Systems dynamics research in management and organization studies: Overview and research agenda

Forliano, Canio; De Bernardi, Paola; Rozsa, Zoltan; Bertello, Alberto Article Systems dynamics research in management and organization studies: Overview and research agenda Journal of Innovation & Knowledge (JIK) Provided in Cooperation with: Elsevier Suggested Citation: Forliano, Canio; De Bernardi, Paola; Rozsa, Zoltan; Bertello, Alberto (2024) : Systems dynamics research in management and organization studies: Overview and research agenda, Journal of Innovation & Knowledge (JIK), ISSN 2444-569X, Elsevier, Amsterdam, Vol. 9, Iss. 3, pp. 1-15, https://doi.org/10.1016/j.jik.2024.100512 This Version is available at: https://hdl.handle.net/10419/327415 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ Systems dynamics research in management and organization studies: Overview and research agenda Canio Forliano a, *, Paola De Bernardi a , Zoltan Rozsa b,c , Alberto Bertello a a Department of Management, University of Turin, Italy b Faculty of Social and Economic Relations, Alexander Dubcek University of Trencin, Slovak Republic c European Centre for Business Research, Pan-European University, Prague, Czechia ARTICLE INFO Article History: Received 31 August 2023 Accepted 6 July 2024 Available online 13 July 2024 ABSTRACT This paper presents a bibliometric analysis of the systems dynamics (SD) research landscape, drawing on 2,091 documents from Scopus and Web of Science. This research employs bibliometric techniques to explore the evolution of the scientific community over the past 50 years and assess research productivity and impact. Through network analysis, the study further reveals the field’s social and conceptual structures. This approach revealed four pivotal thematic clusters, which were discussed based on content analysis: (1) operations research and strategy formulation, (2) behavioral studies and collaborative approaches, (3) dynamic performance management, and (4) systems thinking for sustainable development. The findings reveal a diverse and interdisciplinary trajectory of SD research, reflecting its integration into a broad array of fields and its potential to inform both theoretical and practical applications. The paper concludes by providing targeted recommendations for future SD research, with a particular emphasis on enhancing management and organizational studies through the incorporation of SD methodologies. This includes the potential for SD to influence the design of adaptive strategies, the use of SD in participatory policymaking, and the application of SD tools in promoting organizational learning and sustainability. © 2024 Published by Elsevier España, S.L.U. on behalf of Journal of Innovation & Knowledge. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Keywords: Systems dynamics Systems thinking Bibliometrics Network analysis Research agenda JEL classification: C19 C44 D01 D81 L25 M00 Q01 Introduction In today’s rapidly changing world, where technological, market, and environmental complexities increasingly challenge organizations, the adoption of advanced analytical tools has become indispensable (Bresciani et al., 2022;Forliano et al., 2022). Systems dynamics (SD), which is rooted in the broader discipline of systems thinking, offers a powerful lens through which to understand and navigate these complexities. Unlike traditional econometric models, which often rely on linear assumptions and static relationships, SD excels in modeling dynamic systems characterized by feedback loops, time delays, and nonlinear interactions (Woodside, 2013). In response to these analytical shortcomings, there has been a discernible shift toward complexity theories and asymmetrical techniques that better accommodate the intricate dynamics of organizational systems (Misangyi et al., 2017;Kumar et al., 2022). As a foundational component of SD, systems thinking promotes an understanding of organizations and their environments as interconnected wholes, rather than as collections of isolated parts (Ricciardi et al., 2020). This holistic approach is critical for addressing the multifaceted challenges faced by modern organizations, which are often systemic and cannot be effectively understood through reductionist methods (Dentoni et al., 2021;Mair & Seelos, 2021). SD, as an extension of systems thinking, enable the exploration of how various elements within an organization interact over time, thereby providing insights into potential future behaviors and outcomes. SD uniquely combines qualitative and quantitative methods to enhance the modeling and analysis of complex systems (Sterman, 2000;Bianchi, 2016). Qualitatively, causal loop diagrams help elucidate the relationships and feedback mechanisms within systems, offering insights into the underlying structures and potential behavior patterns. Quantitatively, stock and flow diagrams provide a means to numerically simulate these dynamics, allowing for detailed scenario planning and decision analysis. This integration of qualitative and quantitative dimensions enables a more comprehensive exploration of system dynamics than is possible with traditional methods that rely on static linear * Corresponding author. E-mail address: [email protected] (C. Forliano). https://doi.org/10.1016/j.jik.2024.100512 2444-569X/© 2024 Published by Elsevier España, S.L.U. on behalf of Journal of Innovation & Knowledge. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Journal of Innovation & Knowledge 9 (2024) 100512 Journal of Innovation &Knowledge https://www.journals.elsevier.com/journal-of-innovation-and-knowledge assumptions to estimate net effects (Aminullah, 2024;Hasegan et al., 2018). These capabilities make SD particularly effective in environments where traditional statistical models fail to capture the essence of dynamic interactions. For these reasons, the adoption of SD has become increasingly prominent among scholars and practitioners and has proven to be crucial in decision-making across both the public and private sectors (Borgonovi et al., 2018;Cosenz & Bivona, 2020;Forliano et al., 2020). In our contemporary society, where complex systemic issues underpin major societal challenges, SD is recognized as an indispensable tool for addressing “wicked problems”—challenges characterized by complexity and resistance to straightforward solutions (Wasieleski et al., 2021). SD provides a powerful analytical framework capable of revealing and managing the dynamic and complex interrelations that these problems present. The inherent complexity of such issues necessitates innovative approaches that surpass traditional linear analytical models, advocating for a systemic perspective that is intrinsic to SD (Grewatsch et al., 2023). This approach has been applied across a wide spectrum of domains, effectively addressing persistent issues such as poverty and inequality (Tey et al., 2020), environmental sustainability (Ding et al., 2018), resources and energy management (Del Vecchio et al., 2019;Sun et al., 2017), enhancing public health systems (Darabi & Hosseinichimeh, 2020), improving safety and reducing crime (Xavier & Bianchi, 2020), and catalyzing educational reforms (Maruccia et al., 2020). The broad application of SD and systems thinking in providing insightful analyses and fostering collaborative efforts toward sustainable solutions (Ricciardi et al., 2020) underscores the need for comprehensive research to explore how these perspectives are implemented across diverse fields, thus contributing to the academic discourse and practical applications of systems dynamics. Second, as highlighted by Grewatsch et al. (2023), systems thinking and SD have been conceptualized in various forms over the years. They have been seen as a comprehensive theory aiming for a general understanding of social sciences (Von Bertalanffy, 2010), a paradigm shift from mechanistic or reductionist worldviews to an integrative, systemic approach (Gladwin et al., 1995), a belief system for mindset change (Senge, 1990), a perspective for theory−practice engagement (Lewis, 1991), or a methodological approach for multilevel, complex problem analysis (Forrester, 1994; Sterman, 1994). This conceptual diversity underscores the necessity of synthesizing these various applications and implications, thereby providing clarity and direction for future research within management and organization studies. To capture and illustrate the increasing trend of SD publications in this domain, numerous articles have attempted to systematize SD research. However, these efforts have often been confined to specificfields, such as strategic management (Cosenz & Noto, 2016) or performance management (Oladimeji et al., 2020); specific contexts, such as healthcare (Darabi & Hosseinichimeh, 2020) or tourism planning (Sedarati et al., 2019); or specialized journals, such as the System Dynamics Review (Torres, 2019). This work aims to broaden this perspective by providing a comprehensive and inclusive overview of prior work using SD in management and organization studies. By conducting a bibliometric analysis combined with science mapping techniques, we seek to answer the following pivotal research questions: RQ1. What trends characterize scientific publications on systems thinking and SD as retrievable in management and organization research areas? RQ2. What social and conceptual structures characterize the scientific debate on systems thinking and SD in management and organization research areas? RQ3. How can future research on systems thinking and SD be developed in management and organization research areas? By answering these research questions, this paper contributes to the research by offering a broad and comprehensive systematization of studies on SD as retrievable in management and organization studies, pointing a way forward for future research directions. In addition, practitioners and decision-makers may find a useful blueprint to promote the adoption and development of SD models and tools in organizational and community-level contexts, here considering their managerial and organizational implications. The remainder of the present article is organized as follows: Section 2 systematically describes the research design and the methods employed. Section 3 presents the descriptive results of the bibliometric analysis. Section 4 presents the results of the network analysis and the different thematic clusters that emerged. Based on the previous discussions, Section 5 offers possible future research streams based on several propositions and possible research questions. Finally, Section 6 highlights the paper’s implications, limitations, and further developments. Research design In this work, a bibliometric approach was adopted to investigate the scientific production related to SD, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol, as in other systematic literature reviews in the business and management research fields (Bertello et al., 2023;Kraus et al., 2022). Bibliometrics, which represents a subbranch of informetrics, consists of statistical techniques aimed at measuring both the productivity and impact of scientific research (Cuccurullo et al., 2016;Merig oet al.,2015). By adopting bibliometric methods, more articles than traditional literature reviews could be investigated, ensuring high levels of rigor, transparency, and replicability (Daim et al., 2006;Rey-Martí et al., 2016). In the current study, bibliometrics was used to uncover the underlying structure of research related to the application of SD principles and tools, focusing attention on the business and management domains. Moreover, as in other bibliometric works (Forliano et al., 2021;Secinaro & Calandra, 2020), a network analysis aimed at depicting the underlying structures (i.e., conceptual and social) characterizing the research field was performed. Data