By Thorid Wagenblast1, Nicholas Roxburgh2 and Alessandro Taberna3
1 Delft University of Technology, 0009-0003-5324-3778
2 The James Hutton Institute, 0000-0002-7821-1831
3 CMCC Foundation – Euro-Mediterranean Center on Climate Change, RFF-CMCC European Institute on Economics and the Environment, 0000-0002-0207-4148
Introduction
Most agent-based models (ABMs) are designed around the assumption of a broadly stable system architecture. Whether exploring emergent dynamics or testing the effects of external interventions or stressors, such models typically operate with a fixed ontology – predefined agent types, attribute classes, behavioural repertoires, processes, and social and institutional structures. While this can allow rich exploration of dynamics within the given configuration, it limits the model’s possibility space by excluding forms of change that would require the structure itself to evolve.
Some of the most consequential forms of real-world change involve shifts in the system architecture itself. These forms of change – what we refer to here as deep structural change – reconfigure the underlying logic and potentialities of the system. This may involve, for example, dramatic shifts in the environment in which agents operate, the introduction of novel technologies, or reshaping of the roles and categories through which agents understand and act in the world. Such transformations pose a fundamentally different challenge from those typically addressed in most agent-based modelling studies to date – one that pushes beyond parameter tuning or rule adjustment, and calls for new approaches to ontology design, model construction, and the conceptualisation of structural transformation and uncertainty in simulation.
Various theoretical lenses can be applied to this topic. The concepts of transformations or regime shifts seem particularly pertinent. Transformations, in contrast to incremental or minor changes, are changes that are large-scale and significant, but apart from that do not seem to consist of any specific features (Feola, 2015). The changes we explore here are more closely linked to regime shifts, which are characterised by structural changes, but with a notion of abruptness. Methods to detect and understand these regime shifts and the structural changes in relation to social simulation have been discussed for some time (Filatova, Polhill & van Ewijk, 2016). Nonetheless, there is still a lack of understanding around what this structural change entails and how this applies in social simulation, particularly ABMs.
To explore these issues, the European Social Simulation Association (ESSA) Special Interest Group on Modelling Transformative Change (SIG-MTC) organised a dedicated session at the Social Simulation Fest 2025. The session aimed to elicit experiences, ideas, and emerging practices from the modelling community around how deep structural change is understood and approached in agent-based simulation. Participants brought perspectives from a wide range of modelling contexts – including opinion dynamics, energy systems, climate adaptation, food systems, and pandemic response – with a shared interest in representing deep structural change. A majority of participants (~65%) reported that they were already actively working on, or thinking about, aspects of deep structural change in their modelling practice.
The session was framed as an opportunity to move beyond static ontologies and explore how models might incorporate adaptive structures or generative mechanisms capable of capturing deep structural shifts. As described in the session abstract:
“We will discuss what concepts related to deep structural change we observe and how models can incorporate adaptive ontologies or generative mechanisms to capture deep structural shifts. Furthermore, we want to facilitate discussion on the challenges we face when trying to model these deep changes and what practices are currently used to overcome these.”
This article reflects on key insights from that session, offering a synthesis of participant definitions, identified challenges, and promising directions for advancing the modelling of deep structural change in agent-based social simulation.
Defining deep structural change
Participant perspectives
To explore how participants understood deep structural change and its characteristics, we used both a pre-workshop survey (N=20) and live group discussion activities (N ≈ 20; divided into six discussion groups). The survey asked participants to define “deep structural change” in the context of social systems or simulations, and to explain how it differs from incremental change. During the workshop, groups expanded on these ideas using a collaborative Miro board, where they responded to three prompts: “What is deep structural change?”, “How does it differ from incremental change?”, and they were asked to come up with a “Group definition”. The exercises benefited from the conceptual and disciplinary diversity of participants. Individuals approached the prompts from different angles – shaped by their academic backgrounds and modelling traditions – resulting in a rich and multifaceted view of what deep structural change can entail.
Across the different exercises, a number of common themes emerged. One of the most consistent themes was the idea that deep structural change involves a reconfiguration of the system’s architecture – a shift in its underlying mechanisms, causal relationships, feedback loops, or rules of operation. This perspective goes beyond adjusting parameters; it points to transformations in what the system is, echoing the emphasis in our introductory framing on changes to the system’s underlying logic and potentialities. Participants described this in terms such as “change in causal graph”, “drastic shift in mechanisms and rules”, and “altering the whole architecture”. Some also emphasised the outcomes of such reconfigurations – the emergence of a new order, new dominant feedbacks, or a different equilibrium. As one participant put it, deep structural change is “something that brings out new structure”; others described “profound, systemic shifts that radically reshape underlying structures, processes and relationships”.
Another frequently discussed theme was the role of social and behavioural change in structural transformation – particularly shifts in values, norms, and decision-making. Several groups suggested that changes in attitudes, awareness, or shared meanings could contribute to or signal deeper structural shifts. In some cases, these were framed as indicators of transformation; in others, as contributing factors or intended outcomes of deliberate change efforts. Examples included evolving diets, institutional reform, and shifts in collective priorities. Participants referred to “behavioural change coming from a change in values and/or norms” and “a fundamental shift in values and priorities”.
