Discussing the Use of Complexity Theory in Engineering Management: Implications for Sustainability

What is the state-of-the-art literature regarding the adoption of the complexity theory (CT) in engineering management (EM)? What implications can be derived for future research and practices concerning sustainability issues? In this conceptual article, we critically discuss the current status of complexity research in EM. In this regard, we use IEEE Transactions on Engineering Management, because it is currently considered the leading journal in EM, and is as a reliable, heuristic proxy. From this journal, we analyze 38 representative publications on the topic published since 2000, and extrapolated through a rigorous keyword-based article search. In particular, we show that: (1) the adoption of CT has been associated with a wide range of key themes in EM, such as new product development, supply chain, and project management. (2) The adoption of CT has been witnessed in an increasing amount of publications, with a focus on conceptual modeling based on fuzzy logics, stochastic, or agent-based modeling prevailing. (3) Many key features of CT seem to be quite clearly observable in our dataset, with modeling and optimizing decision making, under uncertainty, as the dominant theme. However, only a limited number of studies appear to formally adhere to CT, to explain the different EM issues investigated. Thus, we derive various implications for EM research (concerning the research in and practice on sustainability issues).


Introduction
What is the state-of-the-art literature regarding the adoption of the complexity theory (CT) in engineering management (EM)? What implications can be derived for future research and practices concerning sustainability issues? In EM, addressing these questions through a critical discussion of extant findings is relevant if we consider two, intertwined aspects.
First, in general, the adoption of approaches based on CT has become, in the 21st century, increasingly popular and highly supported. Concerning sustainability related issues, in particular, this is seemingly evident, especially when research grants, funding opportunities, and/or public tenders are released on themes regarding, for example, technology management, open innovation, circular economy, green procurement, or, more generally, sustainable ecosystems [1].
Second, as also highlighted by our analysis in this article, in the 21st century, the use of complexity approaches recurs in decision-making problems, regarding how to improve the effectiveness and efficiency of new product development (NPD), project management (PM), and supply chain management (SCM), or team organization. We know that these aforementioned problems have always been considered as key themes in EM. At the same time, we are confident that, to date, they also represent key challenges towards more sustainable business models [2].
As a piece of core evidence, our analysis shows that many key features of CAS seem to be clearly observable in the dataset, with modeling and optimizing DM under uncertainty as the dominant theme. Perhaps surprisingly, however, only a limited number of studies still seem to formally adhere to CT, to explain the different EM issues under investigation. This is also why, among the various avenues presented, we suggest that more all-inclusive complexity-based research frameworks would be needed. Accordingly, formally embedding fine-tuned co-evolutionary logics in these frameworks could also add value.

Theoretical Background
As previously mentioned, CT represents a multi-disciplinary, modern approach that studies CAS, following its own specific set of laws, behaviors, and characteristics, such as self-organization, and emergence. In principle, CAS can be considered as open systems consisting of several agents locally interacting in a non-linear manner and forming a unique, organized, and dynamic entity; this entity is capable of adapting to, and evolving within, the environment [16]. In other words, CAS have many features in common with living systems; they adapt and evolve through learning.
As mentioned above, a first important characteristic of CAS is the concept of self-organization. The Austrian biologist Von Bertalanffy [17] seminally coins this term in reference to the growth of organisms over time. Self-organizing reflects the ability of CAS to establish an internal organization through adaptation and evolution, without central control.
Relatedly, emergence is a characteristic showed by CAS, where "the behavior of the whole is much more complex than the behavior of its parts" [18] (p. 12). The peculiarity of emergence is that its nature is not necessarily linked to that of the agents [19]. For example, in PM, it has been conjectured that the complex interactions of various parts of a project can generate a specific behavior of the project itself, which can be explained through systemic analysis.
In order to understand how CAS behave, we need to model them, i.e., identify a set of variables that operationally describe these systems. System theory helps with this operationalization [20][21][22]. In particular, we can define a state variable of CAS as a measurable element of the systems that describes their conditions in a given moment. The state of CAS at a given time is, thus, the set of values held, at that time, by all their state variables [11]. In this regard, there is no formal rule for choosing the appropriate number and type of state variables; however, we can assume that the greater the complexity of CAS (in terms of number of agents and level of interdependence), the greater the variety in type and number of the state variables [13]. Moreover, state variables are represented in an n-dimension space, where n = number of state variables. In this space, each point defines a precise state of the systems (such a state is the state space of CAS). Given a set of state variables, the evolution in time of CAS is a trajectory in its state space [14].
Accordingly, another important characteristic of CAS is that their trajectories in the state space can have three main types of behavior [23]: Order, when the trajectory reaches a point (or an orbit) of the space and then stabilizes. This point or orbit is defined as an attractor. The systems in this regime are stable; 2.
Disorder or chaos, when the trajectory shows a chaotic path. In this regime, CAS are completely unstable; 3.
Complex regime (or edge of chaos), when the trajectory is attracted by a particular region of the state space. This particular region is known as a strange attractor. In this regime, the systems reach their dynamic equilibrium.
The most interesting type of trajectory appears to be the third (i.e., complex regime), since CAS in this regime show their most relevant behaviors. When CAS reach the complex regime, the conditions are set for all of its peculiarities, i.e., self-organization and emergent behavior, respectively, to be present. However, despite the tendency of the trajectory to orbit around its strange attractor, the evolution of CAS is generally unpredictable [11].
To date, CAS may be found in different contexts, such as economics (e.g., a market), sociology (e.g., a human group), biology (e.g., a cell), business (e.g., an organization), or EM (e.g., a NPD process). In this regard, approaching these contexts through the lens of complexity can, appropriately help face uncertainty and unpredictability [24,25]. In particular, complexity can help model the real world through describing its main characteristics, especially when the deterministic approach seemingly unveils its limits. To do so, to date there are many methodological tools available in the scientific arena. Agent-Based Modeling (ABM), for example, allows simulating the actions and interactions of simple agents, and capturing the emergent and usually complex behavior of the system to which they belong [26]. ABM could also generate adaptive-learning models, which assume that agents have non-linear behaviors, generally based on very simple agent rules [27]. Another tool is fuzzy modeling, which helps face the ambiguity of complexity contexts by introducing un-precise values for the selected variables [28,29]. Likewise, stochastic models countervail the inability to accurately measure well-defined parameters, assuming that an optimal representation may be indeed found within a probability distribution of such measures [30]. Finally, a contribution to help understanding and modeling of complex systems can also be provided by the system of the systems approach [31] because of its tendency to pool resources and capabilities from single systems into a more complex entity, which performs more than the sum of the systems taken separately.