collection and extraction After defining the study’s research questions, the second step of a bibliometric study is to determine which keywords must be used to collect raw data. Consistent with previous studies related to the SD topic (Darabi & Hosseinichimeh, 2020;Oladimeji et al., 2020;Sedarati et al., 2019) and the broad investigation aim of this study (Chen & Xiao, 2016), high-level keywords related to systems thinking, SD, and related methods such as causal loops and stock and flow diagrams were used. Third, the database to be investigated to collect the necessary metadata related to publications on SD had to be selected. In the social sciences, the two largest and most reliable databases are Clarivate Analytics’Web of Science (WoS) and Elsevier’s Scopus (Forliano et al., 2021). Therefore, according to the syntax of the two databases, the string was developed using wildcards to truncate the keywords and capture singular and plural variants of the search terms, while it was possible to search for alternatives through the Boolean operator “OR.”The search was conducted in January 2023 to search for documents’titles, abstracts, and keywords, returning 37,605 results to WoS and 55,241 results to Scopus. As shown in Fig. 1, different exclusion criteria were applied, excluding studies not written in English or that passed through a peer-review process. To answer the research questions of this study, all articles that were not classified in the “Business”or “Management”domain were further excluded (Massaro et al., 2016;Tranfield et al., 2003). Although representing a possible limitation of this study, applying these selection criteria also allowed us to perform a more accurate comparison of different productivity and relevance metrics. Indeed, other relevant subdomains concern “STEM”disciplines (i.e., science, technology, engineering, and C. Forliano, P. De Bernardi, Z. Rozsa et al. Journal of Innovation & Knowledge 9 (2024) 100512 2 mathematics) or medicine, whose productivity and relevance metrics are completely different from those of the social sciences. Finally, as in previous systematic literature reviews (Battisti et al., 2021;Zheng et al., 2022), the search was restricted to only articles published in journals ranked 2 or above in the 2021 Academic Journal Guide (formerly the Chartered Association of Business Schools, ABS). In this way, the most relevant and rigorous articles could be collected and analyzed. Hence, after the two databases were merged, 3404 articles were identified, of which 1065 duplicates were removed. Finally, by analyzing the articles’titles and abstracts while maintaining a broad perspective, it was possible to restrict the data collection to a final sample of 2091 records. Bibliometric analysis The final sample of 2091 articles was analyzed through bibliometric analysis. Thus, open-source RStudio software (RStudio Team, 2016) was used for conducting a performance analysis of the scientific literature related to the topic, especially the Bibliometrix package (Aria & Cuccurullo, 2017), which has been increasingly adopted by researchers in similar studies (Forliano et al., 2021;Secinaro & Calandra, 2020) because it enables the creation of a normalized matrix comprising all records extracted from Scopus and the performance of a bibliometric analysis. A performance analysis of studies related to SD was completed by leveraging several indicators built on metadata related to articles, scholars, countries, and journals (Massaro et al., 2016). In this sense, both their productivity and impact on the scientific community could be captured by evaluating general trends characterizing this research field in the business and management domains. Network analysis Another widely used technique in bibliometric studies is network analysis, which maps the underlying structures characterizing a given research field and their evolution over time (Cobo et al., 2012). This analysis is crucial for identifying research trends and gaps in a given field. Thus, the social and conceptual structures of studies on SD were reconstructed. Notably, the former was analyzed by considering the authors’coauthorships. Conversely, assuming that keywords used together refer to themes that are relevant to each other and can be combined into a thematic cluster (Van Eck & Waltman, 2009), the latter was analyzed by investigating keywords’co-occurrences. In particular, both the authors’keywords and index keywords were used. However, to do so, the original sample of 6343 different authors’keywords (out of 10,120 in total) and 7108 different index keywords (out of 16,044 in total) had to be normalized. Indeed, keywords written in different ways but referring to the same term because of singular/plural forms, upper or lowercase letters, British/American English variants, acronyms, hyphens, and similarities had to be reconciled. This analysis was conducted using OpenRefine (ver. 3.3), an opensource tool originally developed by Google for managing and cleaning big data, and successfully utilized in several similar studies (e.g., Montoya et al., 2016). Considering the size of the database, the different specific algorithms embedded in the software and designated for data reconciliation enabled us to obtain more rigorous and replicable results than manual analysis. At the end of the data cleaning phase, 5627 authors’keywords and 6512 index keywords were retained. Thus, the refined dataset was processed in VOSviewer (ver. 1.6.13), which is a powerful tool for visualizing the structure and dynamics of large networks. Indeed, VOSviewer creates distance-based maps of networks based on the similarity measure of the nodes (Van Eck & Waltman, 2010). Descriptive results of the bibliometric analysis This section presents the results of the performance analysis, which was conducted by analyzing articles, authors, countries, and journals as units of analysis. In this way, it is possible