Furthermore, participants discussed how deep structural change differs from incremental change. They described deep structural change as difficult to reverse and characterised by discontinuities or thresholds that shift the system into a new configuration, compared to slow, gradual incremental change. While some noted that incremental changes might accumulate and contribute to structural transformation, deep structural change was more commonly seen as involving a qualitative break from previous patterns. Several responses highlighted periods of instability or disruption as part of this process, in which the system may reorder around new structures or priorities.
Other topics emerging in passing included the distinction between scale and depth, the role of intentionality, and the extent to which a change must be profound or radical to qualify as deeply structural. This diversity of thought reflects both the complexity of deep structural change as a phenomenon and the range of domains in which it is seen as relevant. Rather than producing a single definition, the session surfaced multiple ways in which change can be considered structural, opening up productive space for further conceptual and methodological exploration.
A distilled definition
Drawing on both existing literature and the range of perspectives shared by participants, we propose the following working definition. It aims to clarify what is meant by deep structural change from the standpoint of agent-based modelling, while acknowledging its place within broader discussions of transformative change.
Deep structural change is a type of transformative change: From an agent-based modelling perspective, it entails an ontological reconfiguration. This reconfiguration is related to the emergence, disappearance, or transformation of entities, relationships, structures, and contextual features. While transformative change can occur within a fixed model ontology, deep structural change entails a revision of the ontology itself.
Challenges in modelling deep structural change
To understand the challenges modellers face when trying to incorporate deep structural change in ABMs or social simulations in general, we again asked participants in the pre-conference survey and had them brainstorm using a Miro board. We asked them about the “challenges [they] have encountered in this process” and “how [they] would overcome these challenges”. The points raised by the participants can roughly be grouped into: theory and data, model complexity, definition and detection.
The first challenge relates to availability of data on deep structural change and formalisation of related theory. Social simulations are increasingly based on empirical data to be able to model real-world phenomena more realistically. However, the data is often not good at capturing structural system changes, reflecting the status quo rather than the potential. While there are theories describing change, formalising this qualitative process comes with its own challenges, leading to hypothesising of the mechanisms and large uncertainties about model accuracy.
Second, a fine line has to be struck between keeping the model simple and understandable, while making it complex enough to allow for ontologies to shift and deep structural change to emerge. Participants highlighted the need for flexibility in the model structures, to allow new structures to develop. On the other hand, there is a risk of imposing transformation paths, so basically “telling” the model how to transform. In other words, it is often unclear how to make sure the necessary conditions for modelling deep structural change are there, without imposing the pathway of change.
The final challenge concerns the definition and detection of deep structural change. This article begins to address the question of definition, but detection remains difficult — even with greater conceptual clarity. How can one be confident that an observed change is genuinely deep and structural, and that the system has entered a new regime? This question touches on our ability to characterise system states, dominant feedbacks, necessary preconditions, and the timescales over which change occurs.
Closing remarks
Understanding transformative change in general, but increasingly so with the use of social simulation, is gaining attention to provide insights into complex issues. For social simulation modellers, it is therefore important to model deep structural changes. This workshop serves as a starting point for hopefully a wider discussion within the ESSA community on how to model transformative change. Bringing together social simulation researchers showed us that this is tackled from different angles. The definition provided above is a first attempt to combine these views, but key challenges remain. Thus far, people have approached this in a case-by-case manner; it would be useful to have a set of more systematic approaches.
The SIG-MTC will continue to examine questions around how we might effectively model deep structural change over the coming months and years, working with the ABM community to identify fruitful routes forward. We invite readers to comment below on any further approaches to modelling deep structural change that they view as promising and to provide their own reflections on the topics discussed above. If you are interested in this topic and would like to engage further, please check out our ESSA Special Interest Group on Modelling Transformative Change or reach out to any one of us.
Acknowledgements
We would like to thank the participants of the SimSocFest 2025 Workshop on Modelling Deep Structural Change for their engagement in the workshop and the willingness to think along with us.
References
Feola, G. (2015). Societal transformation in response to global environmental change: A review of emerging concepts. Ambio, 44(5), 376–390. https://doi.org/10.1007/s13280-014-0582-z
Filatova, T., Polhill, J. G., & van Ewijk, S. (2016). Regime shifts in coupled socio-environmental systems: Review of modelling challenges and approaches. Environmental Modelling & Software, 75, 333–347. https://doi.org/10.1016/j.envsoft.2015.04.003
Wagenblast, T., Roxburgh, N. and Taberna, A. (2025) Modelling Deep Structural Change in Agent-Based Social Simulation. Review of Artificial Societies and Social Simulation, 8 Aug 2025. https://rofasss.org/2025/08/08/structch
© The authors under the Creative Commons’ Attribution-NoDerivs (CC BY-ND) Licence (v4.0)
Review of Artificial Societies and Social Simulation 