Analysis
In order to start discussing the impact of CT on EM, since 2000, after different methodological attempts (in terms of search strings and protocols), we ultimately chose to scan only the IEEE TEM journal-considered as the leading journal in the field [15]-through adopting a rigorous keyword-based article search on the EBSCOhost/Business Source Complete research database. In this regard, an initial clarification about the determinants of this methodological choice seems warranted here. This choice happened for two main (intertwined) reasons: First, at the very beginning of our research project, we attempted to adhere to a traditional systematic review protocol (e.g., [32]). In other words, we initially scanned EBSCOhost/Business Source Complete for all of the articles containing, at least, the keyword "complex*" in their abstract (as known, the asterisk at the end of "complex" allows for different, related suffixes [e.g., complex or complexity]). From a strict procedural view, we are confident that, in principle, this methodological choice would have been, perhaps, more appropriate to initially circumscribing the potentially relevant literature in the field. In practice, however, while performing it, this search produced a large amount of results. These results, in substance, would have made the subsequent steps of a traditional systematic review to be rigorously performed in terms of screening, scanning, evaluating, and selecting, substantively not feasible [33].
We then made various attempts to limit the amount of potentially relevant papers through adding more specific filters, e.g., "engineering management", as keywords in their abstract. However, after making some crash checks through looking at the papers' text, we came to the opinion that this choice would have been too risky, in that it would have probably added opacity to the article inclusion (or exclusion) process. For example, various papers focused on complexity-based innovation, PM, or SCM, thus, in line with the focus of the review, do not contain "engineering management" in their abstract. In other words, at least in our view, this choice would have probably brought the risk of biasing the accountability, rigor, and transparency that is at the core of any systematic review process [34].
Second, as a consequence of the above, we attempted to focus only on IEEE TEM to scan EBSCOhost/Business Source Complete for all of the articles containing, at least, the keyword "complex*" in their abstract. This initial step produced 120 results, which then became 111 after eliminating all of the articles published in IEEE TEM before 2000 (our focus is on the 21st century), as well as those articles that could not strictly be considered peer-reviewed (e.g., departmental notes or guest editorials). This initial amount of results, we thought, made the subsequent, needed steps for the article inclusion/exclusion, through a rigorous fit for purpose protocol [35] practically feasible.
On this premise, to ensure substantial relevance for our dataset, we scanned all 111 abstracts. Specifically, to be selected: (i) the article abstracts had to formally adopt CT and/or CAS as their theoretical framework; or (ii) if the formal adoption was absent, the presence of the most vivid characteristics of CT had to be clearly identifiable in the abstracts. In particular, as explained in our theoretical framework, this is the case for characteristics such as ABM, emergence, evolutionary dynamics, fuzzy logics, non-linear dynamics, self-organization, stochastic modeling, system of systems, and uncertainty. Overall, this phase reduced our results to 54. Additionally, to ensure conclusive substantial relevance, we repeated this fit for purpose criterion through reading the article texts of all 54 abstracts selected; 38 articles (2000-September 2019) relevant to our research scope finally emerged. In general, this size is consistent with that of many past (e.g., [36]) and recent (e.g., [37]) more traditional systematic reviews, published in the management arena.
In sum, given the exploratory aims of this conceptual article, we believe that, due to the combined mix between the consistency of our dataset and the IEEE TEM leading reputation in the EM field [15], an IEEE TEM-based initial discussion about the topic coverage can represent: (1) not only a reliable, internationally recognizable, heuristic proxy about the state-of-the-art literature regarding the topic; (2) a (hopefully) challenging starting point to inspire future research efforts in what, as our results show, demonstrates to be a fast-growing, although still not totally conceptually consolidated, area in EM. In this regard, Table 1 synthesizes various, significant items of analysis emerging from our sampled publications. We adapted the thematic areas used in the column "Main Area(s) of Interest" from those present in the ABS 2018 Journal List.     The work provides modeling patterns for the product development process based on its iterative, evolutionary, uncertain, and cooperative characteristics.  The work provides a conceptual model characterizing the complexity features of a supply chain, which is useful to understand the linkage between SCM and delivery performance.
In the four sub-sections below, we analyze these items per key content lines.

