Abstract
Introduction:
The inclusion of a system dynamics course in our medical school curriculum was designed to encourage systems thinking through computational modeling. From anecdotal observations, it soon became evident that something more profound was occurring—rather than simply learning, our students appeared to be constructing knowledge by building computational models in a way that is consistent with Papert's constructionism.
Approach:
In the absence of a reliable tool to identify constructionism, we examine the seminal literature of Forrester's system dynamics and Papert's constructionism by extracting key excerpts to look for evidence supporting the hypothesis that computational modeling may constitute a constructionist activity.
Observations:
The literature suggests that there is substantial convergence between the educational approach of constructionism and the activity of constructing models in system dynamics.
Discussion:
An examination of the seminal literature suggests that system dynamics modeling has features that are consistent with a constructionist approach. By extension, other approaches such as agent-based modeling also embody constructionist principles, and the expanding integration of artificial intelligence into computational modeling may present opportunities for novel approaches to constructionist learning. Formal real-world educational studies will be required to accumulate empirical learner data in order to confirm the constructionist nature of systems modeling.
1 Introduction
The impetus for this paper arose from our ongoing experience with a system dynamics (SD) course that we introduced into our medical curriculum with the objective of empowering our students with the advantages of a systems approach to complexity (Rubin et al., 2012; Rubin, 2018). As our experience in running this course grew, we began to recognize that the computer-based model-building appeared to provide our students with much more than the benefits of systems thinking. From these anecdotal observations, it became evident that through modeling, our students seemed to be engaging in a type of learning that is consistent with Papertian constructionism. This is distinct from social constructionism which will not be discussed in this paper.
Existing evidence supports the educational merits of our system dynamics course (Rubin et al., 2019), and we are continuing to advance this work through systematic evaluation and refinement. However, as a consequence of the anecdotal observations of apparent constructionist features in computational modeling, and in the absence of empirical learner data, we formulated the hypothesis that computational modeling may offer an opportunity to exploit constructionism. Thus, the focus of this paper is not to investigate the utility of a systems education per se in a medical curriculum. Rather, our purpose here is to answer the question: can computational model-building in SD constitute Papertian constructionism? To the extent that it does, it may present an opportunity to include constructionism in a medical curriculum, with all the associated benefits of this influential educational approach extending far beyond those that we originally envisaged.
Papertian constructionism, which we will simply refer to as constructionism in the remainder of this paper, can be loosely defined as “learning by making” although its architect, Seymour Papert, found this characterization to be overly simplistic and lacking in nuance and richness (Harel and Papert, 1991). It is perhaps not surprising that, despite its prominence as a theory of knowing, constructionism, at least of the Papertian flavor, is largely unexplored in the medical education literature. After all, what could a medical student possibly make in the process of learning?
Various authors have made limited, and in some cases, oblique, mention of Papert and constructionism in medical education (Qayumi and Qayumi, 1999; Hedegaard et al., 2007; Keskitalo and Ruokamo, 2021; Hasko et al., 2023). A number of authors have proposed using Lego bricks in medical education and suggest a constructionist element to this activity (Kirby and Pawlikowska, 2019; Capogna et al., 2022).
Ellaway has proposed “making” rather than simply “acquiring” as a potentially valuable aspect of medical education, and points to Papert's idea that computer programming is an important “making” activity (Ellaway, 2011). Harris, on the other hand, argues for the creation of physical artifacts using readily available materials, as a constructionist activity among medical students (Harris et al., 2025). Halan et al. suggest that the creation of a conversational virtual agent serves as a constructionist activity for medical students (Halan et al., 2014). Lavine argues in the context of making videotaped case presentations in neurobiology, that Papertian constructionism is best augmented by instructor guidance (Lavine, 2005).
The broad notion of computational modeling as a constructionist paradigm has been proposed. Gero and Levin (2019) suggest that modeling physical and engineering systems as difference equations in spreadsheets constitutes a constructionist approach to education. Other authors have recognized the constructionist aspects of SD specifically. For example, Gould-Kreutzer (1993) argues that the model had become an “object to think with,” in keeping with Papert's notion of a transitional object (Papert, 1980), while Zuckerman and Resnick (2003), point out the influence of epistemological theorists including Papert, in their development of “system blocks.”
In order to situate this paper, it is useful to say a few words about our SD course. Of the available systems modeling approaches, we chose SD because of its conceptual simplicity and ease-of-use for students who do not have a mathematical or computational background. This facilitates our medical students' engagement in rigorous systems concepts using contextually relevant examples from medicine and public health. In this course, we include a wide range of ever-changing modeling examples, including but not limited to physiological systems (Rubin et al., 2019), pharmacokinetics, epidemics (Rubin et al., 2021), health economics, enzyme kinetics, bioreactors, and dialysis (Rubin et al., 2024).
We continue to study the impact of the SD course on medical students. A survey indicates broad recognition among our students of the benefits of system thinking that flow from their engagement with this course (Rubin et al., 2019). In addition to formal feedback, we regularly receive informal reflections from students during their clinical rotations, attesting to a growing appreciation of their systems training – a phenomenon we intend to study further. Several well-known benefits of the study of SD that we have observed among our students include:
They develop a sense of interconnectedness and dynamic complexity, with a means to explore this on their own (Forrester, 1994). A survey of our students suggests that they are acquiring a sense of integration and systems thinking (Rubin et al., 2019).