to answer RQ1 of this study. The evolution of articles over time Studies on SD started in the late 1950s, when Forrester (1958) leveraged a feedback view and a computer simulation model to investigate complex issues related to order oscillations and subsequent supply chain management. Forrester’s efforts in applying SD in industrial contexts led to his seminal book Industrial Dynamics (Forrester, 1961). Since those prominent studies, research on this topic has been published for more than 70 years, and SD studies have been translated from the engineering and computer science domains to analyze different research areas at varying analysis levels. Additionally, after industrial applications of SD, Forrester focused his attention on managing urban planning issues and applying SD principles to guide the sustainable development of economies, which gave rise to two other seminal books, Urban Dynamics (Forrester, 1970) and World Dynamics (Forrester, 1971). In this sense, as Fig. 2 shows, starting in the late 1990s, scholars began devoting increasing interest to SD, which Fig. 1. PRISMA flow diagram showing the different phases of data extraction activity. C. Forliano, P. De Bernardi, Z. Rozsa et al. Journal of Innovation & Knowledge 9 (2024) 100512 3 reached its first peak in terms of the number of papers published. This increase was further spurred by the formation of a SD research group around Jay W. Forrester’sfigure at the MIT Sloan School of Management. Indeed, some of the most influential scholars from this group have emerged in this research stream. For example, Peter Senge, who wrote The Fifth Discipline (Senge, 1990), a central book in divulgating systems thinking and SD principles to the broad public; John Sterman, who wrote another seminal handbook in explaining SD applied to businesses and organizational learning processes (Sterman, 2000); and John Morecroft, who mainly investigated bounded rationality decision-making problems (Morecroft, 2015). In particular, Morecroft played a fundamental role in spreading SD in Europe by reinvigorating the link between SD and strategic management studies. Although scholars’interest in SD has increased over the years, it has received significant attention only recently, and more than half of the total articles related to this topic have been published only in the last 10 years. This trend can also be explained by considering industry and practitioners’recognition of the relevance of thinking systemically and leveraging SD to address complex and dynamic problems. In conclusion, it can be assumed that SD is still an underdeveloped research stream that offers plenty of progress and studies that can take place. To determine which articles most influenced the scientific debate around SD in the investigated research fields, the number of citations received by each article was considered. Indeed, citations can adequately synthesize the influence of a publication among scholars (Merig o et al., 2015). Table 1 shows the 10 most cited documents of the sample, showing the total citations (TCs) received by other papers in the dataset as accounted for by Scopus and the average citations received per year (TC/Y). Surprisingly, Sterman occurs four times in this ranking and can be considered one of the seminal authors in advancing SD knowledge in the business and management domains. Moreover, as a reference journal for SD studies, it is not surprising that “System Dynamics Review”appears six times in the list of the top ten most influential articles. The most cited paper is a methodological one from Barlas (1996) addressing model validation issues, such as structural and behavioral issues, and counting 950 citations. The second most cited article comes from Daim et al. (2006), in which the authors mixed bibliometric techniques and patent analysis with SD to model the ecosystem surrounding disruptive technologies and forecast their future diffusion. Third, it is possible to find a reflection from Forrester (1994) about the usefulness of SD models to advance theory in the operation research field, followed by two conceptual papers from Sterman (2001,2002), both of which aimed at reinforcing the general awareness about the relevance of adopting a systemic lens to interpret complex systems, as well as using formal models to test decision-makers’mental models and simulating the implementation of different policies. Authors and countries A total of 3552 authors from 65 countries and 2050 different institutions contributed to publishing the 2622 articles in the analyzed dataset. Thus, by leveraging authors as a unit of analysis, their productivity and impact were considered to investigate which scholars mainly influenced business and management studies on SD. Fig. 3 visually portrays the 15 most influential authors, matching their productivity, represented by the number of papers published each year (i.e., bubble size), and impact, represented in terms of citations per year received (i.e., bubble darkness). The TCs per year were preferred to TCs, not to penalize scholars whose careers started in more recent years. Therefore, Saeed, Sterman, Richardson, Andersen, and Morecroft show the most extended timelines, with an unbroken series of publications starting in the early 1980s to date. However, considering the h-index, Sterman (24), Richardson (19), and Lane (18) are among the most influential authors. Indeed, the h-index indicates the minimum number of publications cited at least h times by other scholars in the dataset (Hirsch, 2005) and is considered a well-established and robust indicator that simultaneously combines productivity and relevance (Vanclay, 2007). It is also interesting to note that Rahmandad started publishing in recent years (his first publication in the dataset was released in 2008) but ranks second in terms of citations per year received by the 15 most influential authors. To offer a more precise Fig. 2. Distribution