Themes
In terms of fields, as a premise, we can consider about two-thirds of our sampled publications as falling into traditional EM, one-third into technology management, and substantially none in emerging technologies ( Figure 1). In the four sub-sections below, we analyze these items per key content lines.

Themes
In terms of fields, as a premise, we can consider about two-thirds of our sampled publications as falling into traditional EM, one-third into technology management, and substantially none in emerging technologies (Figure 1). In more detail, as Figure 2 shows, since 2000 CT has been associated with a wide spectrum of topics and themes associated with the fields above. In particular, as Figure 2 shows, innovation, operations, and management science represent, as we could somehow expect, the most investigated areas. In this respect, works on the use of CT in DM processes, regarding NPD, procurement, and supply chain, or PM, specifically prevail. Interestingly, at the same time, considerable (although minor) amounts of observations fall into the areas of human resource management, strategy, and information management. In this instance, for example, the focus is on the use of CT to increase team productivity, competitive capabilities in (technological) environments, or the efficiency/effectiveness of intraorganizational communication.

Timely Distribution and Authorship
As  In more detail, as Figure 2 shows, since 2000 CT has been associated with a wide spectrum of topics and themes associated with the fields above.
Sustainability FOR PEER REVIEW 13 of 24 In the four sub-sections below, we analyze these items per key content lines.

Themes
In terms of fields, as a premise, we can consider about two-thirds of our sampled publications as falling into traditional EM, one-third into technology management, and substantially none in emerging technologies (Figure 1). In more detail, as Figure 2 shows, since 2000 CT has been associated with a wide spectrum of topics and themes associated with the fields above. In particular, as Figure 2 shows, innovation, operations, and management science represent, as we could somehow expect, the most investigated areas. In this respect, works on the use of CT in DM processes, regarding NPD, procurement, and supply chain, or PM, specifically prevail. Interestingly, at the same time, considerable (although minor) amounts of observations fall into the areas of human resource management, strategy, and information management. In this instance, for example, the focus is on the use of CT to increase team productivity, competitive capabilities in (technological) environments, or the efficiency/effectiveness of intraorganizational communication.

Timely Distribution and Authorship
As  In particular, as Figure 2 shows, innovation, operations, and management science represent, as we could somehow expect, the most investigated areas. In this respect, works on the use of CT in DM processes, regarding NPD, procurement, and supply chain, or PM, specifically prevail. Interestingly, at the same time, considerable (although minor) amounts of observations fall into the areas of human resource management, strategy, and information management. In this instance, for example, the focus is on the use of CT to increase team productivity, competitive capabilities in (technological) environments, or the efficiency/effectiveness of intraorganizational communication.