They overcome so-called stock-flow failure, a common cognitive misunderstanding which is so important in medical practice, from fluid management (Brunstein et al., 2010) to understanding obesity (Abdel-Hamid et al., 2014).
They develop an appreciation for the critical importance of feedback, and an intuition for how interventions in complex systems may play out with the ever-present risk of unintended consequences (Forrester, 1994).
Perhaps the most remarkable aspect of SD is its integrating role in learning (Forrester, 1994). For example, similar model structures can represent diverse phenomena such as glucose homeostasis, T-cell response to cancer, and the availability of doctors in a healthcare system.
1.1 A brief review of Papertian constructionism
Notwithstanding his accomplishments as a computer scientist, mathematician and artificial intelligence pioneer, Seymour Papert's parallel career as a globally influential educationalist, remains his most enduring legacy. Constructionism evolved from constructivism as pioneered by Papert's mentor, Jean Piaget (Ackermann, 2001; Stager, 2016).
Papert's constructionism, and its intellectual parent, constructivism, are neither mutually exclusive nor contradictory, and both theories eschew instructionism, or transmission of knowledge (Ackermann, 2001).
While Piaget's constructivism emphasizes the creation and maintenance of mental models and internal structure, constructionism extends these notions to the critical importance of creating a (physical or virtual) public artefact in the process of making meaning and learning—a social activity (Ackermann, 2001). In as much as constructionism constitutes an educational theory, its concreteness as a learning process also makes it a learning methodology. Papert's constructionism is more situational, i.e., contextual, than Piaget's constructivism, and emphasizes the importance of engagement by immersion in the unknown, rather than by remote, stepwise contemplation (Ackermann, 2001). The remarkable ascent of humanity, from early toolmaker to contemporary Homo sapiens (Bronowski, 1973), is essentially a story of “making” within a social context, suggesting to us that our success as a species is deeply anchored in a primal constructionist impulse.
Notwithstanding its prominence as a pedagogical approach, reference to constructionism in the medical education literature is sparse and often oblique. In describing constructivism as the dominant paradigm in the learning sciences, Michael highlights the importance of students actively constructing meaning through engagement and reflection, particularly in a social context (Michael, 2006)—an argument that seems to align with the principles of constructionism, although the term itself is not explicitly mentioned.
1.2 A brief review of system dynamics and a related modeling framework, agent-based modeling
SD is a highly intuitive, visual approach to the world of complex systems. The field has democratized access to the study of systems, because a background in mathematics and computers is not a prerequisite. Its architect, Jay Forrester, recognized the ubiquity of systems in areas as diverse as urban development (Forrester, 1969), commerce, management (Forrester, 1999), economics and social interactions, as an extension of his pioneering work in computers and servomechanisms.
The essence of SD is the simple idea that an accumulation of a quantity over time, known as a level (or stock), is dependent on the rate (or flow) of that quantity, entering and leaving the level (Ford, 2010). A common analogy is the accumulation of water in a bathtub (level) due to the flows (rates) into and out of the tub. The water serves as an analogy for any number of accumulations such as infected people in an epidemic, available doctors in the country, capital available to a medical insurance company, and glucose concentration. Multiple level-rate pairs can be coupled together in the computer work-space to represent more complex systems. Simulation of the model produces graphical output representing the behavior of the variables.
Among the numerous modeling examples in our SD course for medical students, the behavior of epidemics features prominently (Rubin et al., 2021). Figure 1 shows a SD model of a simple Susceptible-Infectious-Recovered (SIR) epidemic. The practitioner may choose to modify the model, e.g., by including an exposed but not yet infected stage, directing a flow from Recovered back to Susceptible to represent loss of immunity, change the probability of getting infected or the rate of recovery, or experiment with the starting fraction of immune individuals to explore herd immunity.
Figure 1
Students learn concepts from their models through experimentation. For example, they discover that incidence rate and point prevalence are simply the (normalized) infection rate and the level of infected individuals (I) in the model respectively.
No discussion on systems modeling in the context of constructionism would be complete without mentioning agent-based modeling (ABM) as another systems modeling paradigm that has an immense influence on learning (Blikstein et al., 2005). SD is a macro, aggregated approach to modeling where the accumulated quantities are collected in a level (stock), whereas ABM is a micro approach (Achachlouei and Hilty, 2015; Cassidy et al., 2019). SD and ABM should not be viewed as competitive modeling approaches. Rather, they are frequently complementary, with each offering a different perspective on the same phenomenon. In ABM, the emphasis is on the individual agents giving rise to emergent behavior, whereas SD emphasizes the aggregated continuous-time (albeit discretely simulated) behavior in the form of coupled first-order differential equations. Both SD and ABM can model highly non-linear behavior.