of publications related to SD over time. Table 1 The 10 most relevant documents in the dataset. # Author(s) Title Year Journal TC TC/Y 1 Barlas Formal Aspects of Model Validity and Validation in System Dynamics 1996 Syst. Dynam. Rev. 950 33.93 2 Daim, Rueda, Martin, & Gerdsri Forecasting emerging technologies: Use of bibliometrics and patent analysis 2006 Technol. Forecast. Soc. Change 765 42.50 3 Forrester System dynamics, systems thinking, and soft OR 1994 Syst. Dynam. Rev. 609 20.30 4 Sterman System Dynamics Modeling: Tools for Learning in a Complex World 2001 Calif. Manage. Rev. 549 26.14 5 Sterman All Models Are Wrong: Reflections on Becoming a Systems Scientist 2002 Syst. Dynam. Rev. 539 24.50 6 Rahmandad & Sterman Heterogeneity and Network Structure in the Dynamics of Diffusion: Comparing Agent-Based and Differential Equation Models 2008 Manage. Sci. 456 28.50 7 Dejonckheer, Disney, Lambrecht, & Towill Measuring and avoiding the bullwhip effect: A control theoretic approach 2003 Eur. J. Oper. Res. 437 20.81 8 Vennix Group model building: tackling messy problems 1999 Syst. Dynam. Rev. 375 15 9 Wilson The impact of transportation disruptions on supply chain performance 2007 Transp. Res. E: Logist. Transp. Rev. 345 20.29 10 Gino & Pisano Toward a Theory of Behavioral Operations 2008 Manuf. Serv. Oper. Manag 341 21.31 Note: Papers are ordered by total citations received by other documents in the dataset (TC). The right column reports the total citations received per year (TC/Y). C. Forliano, P. De Bernardi, Z. Rozsa et al. Journal of Innovation & Knowledge 9 (2024) 100512 4 view of the performance indicators associated with each scholar represented in Fig. 3, the productivity (i.e., total publications in the dataset) and impact measures (i.e., TCs received, h-index, and TCs per year received) are also reported in Table 2. Furthermore, considering the authors’affiliations, both countries’ productivity and impact were analyzed. Hence, the top 15 countries, here based on their productivity, are plotted in Fig. 4. In particular, productivity was differentiated to capture the rate of intracountry collaboration (i.e., single-country publication or SCP) and intercountry collaboration (i.e., multiple-country publication or MCP). Thus, the SCP includes publications with all authors affiliated with the same country, while the MCP includes publications with authors from different countries. Out of the 65 total countries involved, only 457 documents were single-authored (approximately 21.86 % of the dataset), meaning that collaboration is a significant aspect of authors adopting or investigating SD. This assumption is further corroborated by Fig. 4, which shows how both advanced economies (e.g., the USA, the UK, Germany) and developing ones (e.g., China, India, Iran) appear among the most productive countries and are all open to multinational collaboration. Interestingly, neither African nor Latin American (except for Colombia) countries appear on this list. Table 3 reports the top 15 countries in terms of the total number of citations received. Therefore, as shown in Fig. 4, the USA proves to be a leader in both productivity and relevance, followed by the UK. Interestingly, almost all the most prolific countries are also the most influential, with only India and Spain giving way to Sweden and Greece. Journals The 15 most prolific journals in which the dataset’s documents were published are presented in Table 4. Not surprisingly, System Dynamics Review ranks first, with 495 publications (i.e., approximately one-quarter of the dataset). Indeed, this journal focuses exclusively on advancing systems thinking and SD and their applications in a broad range of areas (e.g., societal, technical, managerial, and environmental). However, it is interesting to note that the Journal of Cleaner Production also performs very well in terms of publications related to SD used as a theoretical lens and practical approach to investigating sustainability-related issues. The analysis of the other most productive journals present on the list shows that great attention was given by scholars to manufacturing, industrial engineering, and operations research, together with the sociotechnical implications of adopting a systemic view. In addition to journals’ Fig. 3. Top 15 authors in terms of productivity and impact. Table 2 Top 15 scholars in the dataset based on productivity. # Author NP TC h_index PY_start TC/Y 1Sterman J 34 4238 24 1985 114.54 2Richardson G 30 2138 19 1985 57.78 3Lane D 30 1444 18 1991 46.58 4Andersen D 29 1646 16 1988 48.41 5Saeed K 26 273 10 1982 6.82 6Morecroft J 24 727 15 1983 18.64 7Wolstenholme E 22 1044 14 1982 26.1 8Vennix J 20 1280 16 1992 42.67 9Kunc M 18 520 13 2007 34.67 10 Ford D 17 1032 9 1998 43 11 Larsen E 17 368 12 1993 12.69 12 Gr€ oßler A 17 306 9 2001 14.57 13 Rouwette E 16 971 13 1996 37.35 14 Rahmandad H 16 907 12 2008 64.79 15 Naim M 15 782 12 1991 25.23 Note: Records are ordered by the total number of publications in the dataset (TP). Other performance measures are related to citations received (TC), h-index, first document retrieved in the dataset (PY_start), and total citations per year recjeived (TC/Y). Fig. 4. The 15 most productive countries based on authors’affiliations. Table 3 The top 15 countries were ordered by the total number of citations received. # Country TC TC/TP 1 USA 22,957 48.33 2 United Kingdom 10,762 39.57 3 China 5002 25.26 4 Netherlands 3487 40.55 5 Australia 2215 27.01 6 Germany 2158 24.80 7 Sweden 1829 87.10 8 Italy 1201 23.55 9 Greece 1043 61.35 10 Canada 984 23.43 11 Korea 930 22.14 12 Colombia 854 35.58 13 Switzerland 850 24.29 14 Norway 824 21.13 15 Iran 605 18.33 Table 4 The 15 most relevant journals are ordered by the total number of publications in the dataset. # Journal TP TC 1Syst. Dynam. Rev. 495 17,911 2J. Clean. Prod. 229 6598 3Eur. J. Oper. Res. 103 4896 4J. Oper. Res. Soc. 138 4313 5Technol. Forecast. Soc. Change 97 3107 6Int. J. Prod. Econ. 52 2027 7Syst. Res. Behav. Sci. 150 2200 8Int. J. Prod. Res. 66 2104 9J. Constr. Eng. Manag. 