Timely Distribution and Authorship
As     (2) the country in which s/he was awarded her/his PhD. In this regard, we chose to specifically focus on first authors because of the internationally acknowledged leadership role, which, in general, any first author has in terms of the research design of a publication. At the same time, we preferred to focus on the country in which the first authors were awarded their PhD rather than on their strict nationality because we thought the former could represent a more reliable proxy for the cultural orientation (and associated approach) towards the topic.
Having clarified the above, as shown in Figure 4, the geographical source of our dataset appears substantially balanced between Europe and North America, followed, at the same time, by a significant presence of Far East countries (e.g., China, Japan, Taiwan, Hong Kong, Singapore, and South Korea).
Correspondingly, Figure 5 shows the publications' coverage by author affiliation. In this case, we developed this data-driven figure by considering all of the authors (N = 107, net of duplicates) in our dataset. Interestingly, as shown in the figure, engineering schools/departments prevail, but business schools/departments also occupy a significant portion. at the same time, although in minor On this premise, interesting evidence seemingly emerges if we focus on various features regarding the authorship coverage of our sampled publications (Figures 4 and 5).    Figure 4 substantially shows what we could consider the geographical source of our sampled publications. In particular, we developed this data-driven figure by contemporaneously considering: (1) the first author (N = 37, net of duplicates) of each publication; (2) the country in which s/he was awarded her/his PhD. In this regard, we chose to specifically focus on first authors because of the internationally acknowledged leadership role, which, in general, any first author has in terms of the research design of a publication. At the same time, we preferred to focus on the country in which the first authors were awarded their PhD rather than on their strict nationality because we thought the former could represent a more reliable proxy for the cultural orientation (and associated approach) towards the topic.
Having clarified the above, as shown in Figure 4, the geographical source of our dataset appears substantially balanced between Europe and North America, followed, at the same time, by a significant presence of Far East countries (e.g., China, Japan, Taiwan, Hong Kong, Singapore, and South Korea).
Correspondingly, Figure 5 shows the publications' coverage by author affiliation. In this case, we developed this data-driven figure by considering all of the authors (N = 107, net of duplicates) in our dataset. Interestingly, as shown in the figure, engineering schools/departments prevail, but business schools/departments also occupy a significant portion. at the same time, although in minor     Figure 4 substantially shows what we could consider the geographical source of our sampled publications. In particular, we developed this data-driven figure by contemporaneously considering: (1) the first author (N = 37, net of duplicates) of each publication; (2) the country in which s/he was awarded her/his PhD. In this regard, we chose to specifically focus on first authors because of the internationally acknowledged leadership role, which, in general, any first author has in terms of the research design of a publication. At the same time, we preferred to focus on the country in which the first authors were awarded their PhD rather than on their strict nationality because we thought the former could represent a more reliable proxy for the cultural orientation (and associated approach) towards the topic.
Having clarified the above, as shown in Figure 4, the geographical source of our dataset appears substantially balanced between Europe and North America, followed, at the same time, by a significant presence of Far East countries (e.g., China, Japan, Taiwan, Hong Kong, Singapore, and South Korea).
Correspondingly, Figure 5 shows the publications' coverage by author affiliation. In this case, we developed this data-driven figure by considering all of the authors (N = 107, net of duplicates) in our dataset. Interestingly, as shown in the figure, engineering schools/departments prevail, but business schools/departments also occupy a significant portion. at the same time, although in minor  Figure 4 substantially shows what we could consider the geographical source of our sampled publications. In particular, we developed this data-driven figure by contemporaneously considering: (1) the first author (N = 37, net of duplicates) of each publication; (2) the country in which s/he was awarded her/his PhD. In this regard, we chose to specifically focus on first authors because of the internationally acknowledged leadership role, which, in general, any first author has in terms of the research design of a publication. At the same time, we preferred to focus on the country in which the first authors were awarded their PhD rather than on their strict nationality because we thought the former could represent a more reliable proxy for the cultural orientation (and associated approach) towards the topic.
Having clarified the above, as shown in Figure 4, the geographical source of our dataset appears substantially balanced between Europe and North America, followed, at the same time, by a significant presence of Far East countries (e.g., China, Japan, Taiwan, Hong Kong, Singapore, and South Korea).
Correspondingly, Figure 5 shows the publications' coverage by author affiliation. In this case, we developed this data-driven figure by considering all of the authors (N = 107, net of duplicates) in our dataset. Interestingly, as shown in the figure, engineering schools/departments prevail, but business schools/departments also occupy a significant portion. at the same time, although in minor percentages, Figure 5 also evidences the presence of scholars from other schools/departments, such as information technology or mathematics, and practitioners as well. We could argue that this evidence can be interpreted as consistent, as explained in our theoretical framework, with the multidisciplinary nature of the approaches to CT.

Methodologies, Settings, and Complexity Features
Almost all of the studies are based on conceptual, mathematical modeling, with the vast majority also tested through industrial applications, relying, for the largest part, on quantitative methods. Interestingly, on the one hand, the conceptual modeling is featured by a wide range of techniques, these varying, for example, from genetic algorithms to design structure matrices, or analytical hierarchical/network processes. At the same time, on the other hand, many of these techniques share the common feature of grounding on fuzzy logics, stochastic modeling, or ABM as their basis. From more than one aspect, similar highlighting can also regard the context of the industrial applications. In fact, the general settings are heterogeneous ranging, for example, from aerospace, to automotive, manufacturing, or services. However, almost all of these settings share a strong hi-tech component in what is specifically observed. Figure 6 expands on Table 1, offering statistics about the presence of the inner complexity characteristics in our dataset. In particular, we built this figure through the assumption that more than one characteristic can be simultaneously present in the observed publications.
Sustainability FOR PEER REVIEW 15 of 24 percentages, figure 5 also evidences the presence of scholars from other schools/departments, such as information technology or mathematics, and practitioners as well. We could argue that this evidence can be interpreted as consistent, as explained in our theoretical framework, with the multidisciplinary nature of the approaches to CT.