In ABM, we specify a (usually) large number of “agents” and assign to them a set of probabilistic properties, rules and attributes. Similarly, we specify attributes of a set of small regions making up the background on which the agents “live.”
Once the simulation begins, agents move according to their pre-assigned behavioral probabilities. When they interact with each other or with the background, various transformations happen based on assigned rules and probabilities. Agents could represent any of a wide range of entities such as gas molecules, people, or cells. For example, we could model the social interactions of a crowd or growth of cells in a cancer.
The stochastic aspect inherent in this type of modeling is particularly useful when modeling heterogeneous populations in biological, physical and social processes. It is inspiring to watch how models built on a set of simple probabilities and rules, may evolve astounding behavior patterns, so-called emergent behavior, that one would not predict a priori. Small changes to the underlying attributes may result in amazing behaviors which can be a source of intriguing, exploratory fun.
Models that deal with continuous variables, are often best constructed using SD, e.g., glucose homeostasis. Others, particularly those where the variables are discrete such as cells or people, are well suited to ABM. In many situations, it is instructive to explore models using both techniques.
The NetLogo programming language (Blikstein et al., 2005; Wilensky, 1999) was used to develop an agent-based version of the SIR model shown in Figure 2. The 200 people (agents) in this example, mill around with individually unpredictable but probabilistically defined movements. When a susceptible individual (green) encounters an infected individual (red), they have a probability of becoming infected and turning red. Infected individuals recover (turn blue) with a statistically defined probability, and in this model, they remain immune.
Figure 2
The curves in Figure 2 for S, I and R, are not as smooth as those generated by the SD model, and their shape varies each time the model is run. This is a consequence of the stochastic nature of ABM.
There is an interesting connection between ABM and constructionism. Uri Wilensky, inspired by Papert's epistemology and Resnick's educational computer language, StarLogo (Zuckerman and Resnick, 2003), developed the NetLogo computer language as both an educational and professional tool for ABM. Like Resnick, Wilensky remained faithful to the low-threshold, high-ceiling philosophy of the programming language, Logo, developed by Papert, Feurzeig and Solomon, retaining many of its features (Blikstein et al., 2005). “Low-threshold” refers to ease of access by novices, while “high-ceiling” refers to its utility as an advanced professional research tool. Resnick's long-standing association with Papertian constructionism (Resnick, 1997), and his early work on StarLogo for agent-based modeling, lends further support to the notion of systems modeling as a constructionist activity (Wilensky and Resnick, 1999).
In an extensive review of modeling in the healthcare sector, Cassidy et al. (2019) demonstrate the wide range of domains in which both SD and ABM modeling are applied, particularly in acute, long-term, emergency and elderly care services. Both SD and ABM are frequently applied to understanding and optimizing activities and policies in areas as diverse as management of hospital waste, resource constraints, patient outcomes, and medical insurance (Cassidy et al., 2019). They also point out that, while still limited, there is likely to be a growth in hybrid ABM/SD modeling approaches in healthcare.
1.3 Artificial intelligence in computational modeling
As we become accustomed to operating within the third digital epoch dominated by AI (Levin et al., 2025), the question of how we can and should use AI for computational modeling in an educational context becomes pertinent. Arguably the most obvious use of AI in modeling is for parameter estimation.
SD can be used in two ways. The first is to achieve a broad understanding of model behavior, and the second is for behavior prediction (Ford, 2010). In our SD course for medical students, we focus exclusively on the development of understanding rather than prediction, and as such, we do not require our students to perform model calibration (Rubin, 2018). However, we envisage a time in the near future when the introduction of model calibration becomes feasible for our cohort of students by using AI.
Houghton and Siegel developed the Python-based PySD for the integration of SD modeling and parameter estimation using AI (Houghton and Siegel, 2015). Gadewadikar and Marshall demonstrated the use of PySD to develop parameter estimates for an SIR epidemic model of Covid-19 which included deaths (Gadewadikar and Marshall, 2024). Their model was initially developed in Insight Maker (Fortmann-Roe, 2014) and imported into PySD together with real-world data to achieve calibration.
Ye et al. describe the growing integration of mechanistic epidemiology modeling with AI despite it being somewhat fragmented at this stage (Ye et al., 2025). They highlight areas requiring further research and point to the promise of enhanced public health decision-making. Hu developed chatPySD in an effort to seamlessly integrate SD modeling with a large language model (LLM), not only for calibration, but also for two-way interaction with generative AI to achieve model refinement and suggested model architectures (Hu, 2025). This approach allows for a high degree of interactivity between the modeler and the LLM.
It is evident that we are witnessing the early phases of a rapid expansion of AI in modeling. As AI inevitably becomes an integral aspect of computational modeling, we will need to decide on how to use this to enhance the educational process.
2 Approach
Both Forrester's system dynamics and Papert's constructionism are highly specific knowledge frameworks, that are largely defined by the foundational literature of these two authors. While constructionism, is a theory of knowledge creation, system dynamics is a methodology. However, system dynamics presents an opportunity for knowledge creation through the building of models. In this sense, it begins to look very much like a constructionist activity.
As influential as both Papert and Forrester are, they published most of their critical work in a small number of seminal books and papers accounting for the overwhelming preponderance of their respective citations.
