ArticlePDF Available


The digitalization of human society continues at a relentless rate. However, to develop modern information technologies, the increasing complexity of the real-world must be modeled, suggesting the general need to reconsider how to carry out conceptual modeling. This research proposes that the often-overlooked notion of "system" should be a separate, and core, conceptual modeling construct and argues for incorporating it and related concepts, such as emergence, into existing approaches to conceptual modeling. The work conducts a synthesis of the ontology of systems and general systems theory. These modeling foundations are then used to propose a CESM+ template for conducing systems-grounded conceptual modeling. Several new conceptual modeling notations are introduced. The systemist modeling is then applied to a case study on the development of a citizen science platform. The case demonstrates the potential contributions of the systemist approach and identifies specific implications of explicit modeling with systems for theory and practice. The paper provides recommendations for how to incorporate systems into existing projects and suggests fruitful opportunities for future conceptual modeling research.
Data & Knowledge Engineering 141 (2022) 102062
Contents lists available at ScienceDirect
Data & Knowledge Engineering
journal homepage:
System: A core conceptual modeling construct for capturing
Roman Lukyanenko a,,Veda C. Storey b,Oscar Pastor c
aUniversity of Virginia, United States of America
bGeorgia State University, United States of America
cUniversitat Politècnica de València, Spain
Conceptual modeling
Emergent properties
Bunge Systemist Ontology (BSO)
Retrospective case study
Citizen science
The digitalization of human society continues at a relentless rate. However, to develop modern
information technologies, the increasing complexity of the real-world must be modeled, suggest-
ing the general need to reconsider how to carry out conceptual modeling. This research proposes
that the often-overlooked notion of ‘‘system’’ should be a separate, and core, conceptual
modeling construct and argues for incorporating it and related concepts, such as emergence,
into existing approaches to conceptual modeling. The work conducts a synthesis of the ontology
of systems and general systems theory. These modeling foundations are then used to propose a
CESM+ template for conducing systems-grounded conceptual modeling. Several new conceptual
modeling notations are introduced. The systemist modeling is then applied to a case study on the
development of a citizen science platform. The case demonstrates the potential contributions of
the systemist approach and identifies specific implications of explicit modeling with systems for
theory and practice. The paper provides recommendations for how to incorporate systems into
existing projects and suggests fruitful opportunities for future conceptual modeling research.
1. Introduction
With continued human development, social, economic, political and technological systems are growing more complex [14].
Complexity in systems refers to the number of component-parts along with the way in which these parts are structured and interact
with one another and with other systems [5,6]. Systems are the complex entities which constitute the world, such as atoms, animals,
airplanes, universities, stock markets, and galaxies. Generally, the more complex the system, the more difficult it is to fully predict
its behavior. To create and effectively manage complex systems, improved methods, machinery and knowledge are necessary.
This ‘‘complexity challenge’’ opens new opportunities for information technology (IT) development to support, create and manage
complex systems and their users.
Conceptual modeling is a phase of information technology (IT) development. It traditionally focuses on capturing user require-
ments, facts and beliefs about an application domain [710]. Since the 1970s, database design, especially in large organizations,
relied on conceptual models the products of conceptual modeling to model the data to be stored in relational databases
[1113]. Another important application of conceptual models is to support business process management and engineering [1416].
More broadly, conceptual models are used to improve domain understanding, to facilitate communication among IT developers and
stakeholders, and to help visualize and solve IT design challenges [10,1720]. Our growing reliance on information technologies
and their increased sophistication necessitates an ever greater ability of conceptual modeling to represent both physical (including
mental and social) and digital systems [21].
Corresponding author.
E-mail address: (R. Lukyanenko).
Received 15 August 2021; Received in revised form 2 August 2022; Accepted 3 August 2022
Available online 23 August 2022
0169-023X/©2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
To appreciate the challenge in creating and managing complex systems, consider one of the unrealized solutions for tackling
the COVID-19 pandemic, namely, the development of a social (physical) distancing app. Such an app would sense an approaching
person and vibrate, thereby alerting the user of the need to keep distance. The anonymized and aggregated data from such app could
be used by governments to support data-driven policies and facilitate smarter pandemic response.1If widely practiced, physical
distancing leads to significant reductions in respiratory disease transmissions [2224]. However, an effective physical distancing
technology is incredibly challenging, not only because of its many technological obstacles, but importantly, because of a host of
social, ethical, legal, medical, and psychological challenges [25]. Precise and accurate conceptual modeling of facts and opinions in
this domain could assist in the development of effective solutions. Since distancing technology involves personally sensitive usage
by millions of people in real time, and, assuming such usage is not mandated, but is voluntary, we need to accurately model values,
intentions, motivations and needs of different people in order to align the technology with these needs. The use of such technology
is fundamentally collective, involving coordinated efforts on the part of citizens, governments, and medical establishments [26].
To create and sustain an app at such scale amounts to the development of a highly choregraphed complex socio-technical system.
Worse still, such a system might behave in potentially unpredictable and even, possibly, dangerous ways. Inadvertently, such app
could cause undesirable changes in patterns of human movements and socializing or trigger an expansion of mass surveillance.
Measures need to be put in place (including at the level of technical design) to proactively detect and curb any negative outcomes,
while promoting the positive ones. Conceptual modeling then becomes a valuable tool to help engineer effective IT solutions to the
expanding challenges of humanity.
The objective of this research is to examine existing conceptual modeling capabilities with respect to the challenges of the modern
world and suggest a path for better handling of its complexities. We rethink conceptual modeling theory and practice by investigating
a thus far overlooked conceptual modeling concept, namely that of ‘‘system’’. Specifically, we propose that the construct of ‘‘system’’
should be regarded as a basic conceptual modeling construct, on par with constructs such as ‘‘entity’’, ‘‘attribute’’, ‘‘role’’, ‘‘event’’
or ‘‘relationship’’.
Amending conceptual modeling languages with the construct, ‘‘system’’, follows a long line of research that introduced additional
constructs to increase the expressive power of modeling languages. From early research on conceptual modeling, until present,
researchers have been proposing new constructs (e.g., [12,2732]). Some of these became ingrained in widely used conceptual
modeling languages, such as the entity-relationship diagram (ER), Unified Modeling Language (UML), Business Process Model
and Notation (BPMN), and Object-role modeling (ORM), which are now staple elements in practice. The sub/super classes
(i.e., generalization/specialization relationships) is one such example [33,34]. Similarly, we argue that the construct of ‘‘system’’
has the potential to become another basic and indispensable construct in the world of ever-increasing complexity.
Of course, it is already possible to model system components (e.g., parts of a whole) using conventional approaches, such as ER
diagrams or UML. However, as we demonstrate in the paper, traditional conceptual modeling approaches struggle to model many
aspects of systems such as emergence. Furthermore, even though the notion of system is ubiquitous in the conceptual modeling
discourse, there is little guidance for modelers on how to appropriately model systems. This problem is exacerbated by the lack of
consensus and clarity on what constitutes a system and its related constructs.
To keep up with the relentless pace of digitalization of business and society, it is important to continue refining conceptual
modeling to make it more expressive for cases when more explicit and comprehensive modeling of systems is beneficial. Since these
scenarios are pervasive, modeling systems more explicitly is becoming pressing.
In this research, we propose a set of basic notions that are related to the system construct, position system as a core conceptual
modeling primitive, explain the limitations of existing modeling languages, and outline research initiatives that could further
incorporate the system construct into conceptual modeling. Based on theoretical foundations, we propose a CESM+template for
conducing systems-grounded conceptual modeling. Several new conceptual modeling notations are introduced for practitioners and
as input into future academic research. The systemist modeling is analyzed in a case study of the development of a citizen science
platform. We then provide methodological guidelines for designers and a future conceptual modeling research agenda.
2. Background
2.1. Conceptual modeling constructs
Conceptual modeling research and practice is now over 50 years old, with popular conceptual modeling languages, such as the
entity-relationship model [27] appearing in the 1970s. During this lengthy period, numerous constructs have been proposed as
hundreds of different conceptual modeling languages and approaches were introduced, evaluated and applied [3540].
For example, a core modeling construct, which emerged as early as the first conceptual modeling languages, is that of an entity
type [27]. Entity types or classes (used in entity-relationship diagrams and UML Class Diagrams, respectively), are commonly used
to represent groups of objects of interest in the domain of the information systems being developed [13,41,42]. Debates related
to these constructs centered on how to appropriately select [41,43] and apply them [4446], such as identifying the relationship
between classes or entity types and the objects they represent [4751]. Debates also focused on the nature of instances themselves;
for example, whether classes can be instances of other classes [5254].
Other focal constructs in conceptual modeling dealing with entities include ‘‘attributes’’ (characteristics, dimensions, or features
of entities), ‘‘relationships’’ (associations among entities) (e.g., [27]), and ‘‘roles’’ (behaviors and functions of entitles) [28,38].
1Attempts to develop such technology have been made, but the resulting apps not been widely embraced by society. See,
from-around- ieee/the-institute/ieee- products-services/social- distancing-heres- an-app- for-that.
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
These are common in conceptual models representing the form and structure of domains [18,55]. Modeling approaches representing
processes and dynamics of domains include such constructs as ‘‘events’’, ‘‘activities’’, or ‘‘gateways’’ [15,5658]. Those dealing with
goals, values, intentions, have also became popular, having such constructs as ‘‘goal’’, or ‘‘actor’’ [32,5961].
There have been many debates about the value and limitations of various constructs, as well as proposals for how to use them
effectively in conceptual modeling diagrams [6265]. An overlooked, but extremely important construct is that of ‘‘system’’ and its
associated constructs, including emergent properties, mechanism, environment, among others.
2.2. Conceptual modeling foundations and the system construct
The absence of an explicit ‘‘system’’ construct in mainstream conceptual modeling languages (e.g., UML, BPMN, ORM, ER, i*) is
surprising given the ubiquity of the system concept in discourse related to conceptual modeling.