Methodologies, Settings, and Complexity Features
Almost all of the studies are based on conceptual, mathematical modeling, with the vast majority also tested through industrial applications, relying, for the largest part, on quantitative methods. Interestingly, on the one hand, the conceptual modeling is featured by a wide range of techniques, these varying, for example, from genetic algorithms to design structure matrices, or analytical hierarchical/network processes. At the same time, on the other hand, many of these techniques share the common feature of grounding on fuzzy logics, stochastic modeling, or ABM as their basis. From more than one aspect, similar highlighting can also regard the context of the industrial applications. In fact, the general settings are heterogeneous ranging, for example, from aerospace, to automotive, manufacturing, or services. However, almost all of these settings share a strong hi-tech component in what is specifically observed. Figure 6 expands on Table 1, offering statistics about the presence of the inner complexity characteristics in our dataset. In particular, we built this figure through the assumption that more than one characteristic can be simultaneously present in the observed publications. As evidenced in Figure 6, the study of DM and problem solving under uncertainty (and how to manage it) largely prevails, and generally serves as the ground basis for various lines of inquiry, with one or more complexity characteristic often contemporaneously present with uncertainty itself. In particular, as evidenced in the figure, uncertainty is frequently associated with non-linear dynamics and/or, as previously mentioned, stochastic modeling. The former, for example, is interestingly highlighted by Xirogiannis and Glykas [38] in their study on how performance-driven business reengineering processes work and how they could eventually work better. The latter, in parallel, is used more than once to provide insight on how to model the complexity, towards efficiency and effectiveness, regarding NPD, PM practices, or SCM.
An interesting number of observations also include the use of fuzzy logics in conjunction with uncertainty. In the area of management science, for example, and with a focus on PM, Shafie-Monfared and Jenab [39] use fuzzy modeling to identify different degrees of project complexity, based on the differentiation of managerial and technical features. Their framework can usefully provide support to budgeting, planning, and resource allocation. Similarly, through the case study of a new machining center, Lin and Chen [40] propose a new method to evaluate new product design, based on fuzzy logics in general, and linguistic approximation in particular.
Finally, in our dataset, uncertainty is also repeatedly associated with evolutionary dynamics. For example, Mikaelian et al. [  As evidenced in Figure 6, the study of DM and problem solving under uncertainty (and how to manage it) largely prevails, and generally serves as the ground basis for various lines of inquiry, with one or more complexity characteristic often contemporaneously present with uncertainty itself. In particular, as evidenced in the figure, uncertainty is frequently associated with non-linear dynamics and/or, as previously mentioned, stochastic modeling. The former, for example, is interestingly highlighted by Xirogiannis and Glykas [38] in their study on how performance-driven business reengineering processes work and how they could eventually work better. The latter, in parallel, is used more than once to provide insight on how to model the complexity, towards efficiency and effectiveness, regarding NPD, PM practices, or SCM.
An interesting number of observations also include the use of fuzzy logics in conjunction with uncertainty. In the area of management science, for example, and with a focus on PM, Shafie-Monfared and Jenab [39] use fuzzy modeling to identify different degrees of project complexity, based on the differentiation of managerial and technical features. Their framework can usefully provide support to budgeting, planning, and resource allocation. Similarly, through the case study of a new machining center, Lin and Chen [40] propose a new method to evaluate new product design, based on fuzzy logics in general, and linguistic approximation in particular.
Finally, in our dataset, uncertainty is also repeatedly associated with evolutionary dynamics. For example, Mikaelian et al. [41] develop a holistic, evolutionary approach, based on real option analysis, to manage flexibility and DM under uncertainty. In a similar vein, Giannoccaro and Nair [42] heavily rely on complexity science and evolutionary mechanisms to study what (and how) behavioral traits of project managers can shape their decisions regarding product design.

Complexity-Based Evidences
In relation to the third issue analyzed above, however, it seems that only a limited number of studies still formally adhere to the lenses of CT, and/or CAS, to explain the different EM issues under investigation. For example, in the innovation area, Tripathy and Eppinger [43] focus on complex engineered systems, with particular regard to the offshoring and onshoring activities associated with NPD at a global level. In detail, they use five case studies from electronics, equipment, and aerospace to study the complexity of the interactions between the product and process structures, and the strategies planned and implemented at firm level. On the basis of their findings, these scholars then propose theoretical trajectories aimed at improving the DM configuration regarding global product development in complex engineered systems. As their core idea, the modularity in design and development should be separated from that in manufacturing; furthermore, the development of the system architecture, which is a core capability, should not be offshored.
In a similar vein, Levardy and Browning [44] conjecture the processes of NPD as CAS. These scholars oppose linear, time-based vertical scheduling, in that they theorize these processes as featured by a general class of activities/rules, which can self-organize and adapt to their changing state. The implications of their modeling for DM in EM are interesting; in fact, their adaptive model considers product development as a DM process, in which each decision is potentially able to maximize the expected value of the overall project based on the particular state, in any given moment, of its internal and external variables.
Again, in the context of NPD, the work by Jun and Suh [45] appears particularly worth of explanation. They also provide a theoretical framework for the process, composed not only of iterative but also evolutionary, uncertain, and cooperative characteristics. Through an industrial application in the automotive, electronics, and environmental settings, their modeling demonstrates its potential utility to engineers and project managers involved in planning, organizing, and monitoring the design and implementation of new product initiatives.
Following the above evidences about innovation, in the strategy area, Ndofor et al. [3] use the nonlinear, dynamical system methods from CT to study how different industry environments evolve over time. In particular, adopting three operationalizations, classically utilized to discover nonlinear variable dynamisms, these scholars evidence that many industries evolve in a chaotic regime, where uncertainty increases proportionally to hypercompetitive settings. Similarly, Tsilipanos et al. [30] analyze investments in the telecommunication industry through using a methodological approach typical of CT. Specifically, these scholars model this industry as a system of systems, and use the MATLAB software to create a genetic algorithm able to provide results based on stochastic, emergent modeling. Tested through an industrial application, the more general value of their modeling, also in terms of implications for EM, mainly consist of the possibility to provide prospective investors with theoretical support to efficient DM and budget allocation.
Finally, in the operations and supply chain area, the research by Pathak et al. [46] seemingly deserves attention. Through combining the CAS approach with industrial growth, networks, market structure, and game theories, these scholars investigate how supply network structures can evolve and survive over time. The observations from their agent-based study in the U.S. automotive industry can be of particular appeal to engineers. Specifically, they find that the type of environment and the speed of adaptability both affect the survival chances of supply networks; in peaceful settings, on the one hand, the topological evolution of the networks is relatively stable, with centralized or linear network structures, often able to guarantee survival over the long term. In more competitive settings, however, only the hierarchical structure seems able to provide networks with adequate long-term survival chances.