In this study, we examine a sub-set of the principal works that define SD and constructionism written by Forrester (Forrester, 1992, 1994, 1996), Papert (Harel and Papert, 1991; Papert, 1993, 1980), and two authors who worked directly with Papert in the development of constructionism, namely Harel and Papert (1991) and Stager (2005).
We conducted a structured comparative thematic analysis of the two bodies of literature described above, namely, constructionism, and SD. Drawing heavily on Papert's “eight big ideas” (Stager, 2005), we began by inductively identifying recurrent themes within constructionism, followed by an examination of the SD literature through the lens of these themes, with particular emphasis on convergence. While the analysis was implemented systematically, we did not apply a formal coding methodology—as such, the approach was interpretive.
In order to facilitate the comparison, we extracted salient excerpts from the constructionist literature that define the recurrent themes. For later analysis, each excerpt from the constructionist literature is assigned the code C followed by a numeral, e.g., the second excerpt would be designated C2. Representative excerpts from the SD literature demonstrating convergence were then extracted and assigned codes starting with SD followed by a numeral.
A domain intersection map was constructed by listing the themes identified in the constructionist literature as shown in Figure 3. Selected quotations from the constructionist literature that supported the themes were then connected to the themes via their codes. Selected coded quotations from the SD literature were then also connected to the themes to graphically illustrate the alignment between constructionism and SD.
Figure 3
3 Observations
The defining feature of constructionism is the requirement to “make something,” a so-called artefact, as part of the educational process. The very act of making is, according to Papert, integral to the process of knowledge creation. The product may be physical, virtual (a computer program), or as is so typical of Papert, a hybrid between the computational and the physical. It could be a work of art, a game, a piece of music or a movie.
However, Papert, more than most theorists, views the educational process as an obligate social activity. He requires the artefact to be public in the sense of being a shared focal point of discussion (Stager, 2016), which can be explained by the maker and discussed with their peers. The artefact's workings or an appreciation of its aesthetics should not reside solely in the mind of its maker. As will become evident, the attributes of constructionism that we identify from the literature constitute a subset of Papert's “eight big ideas.”
While both constructionism and constructivism involve the building of knowledge structures, constructionism extends constructivism thus (Harel and Papert, 1991, p. 1):
It then adds the idea that this happens especially felicitously in a context where the learner is consciously engaged in constructing a public entity, whether it's a sand castle on the beach or a theory of the universe. C1
We can thus state the first principle of constructionism as learning by publicly making an artefact.
Papert recognizes, indeed insists, on the immense disruptive power of the digital computer to facilitate learning in ways that were previously unattainable. While computers are not the only means to effect a constructionist educational paradigm, they are inextricably coupled with Papert's constructionism (Harel and Papert, 1991; Papert, 1980). As such, a dominant theme in much of his writings deals with artefacts that are created with computers. For Papert, the digital computer is a profoundly liberating tool. As one of the developers of the educational programming language, Logo, Papert envisioned unleashing the student's creative impulse by placing in their hands, the unbridled power of the digital computer (Papert, 1980).
Most students use computers for writing (word processors), communications (email, presentations and video conferencing), access to information (web browsing) and numerical analysis (spreadsheets). For Papert the real value of computers lies in the immense exploratory power that they provide through model building and recursive thinking, thus massively extending our capacity for creative musings and exploration. Indeed, Levin et al. (2025) examined the centrality of digital computing in the evolution of constructionism across three key digital epochs, culminating in the use of artificial intelligence (AI). A number of excerpts from Papert illustrate his thinking (Papert, 1980, p. 18, viii, and 4):
But I want my readers to be very clear that what is “Utopian” in my vision and in this book is a particular way of using computers, of forging new relationships between computers and people... C2
The computer is the proteus of machines. Its essence is its universality, its power to simulate. Because it can take on a thousand forms and can serve a thousand functions, it can appeal to a thousand tastes. C3
In this book I discuss ways in which the computer presence could contribute to mental processes not only instrumentally but in more essential, conceptual ways, influencing how people think even when they are far removed from physical contact with a computer. . . C4
While the inclusion of digital computers is not an absolute requirement, it is so integral to Papert's thinking that we can reasonably state a second principle of constructionism, namely, it will typically exploit the power of the digital computer.
Papert argues that constructionism should, as far as possible, be fun. He points out that fun does not imply easy. On the contrary, he argues that enjoyment can be derived from hard work in an engaging but challenging task. This theme of fun is not well explored in the medical education literature. While the idea that education can be fun is not controversial, Papert sees this in the way a child becomes intensely engrossed in building something. He is very much concerned with the affective aspect of the process of learning—how it feels. In Mindstorms, he describes how his love for gears as a very young child was the catalyst for his further intellectual development. He suggests that the universality of computers can serve the role of his gears. Papert remarks (Papert, 1980, p. 13):
That all this would be fun needs no argument. But it is more than fun. Very powerful kinds of learning are taking place. C5
In their various writings, Papert and his collaborators relate anecdotes of children becoming deeply engrossed in their projects in ways that they were never able to achieve before (Stager, 2005). It is as if the student, using the computer in this way, develops an intense curiosity to explore the world through the power conferred on them by merging their creativity with the computer's capacity to recursively model their every whim. Reading Papert evokes a sense that the activity becomes fun in the same way as an explorer would experience a sense of wonder at the discovery of new worlds.