First, systems notion is synonymous with the product of IT the software or computer applications are widely recognized
to be information systems. This is well understood in conceptual modeling. As Mayr and Thalheim [8, p. 2] remind us: ‘‘from the
very beginning, conceptual modeling was propagated as a means to improve the design and implementation of whatsoever software
system, especially with regard to a comprehensive and as clear as possible elicitation and analysis of system requirements’’. (p. 2;
emphasis added).
Second, when IT get implemented in real-world settings, they become part of socio-technical systems [6668]. Socio-technical
systems are composed of technical systems (processes, tasks, and technological infrastructure) and social systems (humans, their rela-
tionships and social structures). The two systems, when put together and begin to interact, produce joint outputs (e.g., information,
furniture, raw materials, services) [6971]. For example, implemented into organizational settings enterprise resource planning,
customer service, electronic payments, e-commerce IT become parts of socio-technical systems created by the fusion of humans and
technology. Hence, enterprise diagrams, such as a UML class diagram, BPMN model or ArchiMate diagram [72], commonly model
socio-technical systems [2,3,73,74]. Often enterprise models comprise of layers (e.g., [2,73]), which can be understood as systemic
levels (discussed later).
Third, the domains that are managed by IT are commonly understood as systems. For example, a conceptual model may represent
facts about an inventory control system to facilitate a more efficient inventory management by an ERP developed with the help of
this conceptual model [75]. Similarly, in an i* framework, modelers can represent social systems, which contain potential users of
technology and their goals and intentions [76].
Technologies, including IT, are also viewed as components of work systems; that is, systems in which human participants and/or
machines perform work using information, technology, and other resources [7779]. Similarly, design and use of information
technologies are considered to be ingrained and inseparable from the broader social systems in which they reside [80]. These ideas
are accepted in conceptual modeling. Hence, Yu [76, p. 100] explains the benefits of i* as follows: ‘‘unlike traditional systems analysis
methods which strive to abstract away from the people aspects of systems, i* recognizes the primacy of social actors’’ (emphasis added).
Fourth, theoretical foundations of conceptual modeling engage with the notion of system. Hence, as we already discussed, work
systems theory is positioned as a foundational theory underlying information systems [7779]. Another theoretical foundation of
conceptual modeling is ontology [55,8185]. Ontology is a branch of philosophy that studies what exists. A popular ontology in
conceptual modeling, the Bunge Wand Weber (BWW), contains the notion of system. In BWW, ‘‘a set of things is a system if, for any
bi-partitioning of the set, coupling exist among things in the two subsets’’ [86, p. 222]. Some extensions of this ontology, namely
Bunge Systemist Ontology (BSO) [87] and Realist Ontology of Digital Objects and Digitalized Systems [88], extend and modify BWW
by incorporating additional systems constructs, including emergent properties, mechanism, and levels.
In addition to the ubiquity of systems notion in discourse related to conceptual modeling, as these examples show, there is a
great diversity of ideas, approaches, and theories related to systems. Hence, if a given conceptual modeling project were to explicitly
adopt a ‘‘systemist perspective’’, it is not clear which approach the modeler should choose and what the basic elements of systems
are that need to be represented.
2.3. Conceptual modeling languages and the system construct
Beyond the general presence of systemist notions in conceptual modeling discourse, aspects related to systems are present in
conceptual modeling languages themselves. Many conceptual modeling languages, such as UML Class Diagrams or Extended ER
diagrams, contain constructs such as ‘‘part of ’’. These are systemic notions because they deal with composition of complex entities
(i.e., systems). Other languages may not contain explicit system constructs but can be interpreted as being systemist. Thus, the
dependencies in i* can be considered emergent properties, which emerge as a result of the interactions among actors.
Some niche modeling languages provide greater support for systems; most notably, the Systems Modeling Language (SysML),
a modeling language for systems engineering applications [89,90]. Although it uses the term ‘‘system’’, SysML lacks precise, well-
grounded definition of system. The references to ‘‘system’’ are generic and vague. That is, SysML supports the specification, analysis,
design, verification, and validation of a broad range of systems, without providing a precise conceptual characterization of what,
exactly, a system is. As an extension of a subset of the Unified Modeling Language (UML), SysML inherits the conceptual imprecision
of significant concepts. The SysML language’s extensions were designed to support systems engineering activities from a generic
methodological perspective. Furthermore, SysML does not engage with basic systemic notions, such as emergent properties, except
in a very incidental manner (e.g., [90], p. 335).
Overall, the construct of system and its related constructs have been surprisingly underrepresented in conceptual modeling theory
and practice, including in popular conceptual modeling approaches [36,37,55,91]. Furthermore, there is considerable ambiguity
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
when referring to systems in IT [92]. Remarkably, if one could ask systems engineering experts for a precise definition of a ‘‘system’’,
it is most likely that many different definitions would be provided. Indeed, such is the case among scientists as well [93].
Next, we seek to better understand the notion of system and provide an ontologically supported characterization of the ‘‘system’’
construct and its related constructs.
3. Understanding the nature of systems
The term system is Greek in origin (systema), with original meanings of ‘‘organized whole, body’’ as well as ‘‘standing together,
standing in relation, or togetherness’’ [92, p. 209]. It may, however, have an even older Sanskrit root, from the cognate word
samsthana, which also means ‘‘being, existence, life’’ and ‘‘standing together’’.2
Once introduced in the 17th century English, the term eventually became an integral part of the vocabulary in philosophy, natural
and social sciences, engineering, humanities, and medicine. It acquired an additional sense, subsuming an old saying commonly
attributed to Aristotle: ‘‘The whole is something over and above its parts, and not just the sum of them all’’ [94].
3.1. General understanding of systems in science
Today, system is among the basic scientific notions. Indeed, progress in sciences often occurred when what was once considered
indivisible (e.g., atom) was later understood to be complex and was conceptualized as systems [9598]. It is also notable that, in
the field of information systems research, which deals with the design, use, and impact of IT on individuals and collectives, there
have been repeated calls for more systemist theorizing [66,99101].
System is considered to be a unifying scientific construct [95]. Unfortunately, each discipline, and even sub-disciplines,
understand the notion of system in a somewhat unique way, leading to over 100 different definitions and senses of the term [92].
Considering the interdisciplinary nature of a system-based view of reality, general systems theory (GST) was developed by von
Bertalanffy [95] and became widely applied [79,102106]. The basic tenets of GST are as follows. The GST views systems as a
grouping of interdependent parts of a common whole. Thus, Ackoff [107, p. 662] defines a system as ‘‘an entity which is composed
of at least two elements ... each of a system’s elements is connected to every other element, directly or indirectly. No subset of
elements is unrelated to any other subset’’. Some systems exhibit emergent behavior. This common whole tends to be resilient to
change, or homeostatic, giving systems their stability. In some systems, such as organic or certain artificial systems, support feedback
loops exist, wherein the outputs of the system become its inputs, and hence can modulate or amplify the system’s behavior. Systems
may be closed or open, depending upon whether components of the system may interact with the components of other systems.
3.2. Foundations of ontological systemism
Owing to the adoption and further development of general systems theory and its applications to different scientific domains,
many theories and models of systems emerged [6,103,107111].
We adopt the theoretical lens of general ontology, which has been amongst the most important theoretical foundations for
conceptual modeling [55,84,112114]. It is a source of theoretically grounded, consistent, formalized, and rigorous meaning for
the basic notions of what exists. Indeed, conceptual models via concepts and their relationships (i.e., constructs) seek to represent
facts and beliefs about the world by using constructs assumed to be capable of representing these facts and beliefs.
We use the ontological theory of the philosopher physicist Mario Bunge as our guiding ontological theory. The ideas of Bunge
are especially relevant for four reasons. First, they encapsulate the recent advances in sciences, including the debates around the
notion of system. Being a scientist himself, Bunge contributed to these debates, publishing his work in physics, chemistry and
biology outlets (e.g., [115]). Second, Mario Bunge, mainly via BWW, has been fruitfully used as a reference (or benchmark) in
past conceptual modeling research ([2,84,116119]; cf. [120]). Hence, his ideas have already been considered as relevant for the
development of conceptual modeling. Third, Bunge’s approach to systems is a general one, mostly compatible with the views held
by other proponents of systems thinking and of GST.
Finally, Bunge’s objective was the development of a consistent, formalized, and rigorous ontology of systems. This is very
important considering the many disagreements and debates surrounding the notion of system. Furthermore, even Bunge was at
times inconsistent, owing to the great volume of research and evolution of views [87,121123]. In this paper, we adapt and extend
Bunge’s ontology by formalizing it further. In addition, since Bunge did not engage with certain systemist notions which could be
germane for conceptual modeling (e.g., systems as optional mental abstractions), we synthesize the views of Bunge with select tenets
from GST.
Much of familiarity with Bunge in conceptual modeling stems from the BWW ontology developed by Wand and Weber [119,124]
based on Bunge’s Treatise on Basic Philosophy [125,126]. However, this ontology did not offer extensive elaboration of systems and its
related constructs (although, in addition to the system construct, the ontology contained the construct of emergent properties). Upon
review of Bunge’s broader works, Lukyanenko et al. [87] proposed a new ontology, called the Bunge Systemist Ontology (BSO), based
on the writings of Bunge later in his life [96,121,127]. This ontology, as the name suggests, deals with systems more extensively.
However, some germane constructs were still left out (e.g., system level, semiotic systems). Recently, Lukyanenko and Weber [88]
developed a Realist Ontology of Digital Objects and Digitalized Systems by combining Bunge’s ontological notions (including that
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
of systems, levels and emergent properties) with his theories of semantics and semiotics [128]. To develop the foundations of
systems-grounded conceptual modeling, we adopt and extend these, more recent and extended ideas of Bunge.
The BWW ontology postulates that reality is made of things, which have properties [125, pp. 26–29]. Things are ‘‘substantial
individuals’’, which could be composed of other individuals or be simple, structureless and atomic [118, p. 126]. In his more recent
works, Bunge proposed that every object of existence is likely a system. According to the later Bunge, the world is made of systems.
Lukyanenko et al. [87] provide three explanations of this, somewhat radical, ontological position.
