. doi:10.1109/MC.2004.172. [26] B. Thalheim, The Theory of Conceptual Models, the Theory of Conceptual Modelling and Foundations of Conceptual Modelling, in: Handbook of Conceptual Modeling, Springer, Heidelberg, Germany, 2011, pp. 543–577. [27] K. Sandkuhl, H.-G. Fill, S. J. B. A. Hoppenbrouwers, J. Krogstie, F. Matthes, A. L. Opdahl, G. Schwabe, Ö. Uludag, R. Winter, From Expert Discipline to Common Practice: A Vision and Research Agenda for Extending the Reach of Enterprise Modeling, Business & Information Systems Engineering 60 (2018) 69—80. doi:10.1007/s12599-017-0516-y. [28] S. J. B. A. Hoppenbrouwers, H. A. Proper, T. P. v. d. Weide, A fundamental view on the process of conceptual modeling, in: L. Delcambre, C. Kop, H. C. Mayr, J. Mylopoulos, O. Pastor (Eds.), Conceptual Modeling – ER 2005, 24th International Conference on Conceptual Modeling, Klagenfurt, Austria, October 24-28, 2005, Proceedings, volume 3716 of Lecture Notes in Computer Science, Springer, Heidelberg, Germany, 2005, pp. 128–143. doi:10.1007/11568322_9. [29] H. A. Proper, A. A. Verrijn–Stuart, S. J. B. A. Hoppenbrouwers, On utility-based selection of architecture-modelling concepts, in: S. Hartmann, M. Stumptner (Eds.), Conceptual Modelling 2005, Second Asia-Pacific Conference on Conceptual Modelling (APCCM2005), Newcastle, NSW, Australia, January/February 2005, volume 43 of Conferences in Research and Practice in Information Technology Series, Australian Computer Society, Sydney, New South Wales, Australia, 2005, pp. 25–34. URL: http://crpit.scem.westernsydney.edu.au/ abstracts/CRPITV43Proper.html. [30] G. Guizzardi, On Ontology, ontologies, Conceptualizations, Modeling Languages, and (Meta)Models, in: O. Vasilecas, J. Eder, A. Caplinskas (Eds.), Databases and Information Systems IV – Selected Papers from the Seventh International Baltic Conference, DB&IS 2006, July 3-6, 2006, Vilnius, Lithuania, volume 155 of Frontiers in Artificial Intelligence and Applications, IOS Press, 2006, pp. 18–39. [31] ISO/IEC JTC 1/SC 32 Technical Committee on Data management and interchange, Information processing systems – Concepts and Terminology for the Conceptual Schema and the Information Base, Technical Report ISO/TR 9007:1987, ISO, 1987. [32] H. A. Proper, On Model-Based Coordination of Change in Organizations, Springer, Heidelberg, Germany, 2021, pp. 79–98. doi:10.1007/978-3-030-84655-8_6. [33] M. R. Quillian, Semantic memory, Semantic Information Processing, Ph.D. thesis, MIT,
Massachusetts, 1968. [34] S. Razavi, B. A. Tolson, D. H. Burn, Review of surrogate modeling in water resources,
Water Resources Research 48 (2012). doi:10.1029/2011WR011527. [35] G. Guizzardi, L. Ferreira Pires, M. v. Sinderen, An ontology-based approach for evaluating the domain appropriateness and comprehensibility appropriateness of modeling languages, in: International Conference on Model Driven Engineering Languages and Systems, Springer, 2005, pp. 691–705. [36] G. Guizzardi, Theoretical foundations and engineering tools for building ontologies as reference conceptual models, Semantic Web 1 (2010) 3–10. [37] M. Snoeck, J. Stirna, H. Weigand, H. A. Proper, Panel discussion: Artificial intelligence meets enterprise modelling (summary of panel discussion), volume 2586 of CEUR Workshop Proceedings, CEUR-WS.org, 2019, pp. 88–97. URL: http://ceur-ws.org/Vol-2586/paper8.pdf. [38] C. Feltus, Q. Ma, H. A. Proper, P. Kelsen, Towards AI Assisted Domain Modeling, in: I. Reinhartz-Berger, S. W. Sadiq (Eds.), Advances in Conceptual Modeling ER 2021 Workshops CoMoNoS, EmpER, CMLS, St. John’s, NL, Canada, October 18-21, 2021, Proceedings, volume 13012 of Lecture Notes in Computer Science, Springer, Heidelberg, Germany, 2021. [39] W. M. P. v. d. Aalst, et. al, Process Mining Manifest, in: Business Process Management Workshops, volume 99 of Lecture Notes in Business Information Processing, 2012, pp. 169– 194. [40] G. Xiao, D. Calvanese, R. Kontchakov, D. Lembo, A. Poggi, R. Rosati, M. Zakharyaschev, Ontology-based data access: A survey, International Joint Conferences on Artificial Intelligence, 2018. [41] G. Xiao, D. Lanti, R. Kontchakov, S. Komla-Ebri, E. Güzel-Kalaycı, L. Ding, J. Corman, B. Cogrel, D. Calvanese, E. Botoeva, The virtual knowledge graph system ontop, in: International Semantic Web Conference, Springer, 2020, pp. 259–277. [42] J. R. Searle, J. R. Searle, Speech acts: An essay in the philosophy of language, volume 626,
Cambridge university press, 1969. [43] L. Whitman, B. L. Huf, A living enterprise model, in: Proceedings of the 6th Industrial
Engineering Research Conference, Miami Beach, FL, Citeseer, 1997. [44] E. Schultes, M. Roos, L. O. B. da Silva Santos, G. Guizzardi, J. Bouwman, T. Hankemeier, A. Baak, B. Mons, Fair digital twins for data-intensive research, Frontiers in Big Data 5 (2022).