26 1048 10 Int. J. Proj. Manag. 18 1031 11 J. Manag. Eng. 21 648 12 Decis. Support Syst. 16 658 13 Reliab. Eng. Syst 20 507 14 Manage. Sci. 12 1518 15 Syst. Pract. Act. Res. 27 392 C. Forliano, P. De Bernardi, Z. Rozsa et al. Journal of Innovation & Knowledge 9 (2024) 100512 5 productivity, Table 4 also considers their impact in terms of TCs. In this sense, however, there are no significant differences compared with sorting journals for their productivity, except in the case of Management Science. Indeed, some very influential articles coauthored by Sterman (e.g., Oliva & Sterman, 2001;Rahmandad & Sterman, 2008; Sterman et al., 1997) have been published in that outlet. Moreover, in Fig. 5, the publication trends of the six most productive journals are shown. In this sense, it can be easily noted that, excluding the System Dynamics Review, the Journal of Cleaner Production’spublication trend outstands all other journals. On the other hand, the Journal of the Operational Research Society and Systems Research and Behavioral Science shows an increasing trend, followed by the European Journal of Operational Research and the Technological Forecasting and Social Change. Discussion of the social and conceptual structures To complete the bibliometric analysis of studies about SD in the business and management domains, this section presents the results of the network analysis performed to provide an overview of the social and conceptual structures characterizing such studies and their authors. Therefore, it would be possible to answer RQ2 of this study. Social structure Concerning the social structure of studies related to SD, the cooccurrences of the top 50 authors (i.e., coauthorship) were analyzed (Forliano et al., 2021). Fig. 6 shows the network resulting from applying the normalization of association strength (Van Eck & Waltman, 2009). In particular, the greater the number of documents authored by a scholar, the greater its node; the greater the number of documents coauthored by two or more scholars, the closer their bubbles appear, and the more robust the links connecting them are. Interestingly, by applying Louvain’s cluster algorithm (Blondel et al., 2008), Fig. 6 shows the existence of 15 clusters (each defined by a different color) among the 50 most influential authors. In this sense, most of them act as isolated nodes or as niche research groups, suggesting the existence of few influential communities of scholars. It must be noted that the largest SD community includes some of the fathers of this discipline (e.g., John Sterman, David Andersen, George Richardson) and the founders of the System Dynamic Review, the reference journal in this research field. Interestingly, most of them studied as Ph.D. students at MIT (such as Saeed or Morecroft, which appear in a different cluster), where Forrester started to teach SD in the early years of such a discipline. Moreover, considering the authors’affiliations, in line with the performance analysis results, a high level of engagement exists around SDs worldwide. This collaboration rate is represented by more robust lines connecting countries in Fig. 7, while the countries’ productivity is portrayed based on color intensity. Thus, a very high collaboration rate exists between China and English-speaking countries (i.e., the USA, Australia, Hong Kong, and the UK), which also collaborate with each other. With respect to Europe, except for the UK, the most active communities can be found in Norway, the Netherlands, and Italy, three countries where some consistent research groups on SD are located (i.e., especially Norway, which hosted the first international conference on SD in 1976). In contrast, there is still a paucity of engagement from Latin American (apart from Colombia) and African authors. Conceptual structure This analysis highlights the relational patterns among keywords that frequently co-occur within our dataset, which are known as coFig. 5. The publication trends of the six most productive journals. Fig. 6. Coauthorship analysis depicting the social structure of the discipline. Fig. 7. The rate of collaboration between countries, here based on authors’affiliations. C. Forliano, P. De Bernardi, Z. Rozsa et al. Journal of Innovation & Knowledge 9 (2024) 100512 6 occurrences (Van Eck & Waltman, 2009). In our study, we examined both authors’keywords and index keywords, with the latter added by professional indexers and sometimes deemed more informative than the authors’keywords (Campedelli, 2020). With VOSviewer, we concentrated on the 100 most co-occurring keywords, each occurring at least 18 times. Using the Louvain algorithm with a resolution parameter of one across 10 iterations (Blondel et al., 2008), we identified four distinct thematic clusters, each represented by a different color in Fig. 8. The size of a keyword’s node in the network signifies its frequency of use by scholars, and the proximity and line thickness between keywords indicate their co-occurrence rates. These clusters serve as preliminary automatic structuring facilitated by VOSviewer’s use of the Louvain algorithm, a community detection method that optimizes modularity to partition the network into clusters of densely interconnected nodes with sparser connections between clusters. According to the analysis presented in Fig. 8, the following four thematic clusters emerged: Cluster 1: Operations research and strategy formulation (blue cluster); Cluster 2: Behavioral studies and collaborative approaches (yellow cluster); Cluster 3: Dynamic performance management (red cluster); Cluster 4: Systems thinking to support sustainable development (green cluster). Following other bibliometric studies and systematic literature reviews (e.g., Bertello et al., 2023;Martínez-Climent et al., 2018; S anchez-Robles et