First, the notion of system allowed Bunge to reason about constituents of reality that would be difficult to call things. Bunge found
that system is more consistent with the scientific and day-to-day discourse. Scientists routinely refer to fields or atoms as ‘‘systems’’,
and rarely call them ‘‘things’’. Second, this linguistic practice, for Bunge, reflected deeper ontological reasons. Bunge argued that
there are no simple, structureless entities all constituents of reality are complex. Third, as follows from the premise that the world
is made of systems, Bunge asserted that ontological systemism provides a more faithful approach for describing reality, and better
maps to reality [96,129].
The claims by Bunge have been increasingly supported by modern quantum physics, including the candidate for the unifying
theory, the M-theory, which considers fields (e.g., electromagnetic field) as being made of particles, such as bosons [130]. Hence,
fields, which give rise to the physical forces (e.g., electromagnetism), can indeed be considered systems. The notion of known
elementary particles likely being complex is also supported by other quantum theorists [131].
Bunge provides a broad definition of a system as a ‘‘complex object every part or component of which is connected with other
parts of the same object in such a manner that the whole possesses some features that its components lack that is, emergent
properties’’ [93, p. 20]. We adopt the same definition for our analysis. In essence, this definition suggests that, for something to be
considered a system, it needs to be composed of other components. These components need to be connected to one another and,
through these bonds, emergent properties arise.
Bunge defines emergent properties as properties of systems which the components lack, and which only appear once the
components become part of the whole by interacting with one another [129]. What makes systems especially important and
challenging to investigate is that the emergent properties are not directly derivable from the knowledge of the properties of the
components. These properties emerge in an organic way once the components become part of the whole. For example, humans
individually lack the property ‘‘cohesive’’. This property emerges when the humans form a team or family (both, social systems),
and the team is identified as being ‘‘cohesive’’. Similarly, the concept of ‘‘commitment’’ arises as a result of bonding between several
people. The same holds true for human-made systems. A house can be ‘‘cozy’’ when furniture is put together in a particular manner.
Likewise, music can be considered ‘‘soothing’’ despite individual soundwaves lacking this property.
It should be noted: while all entities per Bunge are complex, they do not necessity bond with all other systems to form bigger
systems. An electron somewhere on the Moon does not form a system with a bird on Earth. A pile of laptops does not make a
socio-technical system with the humans in its vicinity. Still, these systems can be grouped together in a mind, for some purpose.
These unrelated groups of systems can be called aggregates (or collections). For example, we can group together Jupiter’s moons,
Australian marsupials, Mario Bunge, and the ER (Entity-Relationship) 2022 conference, to make a point about systemist ontology.
This group, seemingly unrelated, can have properties in common (such as located in the Solar system). However, it is an aggregate, not
a system. What distinguishes systems from aggregates is the presence of emergent properties resulting from the systemic interactions
among components. Hence, it is not ontologically consistent to treat Jupiter’s moons, Australian marsupials, Mario Bunge, and the ER
2022 conference as a system and thus seek a system-based conceptual modeling construct to show it in a diagram.3
In addition to the emergent properties, systems have hereditary or aggregate properties. These properties are also the properties
of the whole, that is the entire system, but they are directly derivable from the properties of the components. For example, total
family income is a property that is the sum of individual incomes of the family members. Similarly, mass of an organism is the sum
of masses of all organs and tissues.
Bunge distinguishes two kinds of system: conceptual and concrete [93, p. 270]. Concrete systems are systems made of energy-
harboring material components and may undergo change. Concrete systems change in the virtue of energy transfer. Atoms,
organisms, and societies are concrete systems. Humans, Jupiter moons and flowers are concrete systems. Bunge views social systems
as concrete, since they are made of concrete components (i.e., marriage involves two or more physical systems) [93].
Aconceptual or construct-system (see, [88]) is a system in which all of the components are mental ideas bound together in the
mind of a thinking system (e.g., human being) via mental rules. Propositions, classifications, and theories are conceptual systems.
Unlike concrete systems, conceptual systems do not harbor energy and change when they are changed by concrete systems.
Bunge suggested that, to represent a concrete system, four elements need to be described, namely, Composition, Environment,
Structure and Mechanism of the system, which are referred to as the CESM model. The composition of the system is its components;
the environment, the external systems (some of which may be ill-understood or ill-defined) with which the system and its subsystems
interact; and the structure, the relationships among its components as well as among these and the environment [111, p. 4]. Finally,
mechanism is the ‘‘characteristic processes, that make [the system] what it is and the peculiar ways it changes’’ [127, p. 126]. To
illustrate how to describe systems using CESM, Bunge [132, p. 39] offers among several examples, a manufacturing plant, which is
a type of socio-technical system:
a manufacturing plant is composed of workers, engineers, and managers; its environment is a market; it is held together by contracts
and relations of communication and command; and its mechanisms are those of manufacturing, trading, borrowing, and marketing.
3Still, as human knowledge expands, what is considered an unrelated aggregate may later be found to be a system, a point Bunge also makes.
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
Conceptual systems can be represented with the condensed Composition, Environment, and Structure or the CES model. Since
conceptual systems do not change by themselves, they do not have mechanisms of their own. The mechanism component is
not applicable to conceptual systems [129,132]. For example, a theory has components (e.g., propositions, axioms, concepts),
environment (e.g., other theories that use the components of the theory in their theories, or other concepts that refer to the concepts
of the theory), and structure (e.g., logical relationships linking the axioms).
Systems can also be described in terms of the level structure the relationships of composition between system components
[129]. Systems at one level (e.g., socio-technical, assume level 1) are composed of systems at a lower level (i.e., technical, and
social, assume level 2). The social level, in turn, can be decomposed into a biological, and then chemical and psychical levels.
The workers, engineers, and managers in the preceding example of a manufacturing plant illustrate the level structure as level 2
components. If necessary, we may decompose the plant example further by considering the parts of the workers, engineers, and
managers such as their organs (level 3), which can be further decomposed still.
Systems have a variety of relationships with other systems beyond composition. Thus, a system can be a type of another
system (e.g., bird is a type of animal, stock market is a type of social system). In this case birds share the properties of animals
(e.g., multicellularity), in addition to having bird-specific properties (e.g., feathers, laying eggs).
The systems that interact with the environment are open systems. Likely, all systems are open, as even experimental artifacts
cannot be fully isolated. However, some systems are more open and susceptible to environmental forces than others (the isolated
tribes of the Andaman Islands interact with other cultures less frequently than the country of Turkey (Turkiye) which historically
been the meeting grounds of different cultures).
Open systems have boundary those components of the system that directly interact with the environment, whereas components
that only interact with other subsystems of its parent system are internal components. For example, a manufacturing plant is an open
system, whose boundaries include legal, HR, public relations and supply and customer service employees, among others. The internal
components of a manufacturing plan include its line workers, security, and control operators, among others. These people generally
do not interact with the environment as members of this system. By interacting with other systems, the systems may alter the
properties, components, structure, or mechanisms of these systems. For example, when a plant produces a car, it may trigger a
desire to buy the car on the part of prospective consumers.
According to Bunge, mechanism harbors the clues for why a system behaves in a particular way. To reveal the mechanism of a
system is to provide an explanation for how and why the system works as it does, which is referred to by Bunge [96,133] and others
[134] as mechanismic explanation. For concrete systems, the mechanismic explanations involve the description of the inner working
of the system. For Bunge, this entailed detailing the different kinds of energy transfers in concrete systems, such as mechanical,
thermal, kinetic, potential, electric, magnetic, gravitational, chemical (e.g., in [127]). Energy transfer leads to change in states of
systems, as they acquire or lose their properties, resulting in events and processes (sequences of events). These changes may also
occur as feedback loops, which, from the GST, are harbingers of natural systems.
In contrast to concrete systems, conceptual systems do not change since they, themselves, do not possess energy. However, energy
transfer occurs within and between concrete systems (e.g., humans who are thinking and communicating about these conceptual
systems). Conceptual systems may be externalized into some medium (e.g., paper, hard drive) in order to be communicated and
shared with others, thus becoming inputs into the design of concrete systems. In this case, the intent to realize a conceptual systems
triggers change (i.e., via energy transfer) in some concrete system, which then acts in the world to implement or instantiate the
conceptual system into properties of concrete systems [135]. This can be accomplished by direct manipulation or by linguistic
declarations, such as commands and requests, or speech acts [136138]. For example, architectural blueprints, engineering models,
and conceptual modeling diagrams, among others, originate in conceptual systems of humans. In thinking about these systems and
mentally relating their properties to properties of concrete systems, humans devise means of realizing them as buildings, bridges
and software code stored on a hard drive or as electrical pulses.
Note that Bunge mainly focused on changes due to energy transfer (e.g., [127]). Whether all mechanisms can be understood
and modeled as energy transfer is debatable. For example, such concrete systems as governments or universities may undergo
state changes driven by speech acts [114,139142]. To describe such mechanisms, we suggest institutional ontologies that seek to
understand social systems in terms of social and psychological dynamics [138,140,143]. Thus, a mechanism can be represented by
physical, as well as social and psychological explanations (e.g., a contract was terminated because one of the parties felt dissatisfied
with the terms).
Some of the energy transfers follow stable and recurring patterns, hence leading to systemic interactions among components; that
is, those interactions that give rise to the emergent properties. For example, the working relationship among employees within an
organization, such as managing and reporting functions, are systemic relationships. If removed, an organization itself may cease to
In contrast, ad hoc interaction among components happen by chance, and do not follow discernible or predefined patterns. These,
typically, do not give rise to the emergent properties within a system, but are still important to account for, in order to capture the
full complexity of the system. For example, lending a lawnmower to a coworker is an example of such ad hoc relationship. It must
be noted, however, that systems are not static, and they change, in part, when ad hoc relationships become more systemic. New
systems can be born out of these ad hoc interactions.
4This is why some scholars define systems as assembly of components which interact with each other on a regular basis (e.g., [144,145]).