The third principle of constructionism can be stated as, the process should be absorbing and fun.
The critical importance of getting it wrong en route to getting it right is another of Papert's themes which is explicitly included in his eight essential points (Stager, 2005). While reflection, and feedback from facilitators, lecturers and peers is integral to the process of medical education, there is no clear emphasis that we could identify which literally venerates “getting it wrong.” From the constructionist viewpoint, getting it wrong is essential, indeed desirable. Many who engage in creative programming will recognize that identifying and solving the problem or bug can be intensely satisfying, even exhilarating. According to Papert (Stager, 2005, p. 4):
...you can't get it right without getting it wrong. Nothing important works the first time. C6
Thus we state the fourth constructionist principle: getting it wrong is desirable, even essential.
Similarly, computer simulation is a “safe” environment in which to test hypotheses and make assumptions in the process of learning. Learning by discovery trumps learning by being taught. For Papert, the students' sense of exploration and wonder are evident throughout his writings. This is tightly coupled with the importance of getting it wrong and having fun. Explorers take wrong turns but in doing so, they discover new vistas. In relation to an experimental constructionist program for at-risk learners, Stager puts it this way (Stager, 2005, p. 9):
These students engaged in a process of exploration not unlike the men who sailed the high seas or landed on the moon. C7
Perhaps Papert's taste for exploration and minimalism in teaching is best expressed by his qualified endorsement of the idea that (Papert, 1993, p. 139):
...every act of teaching deprives the child [medical student] of an opportunity for discovery.... C8
We can thus add a fifth and final principle of Papertian constructionism, that the learning process is an exploration.
We now turn our attention to Forrester's SD and compare it to constructionism.
The structure of models built in SD depends on the modeler's evolving understanding of the phenomenon being modeled. For example, a model of glucose homeostasis may account for no more than the input of glucose, the proportional production of insulin in response to the glucose level, and the insulin-dependent uptake of glucose by cells. In a second iteration, insulin-independent glucose uptake could be included. A more sophisticated model may include non-linear responses of insulin to glucose and glucose uptake to insulin, with further iterations including glucagon and other hormones. Model-building is a participatory, creative process and there is no single correct model—only a quest for the simplest model that is sufficiently explanatory.
The creation of a model is in every sense, the creation of an artefact and this is indeed Papert's primary constructionist principle. According to (Forrester, 1994, p. 20-21):
Coming to an understanding of systems must be a participative activity. Learning about systems is not a spectator sport, such learning comes from active involvement. SD1
Referring to his Urban Dynamics project (Forrester, 1969), Forrester illustrates the importance of public engagement with Boston residents in the modeling process (Forrester, 1994). He then describes the public engagement of his graduate student in modeling diabetes (Forrester, 1994, p. 10):
He immediately developed a working-colleague relationship with doctors in Boston's research clinic for diabetes because for the first time they were able to put together their fragments of medical knowledge into a meaningful system. SD2
As evidence of the public nature of SD, Forrester quotes a high-school teacher engaging a class (Forrester, 1996, p. 15):
Graphs in hand, the students were arguing positions before I could take attendance, peering over books and tables, pointing out misjudgments and omissions. SD3
Other researchers in SD refer to the importance of stakeholder participation in SD modeling (Freebairn et al., 2018). A well-known SD software tool, InsightMaker (Fortmann-Roe, 2014), is entirely web-based and has a facility for model sharing. ABM also benefits from sharing (Blikstein et al., 2005; Wilensky, 1999).
The computer as a virtual exploratory laboratory is what has made SD feasible (Forrester, 1992). Indeed, the affordability and accessibility of such computing power has allowed us to introduce a SD course into our medical curriculum (Rubin et al., 2012; Rubin, 2018). As one of the key elements in making SD feasible, Forrester points out one of the three foundations on which system dynamics rests (Forrester, 1992, p. 8) namely:
Digital computers, now primarily personal computers, to simulate the behavior of systems that are too complex to attack with conventional mathematics, verbal descriptions, or graphical methods. High school students, using today's computers, can deal with concepts and dynamic behavior that only a few years ago were restricted to work in advanced research laboratories. SD4
Both Forrester and Papert share the view that the democratization of knowledge construction through the creative use of computer-based model building is a foundational concept in contemporary education.
Forrester and his colleagues express the same excitement as Papert in observing the use of SD as an educational tool. He quotes Professor Michael Radzicki who worked with high-school students on SD (Forrester, 1996, p. 27):
Students were VERY engaged in the activities. They refused to take breaks; they preferred to keep working on the problems or to try building their own model. SD5
The fun and curiosity sparked by the hands-on modeling in SD is clearly consistent with Papert's vision of constructionism.
Forrester agrees with Papert about the benefit of students being surprised at the mistakes they discover in their own modeling as part of the learning process (Forrester, 1994, p. 21).