Discussion and Implications
In this conceptual article, we have focused on the adoption of CT in EM since 2000. At the beginning, we introduced the key conceptual pillars of CT (and CAS). Subsequently, because of its status of being a leading journal in the field, we chose IEEE-TEM as a reliable, heuristic proxy to analyze and discuss those publications formally, and/or substantially, referring to complexity approaches. Therefore, we can synthesize the results from our analysis into the following three main evidences.
First, from 2000, the adoption of CT in EM has been associated with a wide range of key themes in the field. NPD, SCM, and PM prevail. At the same time, a considerable number of observations also regards team productivity, competitive capabilities in (technological) environments, and intraorganizational communication.
Second, this adoption was seen in an increasing amount of publications, especially if we consider the years 2011 to 2019. Conceptual modeling developed through a wide-range of techniques largely prevails in our dataset, then quantitatively tested in various (almost hi-tech based) industrial settings. This, again, also appears in line with the plurality and heterogeneity of analytical tools and (high-tech) settings traditionally employed in EM [47]. At the same time, the common feature among these techniques is that they are mostly based on fuzzy logics, stochastic modeling, or ABM.
Third, many key ingredients of CT seem to be quite clearly observable in the analyzed publications. Accordingly, modeling and optimizing DM under uncertainty results as the dominant theme; this theme, at the same time, is not only often associated with the mentioned fuzzy logics, stochastic modeling, and ABM, but also with non-linear and/or evolutionary dynamics. Perhaps surprisingly, however, only a limited number of studies still seems to formally adhere to CT to explain the various EM issues under investigation.
From what is summarized above, some implications for EM (concerning the research in and practice on sustainability issues) can also be derived. These implications are exposed below, sequentially ordered per item of focus.

Areas of Investigation and Leadership
Regarding the areas of investigation, on the one hand, as previously written, our results show good coverage of complexity-based approaches in key EM areas. On the other hand, we think that additional areas could also become objects of research in this field. For example, the emerging technologies/technology intelligence area could be expanded through complexity-based observations concerning artificial intelligence or Internet of Things. In fact, on both of these topics, we could not find any evidence in our analysis. Moreover, further studies could also look into how to develop, from engineers to leaders; correspondingly, we could find good coverage of human resource management in general, but, apart from scant exceptions, we could not find sizeable evidence about complexity-based leadership [48] in our analysis.
Regarding the above, for instance, and with a focus on the potential impact of complexity-based leadership on the effectiveness and efficiency of innovation (e.g., NPD) and change, the recent work by Burnes [49] appears remarkable. In particular, according to this scholar (p. 84), "unless employees have the freedom to act as they see fit, self-organization will be blocked, and organizations will die because they will not be able to achieve continuous and beneficial innovation." Furthermore, he states (p. 84), "neither small-scale incremental change nor radical transformational change works: instead, innovative activity can only be successfully generated through the third kind of change, such as new product and process development brought about by self-organizing teams." Relatedly (p. 84), "because organizations are complex systems, which are radically unpredictable and where even small changes can have massive and unanticipated effects, top-down change cannot deliver the continuous innovation which organizations need in order to survive and prosper. Instead, it is argued that organizations can only achieve continuous innovation if they position themselves at the edge of chaos". According to Burnes, self-organization is the only way to reach and keep this position, and is itself based on rules that are order-generating. The key point here is that, if the latter (i.e., rules) result in no longer fitting the organizational context, they can be re-created exactly because of the existence of the former (i.e., self-organization).
Having explained the above, a noteworthy example of complexity-based leadership can be offered by a recent case study considering a military organization as a CAS [50], with a focus on its inner complex dynamics, as an enabler to increase organizational effectiveness. As the case demonstrates, despite the traditionally hierarchical and linear characteristics of military organizations, in order to face the surrounding complexity, the rapidly changing defense environment has substantially proved to need a more adaptable and flexible structure.
On this basis, the military leader willing to adopt a complex approach to the commanding action will seek to foster those dynamics typical of CT (such as non-linear relationships and feedback) in order to increase adaptability and organizational learning. This also implies the need to drive the organization from hierarchical to network-centered dynamics, thus assuring governance cohesion throughout the organization, thanks to the development of a shared vision across the top management team. In principle, this perspective can also be considered as presenting similarities with many conceptual underpinnings featuring the notion of socio-technical systems (e.g., [51]].