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
3.3. Implications of ontological systemist for conceptual modeling
Consistent with efforts to put conceptual modeling on stronger theoretical foundations, we suggest greater consideration of
systems during conceptual modeling. First, systems of all kinds may need to be represented in a conceptual model using one or
more constructs leading to the notion of system as a conceptual modeling construct. System as a conceptual modeling construct is a
representation in a conceptual modeling artifact (diagram, narrative, use case) of a system as perceived by the designer or elicited
from relevant stakeholders. For example, a conceptual modeling diagram may contain a system construct which is assumed to
represent a manufacturing plant (a real-world system).
Second, to represent aspects of a system, one or more systemist conceptual modeling constructs may be used. These are systemist
conceptual modeling constructs. These constructs may present different views of the same system, as per, CESM/CEM model.
For example, a conceptual modeling diagram may contain a construct (e.g., part-of association), which is assumed to represent
a component of a manufacturing plant.
Third, to ensure the systemist conceptual modeling constructs cover important aspects of systems, we suggest a formalism called
CESM+. CESM+adapts the CESM (CES) models of Bunge together with other populates about systems. CESM+is a conceptual
modeling template or checklist aimed to help designers describe and model essential aspects of systems of all kinds. It suggests
that for a concrete system, its Composition, Environment, Structure and Mechanism should be modeled; for conceptual systems the
elements to be modeled include Composition, Environment and Structure. The plus suggests that, in addition to modeling the above
elements, the properties (hereditary and emergent) as well as other valuable or deemed relevant facts about systems (e.g., history
of the system) should be considered for modeling. Among the latter, attempts to anticipate and model emergent properties should
be made.
Modeling with CESM+amounts to providing a comprehensive view of focal systems in a domain from different and converging
perspectives (i.e., knowing properties of systems allows to better understand how these systems change). As we demonstrate through
a case study below, this description should guide the project development toward more appropriate database, user interface and
process choices. The more we understand the relevant facts about the systems of interest, the more we are able to manage their
behavior, including the possibility of anticipating the elusive emergent properties.
To realize CESM+for a given system, multiple systemist conceptual modeling constructs are needed and multiple conceptual
modeling diagrams may be required. This is consistent with the growing trend toward multi-model and multi-representation
conceptual modeling [146150] and model-driven architecture [117,151].
At the same time, note that, although Bunge admits that every constituent in reality may be a system, it is important to underscore
that conceptual modeling is a social activity. It models reality as perceived by human stakeholders, reflecting their needs and views
of the domain. Therefore, even though everything may indeed be a system from a strict materialist point of view, this does not
imply that human stakeholders would conceptualize these entities as systems (mental abstractions) or automatically benefit from
modeling these entities using system and its related constructs.
We suggest a more nuanced perspective, in line with, for example, [6,106]. Skyttner [106, p. 16], suggests that ‘‘[a] system is not
something presented to the observer, it is something to be recognized by him/her. Usually, this word does not refer to existing things
in the real world but, rather, to a way of organizing our thoughts about the same’’. These mental models are effectively conceptual
systems glued together by mental rules. The conceptual systems may or may not accurate or completely map to properties of the
concrete systems, nor even have concrete counterparts. Ptolemaic and Copernican models of the universe are example of these
systems-mental abstractions. Both proved useful, despite one being less accurate than the other. Likewise, organizational stakeholders
who provide information systems requirements, may have different models of systems which may be important to capture. These
models of systems may not agree with one another. An open challenge is to reconcile these differences into a unified conceptual
model which is effective and acceptable by the stakeholders for facilitating development and use of technology (e.g., see [152]).
Although we appreciate Bunge’s basic postulate of the world of systems and maintain that a systemist approach is fruitful but
underutilized in conceptual modeling, we do not suggest the system construct be immediately applicable to all modeling scenarios.
Systems as modeling constructs are only useful when the systemist properties of emergence, CESM and other system-related notions
are valuable to consider and, when possible, represent. Table 1 details the key systems concepts adopted and adapted in the paper
within the context of conceptual modeling.
4. Illustration and further elaboration: Modeling with and without systems
Equipped with the basic ontological notions, we now discuss further the implications of these ideas for conceptual modeling. We
use a case study to elaborate the representational benefits of systems and to illustrate the implications of not representing systems
4.1. Method
We draw on a real case of information systems development within the context of online citizen science. The first author of this
paper has been the primary developer of a citizen science platform, NLNature (formerly,, between 2009 and
2022. This author developed the platform, initially as a developer, hired by a biology department at a mid-sized North American
university. Consequently, the author conducted the initial planning, requirements elicitation analysis, prototype development,
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
Table 1
Key definitions related to systems as used in this paper.
Concept and its sense, where
Definition Reference(s) Examples
‘‘Ontological System’’ or
‘‘system out there’’
an object in the world
Complex object every part or component of
which is connected with other parts of the
same object in such a manner that the whole
possesses some properties that its components
lack that is, emergent properties
Bunge [93, p. 20], Weber
[75] and Wand and Weber
Atom, animal, airplane,
university, stock market, galaxy,
ERP, Google
‘‘System-mental abstraction’’,
‘‘conceptual system’’ or
a potentially useful abstraction
to reason about the world
A mental model of some part of reality which
refers to some ontological system, existing or
Note, conceptual systems are part of reality,
being property of humans who conceptualize
these systems to organize and act in the world
Skyttner [106], Luhmann
[6], Bunge [111] and
Lukyanenko and Weber [88]
Ptolemaic model of the universe,
Copernican model of the universe,
model of local biodiversity as
understood by a biologist, model
of Tolkien’s lore as understood by
a reader, theory of gravity, CESM
model of a factory, the periodic
table of the (chemical) elements
modeling construct’’
a proposed here conceptual
modeling construct
A representation in a conceptual modeling
artifact (diagram, narrative, use case) of a
system as perceived by the designer or elicited
from relevant stakeholders
The definition proposed in
this paper; implicitly
adopted in some conceptual
modeling languages (see
Section 2.3)
Level Structure Model (LSM)
components, modeling a system
using UML stereotype,
proposed later in the paper
Systemist conceptual
modeling constructs
or systems-constructs
Conceptual modeling constructs which
represent different aspects of systems, such as
CESM or emergent properties
The definition proposed in
this paper
part-of construct in UML and
SysML, modeling aggregate and
emergent properties,
proposed later in the paper
and attribute
Feature, trait or characteristic possessed by a
Note: attributes are human conceptualizations
of properties of systems; here, used
synonymously with properties
Bunge [96,132] and
Lukyanenko and Weber [88]
Mass of human, word count of a
novel, color of vehicle, age of
university, shape of a
mathematical function
Hereditary or aggregate
A property of a system that belongs to a
component of the system
Bunge [111,132] Income of a family member, mass
of airplane components
Emergent property A property of a system that does not belong to
any of the composing parts of the system that
arises when the components are bonded
Bunge [127,129] Cohesiveness of water,
productivity of firms, consistency
of theory
Aggregates or collections To the best of existing knowledge, unrelated
(i.e., not directly and continuously interacting)
things and systems
Bunge [111,129] {Jupiter’s moons and Mario
Bunge}, {ER 2022 conference and
3}, pile of cellphones thrown into
a recycling bin
CESM and CES Models Ontological postulate that to effectively
describe a system, one needs to represent
Composition, Environment, Structure and
Mechanism of concrete systems and
Composition, Environment and Structure of
The composition of the system are its
components; the environment, the external
systems with which the system and its
subsystems interact; the structure is the
relationships among its components as well as
among these and the environment, the
mechanism is the characteristic processes, that
make the system what it is and the peculiar
ways it changes
Bunge [96,127] Composition, environment,
structure and mechanism of a
biological family or composition,
environment, and structure of a
theory of thermodynamics
(continued on next page)
conceptual modeling, design and implementation, maintenance, and several redesign phases of the project. Later, the author joined
the research team of the project. These emic experiences permit a rich insider account of modeling with and without systems. At
the same time, the two remaining authors were not part of the project, hence, adding a less involved and biased perspective to the
following analysis.
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
Table 1 (continued).
CESM+Modeling Template A conceptual modeling template used to
describe essential aspects of systems of all
kinds. A systems-grounded conceptual modeling
should strive to represent Composition,
Environment, Structure and Mechanism,
properties (hereditary and emergent) as well as
other relevant facts about systems
Adaptation of CESM/CES
ontological models of Bunge
CESM+is discussed and
illustrated later in the paper
The NLNature project did not adopt the ontological systemism perspective, as described earlier, and used traditional (as well as
experimental) systems analysis and design approaches. We, thus, engage in retrospective analysis a common method in project
management and systems analysis and design which draws insights from post-mortem evaluations of the successes and failures
[154,155]. Specifically, we evaluate the outcomes of the project in light of the systemist ontology provided earlier and suggest
implications for conceptual modeling research and practice.
4.2. Project description
The NLNature project began in 2009 with the aim to develop a citizen science platform for a region in North America. Citizen
science refers to participation of the members of the general public (citizens) in scientific research, including data collection, analysis,
and, more rarely, project ideation and publishing of scientific articles together with the scientists [156159].
Since the advent of the Internet, online citizen science is emerging as a major societal movement and research approach. For
example, is a citizen science platform with over 1.6 million registered users. The citizens, members of Zooniverse,
work on over 50 research projects, ranging from classification of galaxies and identification of animals in the African savannah to
deciphering ancient texts and locating craters and boulders of the Moon. While Zooniverse is the largest citizen science platform, it
is estimated that there are over 3000 active citizen science projects. These are mainly local projects interested in specific research
questions.5One such platform is NLNature a representative example of a mid-sized regional citizen science platform. Indeed, the
project was the regional node of a national citizen science network and the principal citizen science platform for its region of North
The objective of NLNature was to collect sightings of plants, animals, and other taxa in the local region (area of over
400,000 km2). The aim was to create an evolving database of citizen-reported wildlife to support research and policy making related
to plant and animal distributions, environmental change, impact of anthropogenic factors on natural habitats, and monitoring and
conservation of specific species of interest, such as endangered lichens. Another goal was to raise awareness of local natural history
among the residents and tourists of the region.
To support NLNature’s objectives, a target list of species was identified, which became the focus of the first development stage.