al., 2023), after this algorithmic clustering, we conducted a manual content analysis of the 30 most cited articles in each cluster to extract deeper thematic insights. This manual analysis allowed us to further unveil how the core topics defined the conceptual structure of the topics under investigation. The findings from this detailed manual examination form the basis of the discussions presented in the following subsections. Operations research and strategy formulation The roots of SD as a research field have been deeply connected with operations research since its origins, when Forrester (1958) recognized that a company’s supply chain management could be described as a complex system characterized by feedback loops that imply time delays, nonlinearities, unintended consequences, and suboptimal behavioral decisions. Thus, keywords in the first cluster reflect how SD has been broadly adopted to investigate industrial issues, such as inventory control, capacity building adjustments, oscillations in order backlog, and instability in market shares (Richardson, 1999;Rahmandad & Repenning, 2016). For example, scholars have made great efforts to frame the financial and information flows that can lead to fluctuations in inventories (Barlas & Gunduz, 2011;Fiala, 2005), which were first conceptualized by Forrester (1961) as the famous bullwhip effect. In this sense, several scholars consider SD simulation and mathematical models to be better than traditional linear approaches for framing operation management issues that otherwise would be difficult to identify and handle (Gr€ oßler et al., 2008;Warren, 2005). Indeed, SD can be leveraged to explain every complex system, whose behavior is intimately determined by the interactions occurring among the variables constituting its underlying structure (Sterman, 2000). These variables are mainly related to those resources that can be considered strategic in the closed boundaries of the system under analysis and the capacity to effectively manage the flows between them. For this reason, several SD studies have adopted the resourcebased view (RBV) of a firm and its knowledge-based (knowledgebased view, KBV) or intangible-based (intellectual capital-based view, ICBV) extensions as theoretical lenses for analyzing successful strategies characterizing a firm rather than another (Johnson, 1999; Fig. 8. The conceptual structure of the dataset, here based on co-occurrence keywords. C. Forliano, P. De Bernardi, Z. Rozsa et al. Journal of Innovation & Knowledge 9 (2024) 100512 7 Kunc & Morecroft, 2009;Wassmer & Dussauge, 2012). However, possessing the right resources is not enough to spur firm performance, and the causal relationship between resource acquisition and depletion should also be captured and framed (Bianchi et al., 2010;Kim & Park, 2006;Kunc & O’Brien, 2017). Currently spurred by the rise of the Industry 4.0 paradigm, novel technologies such as big data, cloud computing (Hofmann, 2017;Kochan et al., 2018), and open innovation strategies (Yun et al., 2016;Vignieri, 2020) seem to play a fundamental role in guiding such processes and helping to cope with those regulatory and market-based challenges in existing and emerging markets (Kobos et al., 2018). Given the above, the way managers and decision-makers respond to a given situation largely depends on SD models’capacity to refer to correct assumptions. Indeed, these assumptions guide strategic and operational decisions, such as perceiving orders, evaluating material flows and inventory adjustments, scheduling production, and hiring a new workforce. Hence, SD has strict links with resource accumulation and implementation (Li et al., 2018) and strategy formulation (Gary et al., 2008;Cosenz & Noto, 2016) and can represent a proper approach to avoid capability erosion (Rahmandad & Repenning, 2016). SD modeling can support future research in understanding the nonlinear and delayed effects of supply chain policies and strategies, the causal relationships between the business environment and organizational capabilities, and the role of Industry 4.0 technologies in reshaping operations research in dynamic and complex contexts. Behavioral studies and collaborative approaches The second cluster reveals the interest of scholars in applying SD to understand how models can affect people’s behavioral changes and vice versa; since early studies on SD leveraged some insights from psychology and cognitive sciences (Bendoly, 2014;Bendoly et al., 2010;Gino & Pisano, 2008). For example, Liu et al. (2015) showed how people who act in turn-based simulations are involved in learning processes that iteratively guide their decisions. Therefore, they responded to other actors’decision rules, altering their behavior and, at the same time, the steady state of the system, which, in turn, alters other people’s experiences. Thus, if a great variety of SD studies assume that managers and decision-makers are rational agents, they often do not adequately perceive the underlying structure of the complex and dynamic systems in which they behave. Subsequently, they often suffer from misperception issues, even if they have to deal with simple dynamic systems (Moxnes, 2004;Moxnes & Davidsen, 2016). Although recent studies have focused on revealing the microfoundations of problem solvers’and decision-makers’behaviors (Mohaghegh & Gr€ oßler, 2020) and knowledge management practices (Chen & Fong, 2015), these problems are not new to systems dynamicists. Indeed, they were already recognized in the late 1980s, when Forrester conceptualized a behavioral theory endogenously characterizing actors’decision rules by investigating the experimental scenario of its famous “Beer Distribution Game”(Sterman, 1989). Hence, he recognized how the short-termism and lack of a systemic perspective of people in recognizing the feedback loops characterizing a supply chain could lead to nonlinearities and time delays typical of