Settings of Observation and Research Methodologies
Concerning the settings of observation, in a similar vein as above, we could argue that, together with the key high-tech contexts in EM already emerging from our analysis, other central contexts in the sustainability field, such as energy, healthcare, and construction, could become the basis of complexity-based observations. Regarding these contexts, in fact, apart from a few exceptions our analysis could not evidence any specific focus.
Relatedly, with respect to research methodologies, on the one hand, our findings have shown that conceptual modeling tested through quantitative techniques has largely prevailed in the complexity-based observations in EM. On the other hand, however, we maintain that designing and conducting in-depth qualitative case studies [52] should also be important in the field. In this regard, (a) we are substantially in line with those scholars [53,54] who have, for a long time, generally argued that case studies are highly appropriate in complementing computational methods to understand the distinctive features of CAS; and (b) we are particularly in line with those scholars who have used the properties of case studies to develop complexity-based observations in key EM fields, such as NPD.
Taking the above into account, for example, McCarthy et al. [4] used a comparative analysis of three cases to examine how the CAS features of non-linearity, self-organization, and emergence can occur in NPD processes. In particular, these scholars conceive a model of NPD processes, as CAS, featured by three levels of DM, in stage, review, and strategic, respectively. Taking a middle ground between stage gate, chain linked, and chaotic models of NPD, their analysis produces interesting results. In their view, NPD is not necessarily a fixed process; it can adapt and switch from linear to chaotic (and vice versa), thus producing corresponding degrees of incremental or radical innovation. In the practice of EM, their model would be very helpful to avoid the DM traps, potentially regarding the search for fit between (new) product, (new) process, and market demand.

Conceptual Frameworks
Our analysis has shown that, among the many key ingredients of CT quite clearly observable in the analyzed publications, modeling and optimizing DM under uncertainty appears to prevail. Accordingly, we support the recent argument by Baumann and Siggelkow [6] that, in conditions of rationally bounded problem solving, understanding whether integrated (i.e., entirely and simultaneously performed) or chunky (i.e., incrementally expanded) search processes are the most appropriate could also add value. Again, in a technology innovation context of NPD, these scholars focused on this issue through the application of a simulation model. Their analysis has evidenced interesting results: incremental should be preferred to integrated patterns of search when time pressure is not a variable under consideration; moreover, the larger the chunks added at the beginning of the search process, the less the need of a totally incremental search.
According to our results for EM, complexity-based observations have often associated the uncertainty variable with fuzzy logics, stochastic modeling, and ABM, but also with non-linear and/or evolutionary dynamics. As this association has mostly happened on a separate basis (see Table 1), we argue that all-inclusive, complexity-based frameworks could be developed further. Again, this claim corresponds with other key evidence from our analysis: as previously stated, we have shown that, in EM, only a limited number of studies still seem to formally adhere to CT to explain the EM issues under investigation.
The more comprehensive frameworks elicited above could then be tested in different EM settings to assess their reliability. For example, a recent, remarkable attempt of this kind has been the Generalized Complexity Index developed by Jacobs [55]. Based on the three dimensions of multiplicity, diversity, and interconnectedness, this index can be used as an analytical decision tool to evaluate the pros and cons of potential portfolio diversification and/or product differentiation. Furthermore, especially in these learning-based, innovation contexts, distinguishing between complex adaptive and complex generative systems [56] could also be valuable. While the former systems are able to adapt without the need for radical changes, the latter can witness changes which largely modify their inner features and even generate new entities.