Upon subsequent analysis, it was clear that non-experts struggled to report their sightings using this list, so a new development
philosophy was pursued whereby the citizens were permitted to report their observations without classifying the phenomena as
specific species. The species could later be identified using artificial-intelligence-based techniques, such as machine learning and
natural language processing. Indeed, this second phase was when one of the co-authors of the paper switched roles from developer
to co-investigator to spearhead this approach to citizen science.
The project was sponsored by academia, which is typical of citizen science projects [160,161]. What makes this setting especially
interesting is the nearly full transparency of the project development part of the general commitment to open science (of course,
guided by ethical protocols and appropriate participant consent) [162165]. This permits the kinds of revelations that might be
difficult to achieve when working in corporate settings.
The project has been active from 2010 to 2022. During this period, over 10,000 members registered an account on the platform.
They contributed over 10,000 observations of wildlife, making nearly 3000 comments on existing observations and posting over
15,000 photos to accompany the sightings. These sightings received over 10 million user-views. Some of the observations led to
scientific discoveries and resulted in several publications in scientific journals [166,167].
Over the years, NLNature underwent two major redevelopment phases with the aim to better meet the project’s objectives.
Consistent with prevailing approaches to citizen science development [168170], initially the project was developed by focusing on
the needs of the project sponsors: the scientists. Consequently, most of the requirement elicitation and analysis efforts concentrated
on capturing the requirements of the scientists. Early-stage interviews and focus groups with the citizens were also conducted. The
scientists insisted on the major unit of data collection and analysis of the project biological species (e.g., Lung lichen, American
robin), which became the focal entity type of the platform. Consequentially, the citizens were asked to report their observations
in terms of the biological species observed. Fig. 1 shows a fragment of the conceptual model of the project showing how birds are
classified per requirements of the scientists, and the resulting user interface options.
An evaluation phase began as soon as the project was launched and revealed limitations and negative consequences of
approaching citizen science by privileging the views and requirements of the scientists. Specifically, non-experts could not always
positively identify what was observed. Hence, while non-experts could confidently identify familiar species, such as American robins,
they struggled to positively identify lichens, or unfamiliar plants and birds. The analysis of the project logs revealed that often users
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
Fig. 1. Connection between a conceptual modeling fragment (showing how birds are classified per requirements of the scientists) and user interface design in
Phase 1 of NLNature.
Fig. 2. User interface in Phase 2 of NLNature project with a sample observation.
Source: Taken from
resorted to guessing, which was evidenced by frequent changes of the species field while reporting the sightings. This evidence was
further supported through the analysis of user comments and interviews with the existing users [171].
In 2013 the project was redesigned, but this time consistent with an underlying ontological foundation. We chose the BWW
ontology [86] as a guide for this redesign.6Following BWW, the new platform eschewed collecting observations using a pre-defined
list of species and instead collected reports as unique instances, which citizens could describe using attributes or classes of their
choice. After the data were captured, the scientists could infer species, using, for example, machine learning approaches. Fig. 2
depicts a data collection interface of NLNature developed following these ontological foundations. Fig. 3 shows a basic conceptual
model of the redesigned NLNature’s database to accommodate the new redevelopment philosophy.
To evaluate the utility of these ideas, we conducted a series of experiments and focus groups with the prospective and existing
NLNature users. Collectively, these studies showed that data collection focusing on instances resulted in more observations being
recorded, with less guessing and user frustration.
5The estimate is based on the projects listed on SciStarter, the world’s largest database of citizen science projects; see
6Specifically, the central tenet of BWW is that the world is made of things substantial individuals. Things possess properties, which people can conceptualize
as attributes. Things form classes when they share common properties. These ideas have been interpreted by researchers as the need to model individuals
(instances) irrespective of the classes they belong to Parsons [172] and Parsons and Wand [51]. The redesign of NLNature was inspired by these ideas and
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
Fig. 3. A conceptual model in Phase 2 of NLNature. . (For interpretation of the references to color in this figure legend, the reader is referred to the web
version of this article.)
4.3. Systemist analysis of the project
The systemist perspective enables us to better understand the shortcomings of modeling agnostic to systems. First, we briefly
analyze the first phase of the project, before turning attention to the second one and offering a deeper systemist analysis.
4.3.1. Phase 1 systemist analysis
The benefits of modeling with systems can be evident in the simple case of Fig. 1 from the first phase of the project. Equipped
with the systemist ontology (Section 3), we can now interpret this figure in systemist terms. Fig. 1 is an externalized conceptual
system realized as a UML diagram (concrete system, captured on paper and then in software). The diagram represents concepts
and hierarchies, or conceptual classification systems, of the domain of interest of NLNature, elicited mainly from the scientists.
This diagram ultimately informed the development of concrete technical sub-systems of NLNature, a complex socio-technical system.
These included psychical user interface and database structures (electric and magnetic charges representing binary instructions and
containers) for processing and storing citizen observations of nature, developed in accordance with the concepts and hierarchies
shown in Fig. 1.
Assume Fig. 1 fragment is representative of a complete model of the conceptual classification system upon which the user
interface is built to collect sightings. From the systemist perspective such model shows some of the composition and structure
of this conceptual system. The composition are the classes in the diagram (e.g., bird, seabird). The structure is the type of mental
rules connecting these classes that of inheritance where each class is a type of another and shares the properties of its parent.
The diagram does not show the environment. In particular, the diagram does not show the citizen scientists and others
contributing or using the information organized by these classes. It neither models these systems, nor shows which part of the
conceptual system (and hence, the user interface), with which they interact. In systemist terms, Fig. 1 does not show the system
boundaries; that is, the classes that become the entry points for the citizen scientists (and others) into this conceptual hierarchy.
The figure also does not show any emergent properties of the conceptual system, or the components of the socio-technical systems
shaped by this diagram.
The lack of modeling of system boundaries and of emergent properties may complicate building accessible and usable interfaces.
This realization became apparent to the project development team only after several years of NLNature’s deployment. When the
first phase of NLNature was launched, much of the user frustration and attrition was attributed to the lack of domain expertise on
the part of less knowledgeable citizens. This, we reasoned, manifested in the inability to positively identify the observed organisms
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
Fig. 4. Example of a System Boundary Model (SBM) fragment. Classification structure as a conceptual system (call-out bubble); boundary objects of system
shown as classes with bold outlines; important environment objects shown as icons of citizen scientists and project partners.
as predefined biological species. Several laboratory and field experiments, along with user interviews, corroborated this hypothesis
[167,173,174]. However, systemist perspective reveals additional causes of negative user experience, due to unexpected emergent
properties and underappreciation of the systemist boundaries.
The classification system in Fig. 1 of the project contained several emergent properties. In particular, one emergent property was
overall accessibility. In the context of the project, overall accessibility can be understood as the extent to which the entire arrangement
of the classes accords with expectations and knowledge of its users, and hence is usable and accessible for reporting sightings. This
is an emergent property of the sightings reporting sub-component of the NLNature socio-technical system, which was designed in
accordance with the classification conceptual system of the scientists. This property emerged when all of the classes were arranged
in a particular way in the user interface under the guidance of the conceptual model in Fig. 1. However, while the list of species
was carefully considered (e.g., those species deemed completely inaccessible and esoteric were removed), the overall arrangement
was not.
Indeed, much of research on accessible citizen science deals with a single (e.g., generic or species level of classification), not the
overall arrangement of different classes [43,157,175177]. In addition, consistent with the species focus, the impact of intermediate
classes, such as Seabird or Land Mammal, was also not taken into consideration, despite these classes becoming data collection
options (or, in systemist terms, system boundary).
The later analysis of the project showed that some of the negative experiences of the users were caused not by the familiarity
with the individual classes, but by the overall accessibility of the classification system. For example, when reporting on a polar
bear sighting, it was initially assumed that non-experts would be familiar with this class (which, for the most part, is a reasonable
supposition). However, to get to the polar bear class, a user had to first select other top-level classes. In the project, following
standard biological nomenclature [178], polar bears were modeled as a subclass of ‘‘marine mammals’’. However, for many non-
experts in biology, polar bears are first and foremost, bears, and hence, land-dwellers. Hence, some users became lost upon failing
to find a polar bear under the ‘‘land mammal’’ higher-level category. This and similar examples (e.g., puffins are seabirds, rather
than shorebirds despite commonly being observed by the shores) reveal how the knowledge of a single system component (e.g., of
a given species) may still preclude from a successful or friction-free observation due to issues in overall accessibility and failure to
account for the boundary of the system. These realizations, although somewhat intuitive, escaped the original analysis.
A possible alternative to Fig. 1 diagram is shown in Fig. 4. It modifies a standard UML class diagram to depict the classification
structure as a system (here: a call-out bubble); the boundary objects of the system are shown as classes with bold outlines; the main
environment objects are shown as icons of citizen scientists and project partners. We call it a System Boundary Model (SBM), as its
goal is to show the boundary components of systems. Boundary objects are especially valuable to model since often unanticipated
events occur when components of one system begin interacting with the components of another system [179]. This gives rise to the
emergent properties of the new system that forms as a result of these interactions. As a popular adage goes, innovation happens at
the seams.
4.3.2. Phase 2 systemist analysis and CESM+example
In the second iteration of NLNature, a new conceptual system was developed, shown in Fig. 3 (see earlier). The changes appeared
to have addressed the shortcomings of the first design. We now apply the CESM+modeling template to offer a general systemist
analysis of NLNature as a whole.
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
Table 2
Analysis of NLNature based on CESM+template.