...system dynamics modeling is learning by doing. It is learning through being surprised by the mistakes one makes. System dynamics modeling is a participative activity in which one learns by trial and error and practice. I believe that immersion in such active learning can change mental models. SD6
If one were to simply replace the phrase “system dynamics modeling,” with “constructionism” in the above excerpt, this could easily be Papert speaking rather than Forrester. Forrester points out that not all the surprising outcomes represent mistakes; some represent unanticipated behaviors that lead to new insights.
Forrester shares Papert's view about the exploratory aspect of the process (Forrester, 1994, p. 13):
Computer simulation modeling is a repeating process of trial and error. One learns that progress is made through exploration and by learning from mistakes. SD7
Once again, these words could have been Papert's, yet they are Forrester's. Constructionism and SD inherently imbue students with an exploratory impulse.
Based on the extracted excerpts, we construct a domain intersection map shown in Figure 3, illustrating the alignments between SD and constructionism. From the constructionist literature we identify five principles. Linkages of the SD quotations to these five principles are shown in Figure 3.
4 Discussion
While Forrester clearly appreciated the value of SD as a foundational educational instrument and not simply a tool for professional modeling, many other authors have also reported on the power of SD to achieve deep understanding among students. For example, Fisher (2018) describes the unequivocal power of SD to enhance deep understanding among students who may not yet have attained advanced skills usually associated with mathematical modeling—this closely mirrors our own experience.
Through our engagement with the medical SD course, we recognize that, in addition to the benefits of a systems education, the very act of building models is itself a powerful educational exercise which appears to be consistent with Papert's constructionism. The purpose of this study is to gain insight into the extent to which this is so. A direct comparison of the foundational literature of constructionism and system dynamics suggests that the activity of developing computer-based models in SD, aligns with a constructionist educational approach.
Papert's parallel educational career culminated in the highly influential constructionism. Forrester, on the other hand, is not known as an educational theorist. He could be described as an accidental educationalist in that he recognized that SD, which he intended primarily as a professional systems methodology, has immense educational potential for children and adults.
In addition to SD, ABM stands out as another candidate for this constructionist role. This is not accidental. The most widely used educational software for ABM, NetLogo, is a direct legacy of Papert's work on Logo. We are considering the introduction of ABM together with SD as there is value in students exploring systems from both perspectives.
5 Conclusions
Constructionism remains largely unexplored in the medical education literature despite its prominence as a pedagogical theory. This may be due in part to the difficulty in imagining what medical students could make in a social context that would allow them to create knowledge. Our SD course appears to exhibit many of the hallmarks of a constructionist framework, giving our students an opportunity to learn “by making,” and in so doing, to potentially exploit the immense power of constructionist learning.
The free-ranging, adventurous notion of constructionist learning is at odds with more conventional educational models. While there are clear objectives to be met in any educational regime, including the training of future physicians, the process of getting there could be far more efficient and stimulating.
Through an analysis of the seminal literature, computational modeling emerges as a contender for a constructionist educational approach, potentially presenting an opportunity to include Papertian constructionism into medical education. While SD offers a transformative shift in how future physicians engage with complex systems through active, exploratory learning, the untapped potential for AI to expand the constructionist features of computational modeling may be the next frontier.
We argue that the constructionist and SD literature suggests that the process of computational model-building and simulation is consistent with a constructionist activity, presenting untapped opportunities for medical education. The acquisition of empirical learner data through formal educational research is required to determine the extent to which constructionist features are present in system dynamics modeling in an educational context, including in medical curricula.
Statements
Data availability statement
Publicly available data was analyzed and can be found in the cited articles.
Ethics statement
This study received an ethics waiver (W25/01/05) from our institutional review board, HREC(Medical).
Author contributions
DR: Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Software, Visualization, Writing – original draft, Writing – review & editing. PK: Investigation, Methodology, Writing – original draft, Writing – review & editing. XR: Investigation, Methodology, Writing – original draft, Writing – review & editing. AG: Investigation, Methodology, Writing – original draft, Writing – review & editing.
Funding
The author(s) declared that financial support was not received for this work and/or its publication.
Acknowledgments
We thank Adriano Giovanelli for help with an earlier draft of Figure 3.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was used in the creation of this manuscript to find synonyms.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
References
1
Abdel-HamidT.AnkelF.Battle-FisherM.GibsonB.Gonzalez-ParraG.JalaliM.et al. (2014). Public and health professionals' misconceptions about the dynamics of body weight gain/loss. Syst. Dyn. Rev. 30, 58–74. doi: 10.1002/sdr.1517
2
AchachloueiM. A.HiltyL. M. (2015). System dynamics vs. agent-based modeling-comparing models and approaches: a literature review and a transformation procedure. Preprint. Available online at: https://www.diva-portal.org/smash/get/diva2:844069/fulltext01.pdf
3
AckermannE. (2001). Piaget's constructivism, Papert's constructionism: what's the difference. Fut. Learn. Group Public. 5:438.