Co-Evolutionary Dynamics in Complexity-Based Research Designs
The issue of the interconnectedness brings us to the last item to be discussed in terms of potential implications for sustainability, which is a direct call to embed more fine-tuned co-evolutionary perspectives in complexity-based research designs [57,58]. Specifically, we argue, this call appears to have particular momentum if (and when) hypercompetitive technology environments are under investigation. In fact, recalling what was recently demonstrated by Ndofor et al. [3] on the basis of their 36-year observations of 19 industry sectors, these environments are often chaotic, i.e., featured by a significant degree of a non-linear relationship among elements, together with inter and path dependence. As a fast growing meta-theoretical perspective in social sciences [59][60][61][62], and being generally conceived as the joint and dynamic outcome between industry, managerial, and environmental forces [63][64][65], co-evolution demonstrated effectiveness in capturing all three distinctive features surrounding complexity [66].
In the context of technological entrepreneurship, for example, as maintained by McKelvey ( [67], p. 67), "An entrepreneur could have co-evolutionary dynamics going on in his/her firm; a change in one part of a product leads to a change in another part, which then leads to further change in the part showing the initial change; these changes could affect marketing, production, supply chains, and so on. Finally, it could happen that an entirely new product appears. For example, think of all of the coevolving changes in computer, cell-phone, battery, and touch-screen technologies, computer programming, cell towers, the Internet, and the development of apps that led to current smart-phone products." Similarly, in the context of technological ecosystems, Phillips and Ritala [68] interestingly build (and apply) a specific complexity-based, co-evolutionary framework. In particular, they suggest that three intertwined dimensions, i.e., conceptual (boundary and perspectives), structural (hierarchies and relationships), and temporal (dynamics and co-evolution) should be taken into account to understand (and predict) the behavior of complex ecosystems, especially in the case of an innovation (e.g., NPD) context. Relatedly (and finally), as far as understanding the institutional complexity [69] of co-evolutionary ecosystems is specifically concerned, we are also in line with those scholars [70] who have recently claimed the increasing adoption of a neo-configurational perspective based on qualitative comparative analysis (QCA). Hence, for example, Misangyi [71] recently offered remarkable evidence regarding 28 business facilities projecting and implementing an environmental management system.
More generally, the claim towards the use of QCA is also in line with our claim above (please see Section 4.2.) that more qualitative research methodologies should be adopted to understand the complex nature of innovation-based settings. In this regard, for example, in a novel case study regarding innovation and change in organizational culture, Schlaile et al. [72] used a meme-based approach [73] to investigate the complexity-based interdependencies occurring in a German automotive consultancy firm.

Limitations and Conclusions
Through the results (and proposed implications) of this conceptual article, we do not aim to propose CT as the solution to all of the current EM sustainability-related issues. We also agree with those scholars who, seminally [74,75] or more recently [76,77], have identified the risks of transforming CT, when (even more generally) applied to management, as the fad of modern times. Specifically, we do not believe that this fast-growing approach will totally overwrite all of those theories based on positivism and reductionism [10,78].
Relatedly, we are also conscious that, from a methodological point of view, the results from our analysis present some limits, in that they are, at present, strictly focused on the leading journal in the EM field and on a static explanation. At this stage, in other words, our analysis of the 38 articles should be considered through the lens of a (hopefully useful) initial qualitative assessment, rather than the lens of a quantitative research, which has statistics and trends also aimed at being predictive. In this regard, however, we believe that our results could serve as a heuristic proxy, i.e., a conceptual start to be expanded through more journal-based searches and/or dynamic analyses.
In sum, although aware of the limitations above, and through discussing the implications of our findings, we attempted to explain how CT can contribute to govern many current issues associated with the EM research (concerning the research in, and practice of, sustainability issues). If firms are modeled as CAS, through the identification of agents, their interactions, feedback, and emergent phenomena, CT can then help find novel ways of working to foster a supposed desired emergent behavior (e.g., improved efficiency and effectiveness in NPD, team organization, technology management, or PM); thus, providing engineers and managers with new tools for improving decision-making and performance [79][80][81]. In this regard, for example, Bianchi et al. [82] innovatively deal with complexity management in a recent NPD context through a study of the interaction between stage-gate and agile models (and their associated principles to reduce uncertainty).
Of course, scholars and practitioners argue that, in order to be more than a metaphorical device, a relevant CT framework will need to always be more rigorous from the theoretical, mathematical, and computational modeling points of view [83,84]. We also believe that this modeling will need to be tested in different industry settings to ensure appropriate comparisons between models and real world structures [85][86][87]. In this way, CT may also be taken as a useful approach, for engineers and managers, to test the reliability and consistency of more conventional methods intended to improve sustainability.
In conclusion, firms, clusters, networks, and industries, may be seen, from some aspects, as similar to living organisms [88,89], which grow, evolve, and die [90,91]. They can be healthy or sick [92][93][94] and their behavior emerges from their internal qualities and dynamics, which provide complexity to the system, and from their interactions with the environment [95][96][97]. A firm's behavior is both affected by linear control, such as that imposed by bureaucracy or top-down management decisions, and natural, uncontrolled dynamics. If enterprise complexity fits the complexity of the environment, then desired behaviors, such as high performance and synergy, emerge [98,99].
To date, complexity represents one of the main problems surrounding sustainable business. While we think that the application of CT to business cannot eliminate this problem, we believe that it can help reduce it to a satisfying level.
Author Contributions: Conceptualization, G.A. and A.S.; methodology, G.A. and A.S.; investigation, A.S.; writing-original draft preparation, G.A. and A.S.; writing-review and editing, G.A. and A.S.; visualization, G.A. and A.S.; supervision, G.A. All authors have read and agreed to the published version of the manuscript.