System name and type NLNature, an information system, a socio-technical system
Components NLNature technology, composed of such components as programming code, database, application programming
interfaces (APIs, such as Google Map, social media connections), and a series of hardware components (e.g., a
webserver), and the social systems (scientists and citizens)
Environment NLNature partners, media, conservation agencies, and government, as well as physical systems objects of observation
plants, animals and other taxa observed and reported by the citizens which are analyzed by scientists
Structure Reporting a sighting, emailing another member and asking questions, and the relationships between the users of the
project and the objects in the environment
Mechanism Making and posting of observations, reporting on information using the features of the project, communicating among
project members, helping others to identify species, using the data for scientific analysis
Emergent properties (select) Shared sense of the project’s purpose (discussed later)
Observations-anchors (i.e., existing observations which provided strong examples and influenced citizens for future
observations, discussed later)
Socializing (use of platform features in unanticipated ways to elicit off-line encounters)
Research productivity
We suggest following a CESM+checklist at the onset of a project. Effectively, it guides designers in considering what systems are
under consideration and what components of these systems need to be considered and, possibly, represented in conceptual modeling
diagrams. As with most conceptual models, CESM+template can be applied retrospectively, to better understand an existing system
(much like an entity-relationship diagram can be used to describe an existing database). Thus, Table 2 uses CESM+template for
post-hoc analysis of an existing system. By glancing at the table, the scope and essence of NLNature become more apparent.
CESM+Analysis of NLNature. We also use the CESM+checklist to guide the discussion on the readiness of existing conceptual
modeling approaches for systems modeling. Guided by the structure and contents of CESM+as shown in Table 2, we now: (a)
offer a systemist analysis of the diagrams in Figs. 1 and 3, and (b) conduct a general assessment of existing conceptual modeling
capabilities for modeling systems.
Components, Environment, Structure and Mechanism. First, without an explicit systemist perspective, many projects would
lack a diagram which represents all the components of the system. Indeed, Fig. 3, which is the structure of the database, does
represent some of the components of NLNature. Its vast coverage, however, may give a false impression that most focal components
about which information is stored are captured. However, this is not the case. Not all informational components are shown in
the conceptual model, and thus either need to be inferred, or found in other diagrams. For example, we do not see such NLNature
components as the scientists. It should be noted that NLNature had dedicated software elements focused on scientists, such as analysis
and reporting and project administration. Hence, they were objects of a database, but were modeled separately and informally. The
database schema for scientists was created in an ad hoc manner without formal conceptual modeling. Not modeling some database
objects is a common practice that often complicates documentation and may undermine security [180]. Had the explicit systemist
perspective been adopted, such omission would be inconsistent with systemist philosophy and constitute a methodological error.
Hence, adopting a systemist perspective makes modeling more disciplined and systematic.
Of course, it is possible to envision additional diagrams, which could represent these components. For example, an i*
Strategic Dependency Model [32,76] may include scientists together with citizens. Such a diagram may be particularly useful for
understanding the goals, dependencies, and resources involved in the interaction between citizens and scientists. However, absence
a systemist approach, there is no reason to expect that, for a specific project, the system is modeled as a whole and its components
are analyzed and represented. The result is a lack of a holistic view of the entire system. Among other limitations, not having the
holistic view of a system makes it is even more difficult to predict the emergent properties.
Absent an explicit systemist perspective (and a checklist such as CESM+), there is no guarantee that the focal objects of
the environment of NLNature are modeled. These include such important entities as the federal agencies which sponsored the
project, several departments within the university, and public agencies which relied on the data from the project. Among the
important environment entities are other socio-technical systems that are partners of the NLNature project. For example, one such
partner had an agreement to extract data related to oceans (such as ocean currents and pollution). Having this information in
the model, for example, would alert systems developers as to which objects are boundary. Inattentive changes to code of such
objects may undermine interoperability between NLNature and its partners. As shown in Fig. 4, it is generally possible to represent
the environment, with modifications to existing conceptual modeling grammars (constructs and rules for how to use them for a
particular conceptual modeling language). However, very little is known how to do so effectively, while balancing other competing
objectives of conceptual models, such as parsimony and clarity.
Existing conceptual modeling languages offer support for the structure of systems. The structure is commonly modeled as
relationships among system components. These can be represented using, for example, an ER diagram or UML object or class
diagram. Hence, in Fig. 3, some of these relationships are shown by the associations among classes (e.g., Observer and Observation
implies an observe relationship). Existing conceptual modeling languages can make these relationship links more explicit by naming
them, as well as identifying their directions.
However, no method exists for showing the impact of these relationships on the growth and evolution of a system a key
point in describing these relationships under a systemist approach. Indeed, some of these relationships may be more important
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
than others for ensuring system stability, whereas some relationships may be more transient and ad hoc, with less impact on the
longevity of the system. Intuitively, we can reason that the posting of comments is less important than the posting of observations.
Knowing this, suggests additional care and resources dedicated to the observation (technical) sub-system of the project, compared
to the comments sub-system. This could be an important information for developers who lack deep domain knowledge (which can
be the case in outsourcing) [181183].
The final CESM component is mechanism. For example, the making and posting of observations is a key mechanism, which, if
absent, or substantially impeded, nullifies the entire project. For the NLNature system, this is a foundational mechanism. Indeed,
the evolution of this mechanism accounted for most of the code changes during the different iterations of the project. Again,
such realization could help prioritize development efforts and resources. In contrast, the mechanisms involved in contacting other
members are secondary to the project, and thus, are on the periphery of the NLNature system. Hence, any changes to this mechanism
may occur without affecting the functioning of the entire system, and may not require as much care.
Modeling of mechanisms is possible using, for example, process oriented conceptual modeling languages, such as BPMN, EPCs,
or statecharts. However, these notations are not specifically designed for representation of mechanisms, as understood by the
ontological theory. Rather, they are focused on the representation of information flow or decision logic. From the point of view of
systems theory, mechanisms are the explanations for how and why the system works and evolves. The process models are presently
equipped at handling the how part (see, e.g., [184]). They do not deal with the why. For example, these models would not explain
why some observations by one member were similar to observations by another, which, we hypothesized were due to anchoring
effects, as discussed below. Potentially, other conceptual modeling languages, such as goal-oriented, or actor-oriented, languages
and narratives [147,185], may be used for the why question. The challenge then becomes to combine the how and why in an
effective manner. There is no answer to this in the extant conceptual modeling theory.
Sub-systems. Many components of the NLNature system can themselves be modeled as systems following own CESM+template.
Fig. 3 shows some of these distinguishable sub-systems, such as user observation system, user communication system, or classification
system, among others. Of course, even users themselves are systems. However, it is difficult to imagine a scenario where modeling
them as systems may be advantageous for this project; yet this may become important for other projects. Indeed, such crowdsourcing
platforms as PatientsLikeMe (, the world’s largest personalized health network that helps people
find new treatments [186189], may benefit from modeling human organs and tissues. When beneficial, CESM+can be applied
recursively, to model these subsystems. Notable here is existing conceptual modeling language do not explicitly have an ability to
connect these different CESM+representations together.
We now illustrate modeling challenges related to one specific sub-system: user observation system and its referent objects in
the NLNature environment. It was implicitly clear to the development team and the scientists that the plants, animals and other
taxa represented by Natural Object in Fig. 3 are complex; that is, systems. However, they were all modeled as individuals, atomic
entities, since, for a project which had hundreds of species, it was not practical to have hundreds of conceptual models of puffins,
lung lichens, polar bears, and others.
Still, it could have been useful to indicate that the organisms of the project were systems, without engaging in full-blown
complexity modeling. As a result of modeling entities in Fig. 3 as structureless classes, neither the database structures nor the
user interface supported the collection and storage of the attributes based on the parts of the organism (system) being described.
Hence, some of the attributes reported were applicable to the entire organism (e.g., large, beautiful), whereas other attributes were
attributes of the parts (e.g., blue beak, yellow feet). This meant that interpreting these attributes was difficult, because it was not
intrinsically clear (especially when the processing was done automatically, without human interpretation) whether this was an
attribute of the entire organism or its parts.
Furthermore, frequently, the organisms reported were observed as part of larger biological systems. This too escaped the
appropriate capturing by the interface which implemented the conceptual model in Fig. 3. To illustrate, Fig. 5 provides three
observations. Since the project was modeled on the premise of representing individuals, it was very difficult to represent the object
of these sightings as systems.
Modeling all NLNature organisms as systems was not necessary. However, had there been an ability to simply convey that
the Natural Object in Fig. 3 was a system, it could have sent a signal that more complexity needed to be represented in the
database. This could have been achieved by having flexible interface choices permitting, for example, key–value pairs of attribute-
system parts. These could be stored following a key–value pair data model, such as that of AmazonDB or MongoDB, which permits
unbounded variation, thus supporting the system diversity of NLNature [190,191]. As we can see from this analysis, adopting a
systemist perspective does not always entail elaborate system modeling. Small signals from a conceptual modeling diagram, when
appropriately interpreted, can be valuable. However, to create and appropriately interpret these small changes to modeling diagrams,
an update to conceptual modeling methods is needed.
Emergent properties. Finally, the CESM+checklist suggests to consider and attempt modeling emergent properties. To
appreciate the benefits of such modeling, we now consider some emergent properties of NLNature (see Table 2). Two notable
emergent properties, which became apparent after the implementation, are the shared sense of the project’s purpose and observations-
anchors. Specifically, by design, the project was intentionally broad and accepted all kinds of organisms in the specified geographic
area. Over time, as citizens reported thousands of organisms, it became clear that the project began to acquire a crowd-generated
We further hypothesized, this emerged project identity was based on certain popular observations that acted as psychological
anchors [192,193]. These observations shaped the perception of what is interesting to observe, how to describe organisms, and
potentially affected subsequent observations [167]. An analysis showed that most of the observations on the project were of
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
Fig. 5. Real observations where users intended to provide more descriptions of systemic aspects but could not do so.
Source: Taken from
charismatic species, such as fox, eagle, moose, coyote, bear, mink, and seagulls. We suspected having these observations publicly
visible created a grass-roots project identity and biased future observations.
This is not what the project owners wanted. They had hoped to observe a more uniform and representative map of sightings.
The shared sense of the project’s purpose emergent property was not modeled in advance, and hence no mechanisms for promptly
detecting and correcting the drift toward charismatic species was envisioned during the development of the project. The paucity of
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
Fig. 6. Level Structure Models of pre-NLNature systems (left) and NLNature socio-technical system (right) (adapted formalism from [75]); boxes represent systems
at different levels; lines represent composition relationships. Note: in NLNature user interface and data code layers were designed as separate and interacting
subsystems, to permit implementation on multiple devices.
systemist thinking during conceptual modeling dissuaded the conversations about emergent properties of the entire project, as well
as its subsystems.