4
BliksteinP.AbrahamsonD.WilenskyU. (2005). “Netlogo: where we are, where we're going,” in Proceedings of the annual meeting of Interaction Design and Children, vol. 23 (Northwestern University; The Center for Connected Learning and Computer-Based Modeling).
5
BronowskiJ. (1973). The Ascent of Man. British Broadcasting Corporation.
6
BrunsteinA.GonzalezC.KanterS. (2010). Effects of domain experience in the stock–flow failure. Syst. Dyn. Rev. 26, 347–354. doi: 10.1002/sdr.448
7
CapognaG.IngrassiaP. L.CapognaE.BernardiniM. (2022). The use of lego®bricks to train novice debriefers. Int. J. Healthc. Simul. 1, 42–43. doi: 10.54531/kupp8710
8
CassidyR.SinghN. S.SchirattiP. R.SemwangaA.BinyarukaP.SachingonguN.et al. (2019). Mathematical modelling for health systems research: a systematic review of system dynamics and agent-based models. BMC Health Serv. Res. 19, 1–24. doi: 10.1186/s12913-019-4627-7
9
EllawayR. (2011). eMedical teacher. Med. Teach. 33, 865–867. doi: 10.3109/0142159X.2011.618071
10
FisherD. M. (2018). Reflections on teaching system dynamics modeling to secondary school students for over 20 years. Systems6:12. doi: 10.3390/systems6020012
11
FordA. (2010). Modeling the Environment, 2nd Edn. Washington, DC: Island Press.
12
ForresterJ. (1992). System Dynamics and Learner Centered Learning in Kindergarten Through 12th Grade Education. Technical report. MIT OCW.
13
ForresterJ. W. (1969). Urban Dynamics. Waltham, MA: Pegasus Communications.
14
ForresterJ. W. (1994). Learning Through System Dynamics as Preparation for the 21st Century. Technical report. Keynote address for systems thinking and dynamic modeling conference for K-12 Education. Revised 2009.
15
ForresterJ. W. (1996). System Dynamics and k-12 Teachers. Technical Report. MIT OCW.
16
ForresterJ. W. (1999). Industrial Dynamics. Waltham: Pegasus Communications, Inc. Original work published 1961.
17
Fortmann-RoeS. (2014). Insight maker: a general-purpose tool for web-based modeling and simulation. Simul. Modell. Pract. Theory47, 28–45. doi: 10.1016/j.simpat.2014.03.013
18
FreebairnL.AtkinsonJ. A.KellyP. M.McDonnellG.RychetnikL. (2018). Decision makers' experience of participatory dynamic simulation modelling: methods for public health policy. BMC Med. Inform. Decis. Mak. 18:131. doi: 10.1186/s12911-018-0707-6
19
GadewadikarJ.MarshallJ. (2024). A methodology for parameter estimation in system dynamics models using artificial intelligence. Syst. Eng. 27, 253–266. doi: 10.1002/sys.21718
20
GeroA.LevinI. (2019). Computational thinking and constructionism: creating difference equations in spreadsheets. Int. J. Math. Educ. Sci. Technol. 50, 779–787. doi: 10.1080/0020739X.2018.1501827
21
Gould-KreutzerJ. M. (1993). Foreword: system dynamics in education. Syst. Dyn. Rev. 9, 101–112. doi: 10.1002/sdr.4260090202
22
HalanS.LokB.SiaI.CraryM. (2014). “Virtual agent constructionism: experiences from health professions students creating virtual conversational agent representations of patients,” in 2014 IEEE 14th International Conference on Advanced Learning Technologies (Athens: IEEE), 249–253. doi: 10.1109/ICALT.2014.79
23
HarelI.PapertS. (1991). “Situating constructionism,” in Constructionism, eds. I. Harel and S. Papert (Norwood, NJ: Ablex Publishing), 1–10.
24
HarrisA.BardelliM.BrancaleoneG.CostaN.HrubyL.PoeliejoeR. (2025). Making as method in teaching: do-it-yourself (diy) objects and hands-on learning with materials. Perspect. Med. Educ. 14:309. doi: 10.5334/pme.1575
25
HaskoR.HaskoO.HakanK. (2023). “Internet of robotic things (IoRT) approach to lifelong learning and medical education with internet of medical things (IoMT),” in Workshop Proceedings of 7th International Conference on Informatics and Data-Driven Medicine (Birmingham), 287–294. Available online at: https://ceur-ws.org/Vol-3892/short5.pdf
26
HedegaardH.DahlM. R.GrønbækK. (2007). “EKGAR: Augmented reality support for learning of electrocardiography (ECG) analysis,” in EdMedia+ Innovate Learning (Association for the Advancement of Computing in Education (AACE)), 3544–3549.
27
HoughtonJ.SiegelM. (2015). “Advanced data analytics for system dynamics models using pysd,” in 33rd International Conference of the System Dynamics Society, vol. 2 (Cambridge, MA: System Dynamics Society).