the bullwhip effect. Based on this conclusion, several articles have investigated misperception problems in experimental settings. For example, Weinhardt et al. (2015) investigated how people’s different cognitive styles and analytical orientations affect their understanding of the accumulation and depletion processes typical of systems characterized by the presence of stocks and flows. In this sense, they confirmed the same results as Cronin and Gonzalez (2007), who found that even highly educated people often do not understand the basic principles guiding stock and flow model behavior. In addition, it must be highlighted that the first applications of SD adopted in organizational settings were mostly leveraging SD specialists as consultants who used to build models without involving the impacted stakeholders in the process (Cosenz & Noto, 2016). However, thanks to the book “The Fifth Discipline”by Senge (1990), which shed light on the importance of systems thinking and the rise of a new public governance paradigm, scholars, and practitioners started giving more attention to collaborative methods aimed at model building and value cocreation processes. Indeed, systems thinking serves as a theoretical framework for guiding human actions and mental models to understand the big picture around specific issues and avoid concentrating on direct, short-term, and linear causal relationships (Meadows, 1989;Ricciardi et al., 2020). Collaborative approaches and group model building came to the fore as methods to involve the relevant stakeholders of determined processes in framing the feedback loops characterizing the complex and interconnected systems under analysis in which they are embedded (Rouwette et al., 2002). Thus, involving such actors in a prior exploration of a model’s result or building a causal loop or stock and flow diagram would help untangle the complexity of specific systems and raise SD models to full potential, hence taking care of their different interests and logics, which often compete with each other (Kopainsky et al., 2014;Forliano et al., 2020). In this vein, by leveraging institutional theories, adaptive comanagement, and the body of knowledge on the (new) commons, Ricciardi et al. (2020) recently proposed a conceptual causal loop diagram aimed at offering a participatory SD modeling method to overcome the fragilities raised when common resources are at stake. Therefore, involving managers, decision-makers, practitioners, or even citizens through participatory techniques and group model building could represent a critical step in achieving shared consensus behind SD models and different stakeholders’ understanding of the system of interest, effectively guiding behavioral change processes. The insights from this cluster suggest many ways to advance SD research in management and organization studies. SD must be used, for instance, to include/mitigate human biases in complex decisionmaking processes, to explore the dynamics of sense-making processes in terms of stakeholders impacted, and to shed light on the interplay between individuals, organizations, and communities in value cocreation processes. Dynamic performance management As with the other thematic clusters, performance management has also been a topic that has characterized SD studies since the origins of this research field. To overcome the difficulties related to applying SD principles by solving differential equations and using spreadsheets, the development of simulation software and computer-aided modeling is a critical step (Richmond, 1994). Indeed, the possibility of graphically representing system archetypes, causal loop diagrams, and stock and flow models is fundamental to disclosing SD to a broader public than specialists and mathematics (Wolstenholme, 2003). Following this idea, scholars started building “management flight simulators”for applying SD methods to business management in the 1980s (Forrester, 2007;Sterman, 2014). Through user-friendly dashboards and key performance indicators, these tools provide inexperienced users with an interactive learning environment that can be used to design and test different policies, evaluate diverse scenarios, and increase their acceptance of complex SD models (Bianchi & Bivona, 2000;Davidsen, 2000;Gr€ oßler et al., 2000). Conversely, in other cases, it was found that participants’performances can be leveraged by involving them in a prior exploration of the model, even if it is not in its final form (Kopainsky et al., 2014). Whereas its underlying structure characterizes the behavior of a system, organizational performance results from that behavior. Thus, understanding and communicating how that behavior is related to a system’sprocesses and activities represent critical steps for ensuring participants’performance (Schoenberg et al., 2020). C. Forliano, P. De Bernardi, Z. Rozsa et al. Journal of Innovation & Knowledge 9 (2024) 100512 8 Woodside, A. G. (2013). Moving beyond multiple regression analysis to algorithms: Calling for adoption of a paradigm shift from symmetric to asymmetric thinking in data analysis and crafting theory. Journal of Business Research, 66(4), 463–472. Xavier, J. A., & Bianchi, C. (2020). An outcome-based dynamic performance management approach to collaborative governance in crime control: Insights from Malaysia. Journal of Management and Governance, 24(4), 1089–1114. Yun,J.J.,Won,D.,Jeong,E.,Park,K.,Yang,J.,&Park,J.(2016).Therelationship between technology, business model, and market in autonomous car and intelligent robot industries. Technological Forecasting and Social Change, 103, 142–155. Zheng, L., & Iatridis, K. (2022). Friends or foes? A systematic literature review and meta-analysis of the relationship between eco-innovation and firm performance. Business Strategy and the Environment, 31(4), 1838–1855. C. Forliano, P. De Bernardi, Z. Rozsa et al. Journal of Innovation & Knowledge 9 (2024) 100512 15