Presently, conceptual modeling lacks established and robust abilities to detect and model emergence. In the context of the project,
for example, observations-anchors can be shown as asterisks after the names of classes in a UML diagram. However, this does not
permit to quantify the bias due to these anchors. A more comprehensive representation could be based on the visualization of
Markov chains [194] known as Markov network [195]. Markov network is an undirected graphical model used to visualize stochastic
processes as a sequence of possible events where the probability of an event depends on the previous event [196]. Applied to
NLNature, Markov networks could model how a user viewing a set of popular observations, then has a certain probability of reporting
observation with similar properties.7The analysis of the entire network can then shed the light on the emergence of the shared sense
of project’s identity. While Markov networks as a solution would not work for all scenarios, it offers a glimpse of the opportunities
involved in modeling emergent properties.
Level Structure Model. To ensure complex objects are considered in modeling, it would have been helpful to have a map
of components conceptualized in a project as systems. Presently, established conceptual modeling approaches do not permit such
explicit expositions. To appreciate how such diagram could be constructed, we introduce a systemist diagram designed to show the
components of a target system. We call it, Level Structure Model (LSM) of systems with the representation adapted from a formalism in
[75, Chapter 2]. The LSM shows the main higher-level systems in a project. The goal of LSM is to depict the horizontal relationships
between system components related via composition.
There can be multiple versions of an LSM diagram,8as the project progresses from the problem to the solutions space. To illustrate
the usage of level structure models, Fig. 6 shows an LSM of NLNature before and after its creation. Indeed, before NLNature is
7We showed this effect through a controlled field experiment on NLNature [167], although it did not involve visualizing using Markov networks.
8The LSM envisioned here is based on structural decomposition based on hierarchical relations among sub-systems. Other variations of component diagrams
are possible, such as those based on functional decomposition the modeling of systems based on the functions they perform [2]. In organizational design,
which also deals with systems, it was found fruitful to combine structural and functional models (of firms) into matrix models [197] a solution which may
prove useful for conceptual modeling as well.
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
implemented, the corresponding socio-technical system does not exist. The socio-technical system arises only when technology and
people begin to work together to contribute observations of wildlife, as well as to use these observations in their research activities.
Hence, an early LSM version (left) shows two disparate systems the scientific group and a single box for citizens, as although a
citizen is a socio-biological system, citizens are not organized into cohesive systems. With respect to one another, they are aggregates.
Once NLNature is born, citizens become linked with other systems via the technological platform. These observations enabled by
the simple LSM fragment are striking because they help explain some of the future dynamics of NLNature, such as the difficulty in
reaching citizens, attracting them and motivating them to join and continue contributing. Furthermore, the LSM also shows that
citizens in this domain do not form a supersystem with the wildlife, which means they are not intrinsically connected with the
plants and animals. Once NLNature is put in place, it calls upon the people to go out and observe their surroundings.
Finally, this simple diagram underscores the critical importance of design choices for these types of technologies. With the weak
links between citizens and scientists, the technology is a key mediator between them. Any technological barriers in communication
become difficult to detect and overcome. Furthermore, absent NLNature, the incentives for citizens to make observations may be
removed. The second LSM model reinforces these observations.
The post-implementation LSM model offers a high-level overview of the new socio-technical system. From LSM, we can quickly
ascertain the components of NLNature we choose (e.g., based on stakeholder input and domain analysis) to conceptualize post-hoc
as systems in order to reveal their complexity. Hence, Fig. 6 shows that the scientists formed a social system of their own, broken
down into two departments, biology and information systems. Indeed, the scientists created a tightknit network around the project,
shepherding its development and evolution.
In contrast, citizens were geographically dispersed, and largely anonymous to each other (and, to a degree, to the scientists).
Unlike other Internet platforms, such as social media websites Facebook, Twitter or Youtube [198200] or collaborative crowdsourc-
ing, such as Wikipedia [201203], by design, citizens never had an organizing system of their own. Any networks and connections
grew organically by finding secondary uses of the design platform features. LSM shows this by not modeling a separate citizens
sub-system of NLNature. Effectively, citizens, unless self-organized, were aggregates, parts of NLNature individually.
By analyzing the LSMs in Fig. 6, we better appreciate the interaction dynamics of the project, including its information and
power imbalances. The development decisions taken in the past can be better understood with this hindsight model. Indeed, for the
developers, it was much more straightforward to adopt a scientist view of reality (as in Phase 1 development), when the scientists
were a tightknit and well-organized group. In contrast, conducting requirements, and then reconciling and modeling goals, values,
needs of the highly dispersed, heterogeneous, and unorganized citizens presented a significant challenge. With no organization of
their own, the voices of citizens were systematically ignored. This is a notable hindsight, which matches findings on power imbalances
and conflicts in online communities, open source software, and other development settings [204208].
4.4. Case conclusions and further implications for conceptual modeling
From the analysis of the case of NLNature development, the following conclusions can be drawn, with implications for conceptual
modeling. First, we conclude that just about any entity can be conceptualized as a system. Based on Bunge and modern science,
systems are considered omnipresent, and can be found in almost any conceptual modeling diagram. In Fig. 3, strictly speaking,
this includes all classes of the diagram. This, of course, does not mean that every class needs to be shown with a system construct.
However, as the case illustrates, when the complexity of some of these objects becomes important to capture, representing these as
systems becomes valuable.
Second, representing systems appears to go beyond merely showing the components. Hence, the tacit assumption that existing
conceptual modeling constructs are sufficient for representing systems, may not hold. Note that popular conceptual modeling
languages have used the composition construct to depict the relationship between parts and wholes [65,209,210]. Representing
systems also involves capturing the environment, the structure and mechanism of a system, the system’s boundary, the internal
components and emergent properties, among other things. This is not often done in projects, hence, the CESM+checklist can make
modeling more disciplined and methodical.
Third, a key notion of emergence carries implications for conceptual modeling. As the ontology suggests, emergence is something
that happens when the components are put together and begin to operate as a whole. Emergent properties are not directly, or easily,
deducible from the properties of the components, because they emerge over time, as shown in the NLNature case. Herein lies a grand
challenge: conceptual modeling occurs at the early stages of information technology development before the IT is put to use.
This means that, as information systems development shapes systems (e.g., work systems, enterprise resource systems, e-commerce
systems), the a priori modeling of emergent properties may be extremely challenging, if not impossible. Hence, potentially critical
properties of the systems developed with the help of conceptual modeling may escape modeling, and emerge afterwards.
Fourth, an important aspect of systems is the mechanism which, according to Bunge, gives the system its essence and is
responsible for the interaction among the components as well as between the components and other systems (the environment).
To capture mechanism is to explain how and why an event or process happened. For example, what is the mechanism by which
social cohesion among members emerged on NLNature? Why did some observations reach identification consensus and others did
not? Presently, this is an unchartered territory for conceptual modeling.
Finally, the systemic analysis does not mean that the last design iteration of NLNature was a failure. The ontological perspective
taken by the project, which focused on the individuals, appeared to have addressed many of the important shortcomings of the
previous approach. By focusing on the individuals, it allowed users with different backgrounds, levels of motivation, as well as
familiarity and expertise with the natural history domain, to contribute observations using attributes and their own categories.
Still, by ignoring systems, many valuable contributions were not appropriately captured, and many complex nuances lost. Hence,
the adoption of the systemist modeling perspective, while still permitting the users to describe what they observed in terms of
attributes and categories, appears to be a fruitful future design strategy.
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
5. Systems-aware methodological guidelines for conceptual model designers
The implicit treatment of systems ignores the fundamental ontological, and related cognitive and social status of systems
in reality. While there are many outstanding questions regarding how to incorporate systems in conceptual modeling, existing
conceptual modeling languages and methods already permit greater consideration of systems. By synthesizing the theoretical
foundations, as well as the results from the analysis of the case study, we propose the following guidelines for conceptual model
Guideline 1: Every modeling project may entail modeling systems. As the ontological theory claims, as well as evident in
our case study, every entity type (or object, class), can be potentially conceptualized as a system, and hence, can be modeled using
systemist constructs. Furthermore, systems may span multiple entity types or classes (discussed below), so there could be more
systems that are valuable to model than there are classes or entity types. Systems are more ubiquitous than assumed by traditional
conceptual modeling languages, approaches, and methodological guidelines. For some scenarios, such as those found in biology,
complex engineering, or medicine, it may be prudent to assume a default status of all entities as systems. Nevertheless, not every
actual system should be conceptualized as a system. This leads to Guideline 2.
Guideline 2: Model systems when complexity needs explicit representation. Modeling involves abstracting from irrelevant
information that does not advance the purpose of modeling. The same applies to systems. Modeling something as systems should be
beneficial when: the internal complexity of an entity needs to be shown; the emergent properties are important to capture; or when
different system details (belonging to different levels) must be considered. In such scenarios, the additional cognitive and learning
effort, as well as a potential increase in diagrammatic complexity, or the need to develop and consult additional diagrams, may be
offset by the benefits of exposing the system complexity.
As we showed in the NLNature citizen science case, a useful tool for scoping the systemist analysis is the Level Structure Model
(LSM) as introduced in this paper. An LSM depicts horizontal relationships between system components and provides a high-level
overview of the entire system. It can be used to delineate the scope of the systemist analysis for projects, which then guides the
subsequent deeper inquiries covered by the Guidelines 3–5.
Guideline 3: Follow CESM+.Once systemist conceptual modeling is adopted, analysts can follow the CESM+checklist. Bunge’s
conception of systems entails describing them using the CESM model. We adapted this idea into conceptual modeling as CESM+. This
new conceptual modeling formalism should act as a guide for modelers on how to approach systems-grounded conceptual modeling.
It is a roadmap that can help ensure the conceptual modeling diagrams end up with a comprehensive view of focal systems in a
domain from different and converging perspectives.