28
HuB. (2025). ChatPySD: embedding and simulating system dynamics models in chatgpt-4. Syst. Dyn. Rev. 41:e1797. doi: 10.1002/sdr.1797
29
KeskitaloT.RuokamoH. (2021). Exploring learners' emotions and emotional profiles in simulation-based medical education. Austral. J. Educ. Technol. 37, 15–26. doi: 10.14742/ajet.5761
30
KirbyB. P.PawlikowskaT. (2019). Pharmacology through play: using Lego® to revise core concepts for undergraduates. MedEdPublish8:201. doi: 10.15694/mep.2019.000201.1
31
LavineR. A. (2005). Guided discovery learning with videotaped case presentation in neurobiology. J. Med. Sci. Educ. 15, 4–7.
32
LevinI.SemenovA. L.GorskyM. (2025). Smart learning in the 21st century: advancing constructionism across three digital epochs. Educ. Sci. 15:45. doi: 10.3390/educsci15010045
33
MichaelJ. (2006). Where's the evidence that active learning works?Adv. Physiol. Educ. 30, 159–167. doi: 10.1152/advan.00053.2006
34
PapertS. (1993). The Children's Machine: Rethinking School in the Age of the Computer. New York, NY: Basic Books.
35
PapertS. A. (1980). Mindstorms: Children, Computers, and Powerful Ideas. New York, NY: Basic Books.
36
QayumiA. K.QayumiT. (1999). Computer-assisted learning: cyberpatient™-a step in the future of surgical education. J. Investig. Surg. 12, 307–317. doi: 10.1080/089419399272296
37
ResnickM. (1997). Turtles, Termites, and Traffic Jams: Explorations in Massively Parallel Microworlds. Cambridge, MA: MIT Press.
38
RubinD. M. (2018). “Inclusion of system dynamics training in the education of healthcare workers in Africa,” in System Dynamics Models for Africa's Developmental Planning, eds. A. Brent and T. Simelane (Pretoria: Africa Institute of South Africa), 213–234.
39
RubinD. M.AchariS.CarlsonC. S.LettsR. F. R.PantanowitzA.PostemaM.et al. (2021). Facilitating understanding, modeling and simulation of infectious disease epidemics in the age of COVID-19. Front. Public Health9:593417. doi: 10.3389/fpubh.2021.593417
40
RubinD. M.LettsR. F.RichardsX. L.AchariS.PantanowitzA. (2024). Rapid prototyping of multi-compartment models for urea kinetics in hemodialysis: a system dynamics approach. J. Artif. Organs27, 316–320. doi: 10.1007/s10047-023-01416-w
41
RubinD. M.LettsR. F. R.RichardsX. L. (2019). Teaching physiology within a system dynamics framework. Adv. Physiol. Educ. 43, 435–440. doi: 10.1152/advan.00198.2018
42
RubinD. M.RichardsC. L.KeeneP. A.PaikerJ. E.GrayA. R.HerronR. F.et al. (2012). System dynamics in medical education: a tool for life. Adv. Health Sci. Educ. 17, 203–210. doi: 10.1007/s10459-010-9237-4
43
StagerG. (2005). “Papertian constructionism and the design of productive contexts for learning,” in Proc. of EuroLogo (Warsaw), the 10th European Logo Conference (Warsaw: EuroLogo), 43–53.
44
StagerG. S. (2016). Seymour Papert (1928-2016). Nature537, 308–308. doi: 10.1038/537308a
45
WilenskyU. (1999). Netlogo. Technical Report. Evanston, IL: Center for Connected Learning and Computer-based Modeling, Northwestern University.
46
WilenskyU.ResnickM. (1999). Thinking in levels: a dynamic systems approach to making sense of the world. J. Sci. Educ. Technol. 8, 3–19. doi: 10.1023/A:1009421303064
47
YeY.PandeyA.BawdenC.SumsuzzmanD. M.RajputR.ShoukatA.et al. (2025). Integrating artificial intelligence with mechanistic epidemiological modeling: a scoping review of opportunities and challenges. Nat. Commun. 16:581. doi: 10.1038/s41467-024-55461-x
48
ZuckermanO.ResnickM. (2003). “System blocks: a physical interface for system dynamics learning,” in Proceedings of the 21st International System Dynamics Conference, vol. 810811 (New York, NY: System Dynamics Society), 765891–766005.
Summary
Keywords
agent-based modeling, computational modeling, constructionism, medical education, system dynamics
Citation
Rubin DM, Keene PAC, Richards XL and George A (2026) Can computational modeling in medical education support a constructionist educational framework? Insights from the seminal literature in Papertian constructionism and system dynamics. Front. Educ. 11:1743544. doi: 10.3389/feduc.2026.1743544
Received
10 November 2025
Revised
19 February 2026
Accepted
25 February 2026
Published
13 April 2026
Volume
11 - 2026
Edited by
Ilya Levin, Holon Institute of Technology, HIT, Israel
Reviewed by
Forman Erwin Siagian, Christian University of Indonesia, Indonesia
Evgeny Patarakin, Moscow City University, Russia
Atsushi Yoshikawa, Kanto Gakuin University, Japan
Updates
Copyright
© 2026 Rubin, Keene, Richards and George.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: David M. Rubin, david.rubin@wits.ac.za
Disclaimer
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.