To realize CESM+for a given system, multiple systemist conceptual modeling constructs are needed and multiple conceptual
modeling diagrams may be required. To develop CESM+conceptual models, analysts are advised to seek most effective and
reasonable [211,212] ways to represent each element of CESM+.
Producing CESM+conceptual models can partially be accomplished without the need to modify existing conceptual modeling
grammars. Hence, the composition of the system can be shown using a part-of relationship in languages that support it (e.g., UML,
ArchiMate). The environment may be shown as other entities that interact with the focal system (as in the example in Fig. 4).
The structure can be shown using relationships. For this, relationship-focused languages, such as ER, ORM, or UML may be used,
but some extensions to these languages may be required. For example, it could be helpful to indicate whether these are systemic vs.
ad-hoc relationships.
Finally, although no direct strategy appears to fully support showing mechanisms and their explanations, elements of mechanism
can be shown using existing methods. For example, to show how systems conditionally react to different inputs, process models
(e.g., BPMN, EPCs, petri nets) can be used. For technical systems, data flow diagrams (DFDs) could be applicable. For discrete-event
systems (such as electric devices), statecharts can be applied [184,213].
To understand why systems change, languages that take a social or agent-oriented perspective are best suited. These include
actor, intention and goal models (e.g., Telos, i*) [61,76,214216]. Finally, auxiliary conceptual modeling tools, such as narratives
[217,218], can also be used to capture the nuances of mechanismic explanations. Other aspects of CESM+, such as emergent
properties, can be represented following the considerations provided in the next guideline.
Guideline 4: Anticipate and model emergence. The plus component of CESM+suggests to model emergence. Emergent
properties are not straightforward to derive and may not even manifest themselves at the time of modeling. At the same time,
some strategies can be effective for anticipating and modeling emergent properties.
We suggest that designers should simulate the lifespan of a system, using tools or imagination. This can be, for example, the
imagination or simulation of the implementation and usage of the artifact built with aid of a conceptual model. Here, such techniques
as agent-based modeling and dynamic system simulation can be useful [99,101,219,220].
Some emergent properties can already be modeled using existing grammars. The dependencies in i* [32,76] are indeed emergent
properties that arise from the agents interacting together. Hence, at least for some domain semantics, such as those dealing with
goals, tasks and resources, a Strategic Dependency models may be used.
Another potentially relevant technique is disciplined imagination proposed by Weick [221,222] within the context of theory
development. Indeed, the anticipating of the application and use of a theory corresponds to the challenge of capturing emergent
properties. In this context, the technique implies a deliberate, and persistent mental simulation of a development or use of the
modeled system as an attempt to foresee its emergent behavior.
Finally, although not definitive, another approach is small-scale ‘‘pilot’’ or ‘‘beta’’ realization and deployment of the technology
based on the conceptual model, in order to observe its emergent behavior in vivo. This technique may prove useful for some scenarios;
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
Fig. 7. Modeling a system using UML stereotype.
Fig. 8. Modeling a system by extending UML grammar to distinguish aggregate (a) and emergent (e) properties.
however, the behavior of a scaled-down system may not always match the behavior of the full-blown system. As Bunge explains,
simulations can be valuable, but they cannot definitively capture all possible emergent properties of systems [129]. Generally,
artificial systems, such as software components of socio-technical systems, under idealized and controlled conditions, are more
amenable to simulation. However, the knowledge resulting from simulations of natural systems (including human social and socio-
technical systems) will not offer a full view of the system since complete reduction of natural systems to its components is impossible.
For a full understanding of the behavior and impact of a system, the system as a whole in its natural setting needs to be observed
(see also, [223]). Still, even limited understanding of system’s emergence can be much more helpful than complete ignorance.
Guideline 5: Model systems by re-interpreting or modifying existing notations
Although a comprehensive conceptual modeling support for CESM+does not yet exist, there are, in fact, many possibilities for
modeling systems by re-interpreting or making minor modifications to existing conceptual modeling languages. Below we highlight
some of the possible options.
Option 1: Model using patterns or templates. Patterns and templates can be used with many existing conceptual modeling
languages (e.g., UML class diagrams) to show typical, representative or anomalous systems in a domain. Hence part-of associations
can be used to show composition; relationships can be used to show structure; activities and gateways can be used to show some
aspects of the mechanism. For example, typical, or most common, species of NLNature can be modeled using patterns. For birds, a
pattern could indicate parts, such as wings, beak, legs, and breast, which are the most common components that users describe using
attributes. For flowering plants, stalk, leaves, and flowers could be modeled as parts. Such models could dictate the database structures
and user interface features to accommodate a more nuanced user input. Hence, when a user attempts to enter an observation of a
bird, a system could present a systemic schema showing the bird components and elicit attributes of the parts as well as the whole.
Option 2: A basic system construct in the diagram. In cases where is too much variations, an option is to alert the interface
developers of the complexly by representing a particular class using an explicit system construct. This may or may not require
significant modifications to the existing conceptual modeling grammars; that is, rules for creating conceptual models (discussed
below). For example, a simple way to show a system could be based on a UML stereotype, as shown in Fig. 7. This may be sufficient
in some cases as a simple way to signal complexity and potential emergence, although none are explicitly shown. Such a construct
could be interpreted as, for example, the need to provide flexible database and user interface capabilities. For example, this modeling
approach could indicate the need to provide key–value pairs or ontology-based data collection to better relate the part attributes to
the whole; or emergent properties that are also expected as attributes of such classes.
Option 3: Extended system construct in diagram. To show emergent properties, more nuanced representations may be needed,
which would go beyond merely indicating that something is a complex object. To illustrate, we propose a tentative multi-entity
system construct shown in Fig. 8. The multi-entity system construct allows to represent cases when a system covers multiple entities,
which in addition to being able to show system components and their relationships (or structure in CESM+), permits distinguishing
between aggregate and emergent properties. Naturally, the introduction of the multi-entity system construct requires a deeper
reengineering of existing conceptual modeling grammars. This is a point we revisit later.
Summary. As evident from the guidelines provided here, adopting a systemic perspective in conceptual modeling can be
achieved without waiting for more extensive research on the various aspects of representing systems. Table 3 summarizes the above
approaches to introducing systemist-constructs into conceptual modeling and their expected benefits. The table also outlines the
general the benefits of systems-grounded conceptual modeling, based on the preceding discussion and examples. This is by no
R. Lukyanenko, V.C. Storey and O. Pastor Data & Knowledge Engineering 141 (2022) 102062
Table 3
Examples of systems-related constructs and their expected benefits along with the benefits of systems-grounded conceptual modeling.
Diagram or pattern
Description and possible implementation Suggested common use cases and modeling benefits
System Boundary
Model (SBM)
Representation of the boundary components of systems;
These can be shown using a proposed System Boundary
Model (e.g., Fig. 4) or by annotating existing structural
diagrams, such as an entry-relationship or UML object or
class diagram
Boundary objects can be valuable to model since unanticipated
events often occur when components of one system begin
interacting with the components of another system
The potential impact of the boundaries on the ways humans
interact with them can be made more explicit
Level Structure Model
The LSM shows the main higher-level systems in a project.
The goal of LSM is to depict the horizontal relationships
between system components related via composition
Provides scope of the systemist analysis (i.e., what are the
systems in the domain or the system-to-be-built worthy of
systemist analysis)
Offers an overview of the larger system
Can be used both before and after the system is being
Systems design
Representations using existing conceptual modeling constructs
(e.g., part-of associations) to show CESM+components of
typical, representative or anomalous systems in a domain
When the diversity of systems is large and it is impractical to
model every system, design patterns can be used for typical,
representative or anomalous systems
Modeling typical or representative systems allows to signal
typical use case scenarios in a domain
UML system
Shows that a class or object in question is complex, without
revealing the complexity
Alerts the interface developers of the complexly of the object,
and hence, the need to have flexible input
Allows to reduce diagram complexity
Useful when diversity of systems can be large (e.g., many
kinds of products), but this diversity does not need to be
explicitly modeled
Multi-entity systems
The multi-entity system construct allows to represent cases
when a system covers multiple entities
Shows emergent properties
Shows system components and their relationships
Permits distinguishing between aggregate and emergent
CESM+Roadmap and checklist for systems-grounded conceptual
modeling. It is a template which reminds modelers to
represent Composition, Environment, Structure and
Mechanism and other facts about systems
A guide for modelers on how to approach systems-grounded
conceptual modeling
Ensures key facts about systems are considered for
Allows to briefly summarize key facts about systems
General benefits of
conceptual modeling
Greater systematization of conceptual modeling activities, especially related to representation of systems
Common concepts and vocabulary for communicating about systems of various kinds
Greater appreciation of the boundaries of systems, and the potential opportunities and challenges at the project’s ‘‘seams’’
Explicit representation of the often-tacit facts in a domain (e.g., systemic interactions, key mechanisms), which could help
guide and prioritize development efforts
More systematic examination of emergence, with potential to anticipate potentially harmful or challenging emergent
More explicit understanding of the relationships between the components of systems and the emergent and aggregate
properties and behavior of these systems
Better guidance for user interface and database design (e.g., by suggesting which complex objects require flexible design
Increased ability to understand, create and manage social and organizational complexity
means an exhaustive list of possible constructs to represent systems. One area of future research is to investigate these additional
means of representing systems, producing an entire agenda for future conceptual modeling scholarship, which we highlight in the
following section.
6. Agenda for systems-focused conceptual modeling research
Systems are the ontological primitives upon which, one could argue, other conceptual modeling constructs can be built. This, we
propose, is a new approach to conceptual modeling, which brings exciting opportunities for future conceptual modeling research.
Below we suggest several fruitful research directions to better incorporate systemist notions into conceptual modeling.
6.1. Research Direction 1. When to use the system construct?
Under the ontological assumption that virtually every entity in existence (even admitting a few exceptions, such as photons or
quarks9) are systems, any object could be conceptualized by stakeholders as a system and hence may need to be represented using one
or more system constructs. This applies both to modeling using abstractions (such as entity types or classes) [34,39] and to instance-
based modeling, in which individual occurrences or instantiations of things form the basis for the modeling [50,51,224,225]. Yet,
9Subatomic </