# 5.1: Introduction What makes an agent-based model *empirical*, and how are empirical ABMs built, grounded, and validated? This chapter answers that question. Agent-based models (ABMs) have been successfully applied to a variety of complex real-world problems for nearly 20 years, with a particular focus on coupled socio-technical and socio-environmental systems (STE systems). Some newcomers to the field still echo outdated critiques suggesting that ABMs lack empirical grounding, and this chapter exists in part to answer them: the reality is quite the opposite, because the field has moved towards a sophisticated, data-driven paradigm in which modern ABMs are recognised as essential tools for capturing the "messy" reality of human and environmental interactions that traditional models often overlook. This introductory section defines what an empirical ABM is, sets out why such models matter, traces how their empirical rigour developed, and previews the methodological questions that the rest of the chapter takes up. ## 5.1.1: What is an empirical ABM? Before examining why empirical ABMs matter, it helps to fix precisely what the term means. An empirical model does more than place real numbers in an otherwise abstract simulation. Evidence may inform which agents are represented, their initial attributes, the environment and networks they occupy, the rules by which they decide, and the observations against which the output is calibrated or validated. Different studies ground different combinations of these elements, depending on the question and available data. Empirical grounding also does not mean copying every observed detail. Data still have to be selected, interpreted, and translated into modelling assumptions. A useful empirical ABM makes that translation visible: it states which evidence supports each important choice, where evidence is missing, and how uncertainty in the data affects the conclusions. Section 5.2 develops these dimensions in detail. ```{admonition} In a nutshell :class: tip An *empirical* ABM is one grounded in real-world observations rather than abstract assumptions, and today's ABMs are firmly data-driven: they use empirical evidence to represent heterogeneous decision-makers and their environment, and to explain how micro-level behaviour gives rise to macro-level outcomes. ``` The vocabulary below recurs throughout the chapter; the flip-cards introduce each term where it is first needed, before the argument puts it to work. ```{raw} html :file: ../learning_activities_and_widgets/widgets/widget_5_1a_key_terms_at_a_glance.html ``` ## 5.1.2: Why empirical ABMs matter With that definition in hand, we can see what gives empirical ABMs their distinctive value. They are particularly valuable for exploring the heterogeneous micro-level behaviours of actors that represent a variety of real-world decision-makers and stakeholders in a dynamic environment, enabling a generative understanding of how micro-level behaviours give rise to macro-level outcomes (Arthur, 2021[^1]; Balint et al., 2017[^2]; Epstein, 2012[^3]; Filatova et al., 2013[^4]; Gallagher et al., 2021[^5]). These models allow researchers to simulate learning, negotiations, and adaptation, as well as the intricate interactions between agents and their environment, whether that environment is natural (such as spatial maps of real land use), technical (infrastructure), or institutional. The reasons above become sharper as concrete choices. In the activity below, each scenario poses a modelling situation, and the task is to identify what empirical grounding actually buys the model. ```{raw} html :file: ../learning_activities_and_widgets/learning_activities/activity_5_1b_why_ground_a_model.html ``` The exercise makes the section's point tangible: grounding is what lets a model represent real decision-makers, earn the trust of stakeholders, explain how micro-level behaviour produces macro-level outcomes, and avoid resting on assumption. ## 5.1.3: Endogenous institutions and a track record of empirical validation That same capacity extends beyond individual actors to the institutions they inhabit. ABMs have been used to model various institutions endogenously, ranging from formal structures such as markets and policy frameworks to informal ones such as social norms. Modelling these institutions as *endogenous* means treating them as arising from agents interacting rather than as fixed conditions imposed from outside the model. This ambition is not new, and it has a substantial track record of empirical validation behind it. Since the early 1990s, financial and energy market ABMs have led the way in employing empirical data for both design and validation. This empirical rigour is now standard in the study of social-ecological systems (SES). As noted by Schulze et al. (2017)[^6], ABMs are uniquely positioned to represent the behaviour of human actors and institutions in a way that provides general insights for sustainable resource management. The contrast between endogenous and exogenous is easier to grasp through examples, so the activity below asks you to sort concrete model features into the two categories. ```{raw} html :file: ../learning_activities_and_widgets/learning_activities/activity_5_1c_endogenous_or_given.html ``` Sorting the features clarifies the section's claim: an ABM is most illuminating when institutions such as markets and norms grow from agent interaction rather than being assumed from the outset. ## 5.1.4: From theory to grounded theory: the empirically stylised approach Yet empirical maturity in data use does not by itself resolve how to reason about complex causality, and it is here that a methodological gap remains. While the potential of ABM for developing theory is well recognised, the field has historically lacked specific methodologies to handle complex causality and social-ecological interdependencies. To address this, a new approach has been proposed that combines case-based empirical research with collaborative modelling to ensure that theories are contextually embedded, collaboratively abductive, and *empirically stylised* (Schlüter et al., 2025)[^7]. This approach yields better-grounded, transparent assumptions and opens new interdisciplinary avenues for discovering how social and environmental systems truly interact. The empirically stylised approach is characterised by three traits; the cards below turn each trait over to show what it means in practice. ```{raw} html :file: ../learning_activities_and_widgets/learning_activities/activity_5_1d_marks_of_a_grounded_theory.html ``` Taken together, these three traits are what separate a grounded, transparent theory from an abstract one, which is precisely the shift this section describes. ## 5.1.5: What this chapter covers Together these threads answer the question the chapter opened with: an empirical ABM is one whose heterogeneous agents, environments, and institutions are grounded in evidence rather than assumption, and building one well is a methodological discipline in its own right. Having established what empirical ABMs are and why they matter, the next section (5.2) turns to the *dimensions* along which a model can be grounded, that is, exactly what parts of a model empirical evidence is used to support. --- ## Key takeaways ```{admonition} Key takeaways :class: important - An empirical ABM is one whose agents, environment, and rules are grounded in real-world data rather than assumed. - Far from lacking empirical grounding, modern agent-based modelling has become a data-driven approach for capturing socio-technical and socio-environmental systems that traditional models often overlook. - The rest of the chapter examines what can be grounded, what data can ground it, and the practices that keep an empirical model credible. ``` ## References [^1]: Arthur, W. B. (2021). Foundations of complexity economics. *Nature Reviews Physics*, 3(2), 136–145. https://doi.org/10.1038/s42254-020-00273-3 [^2]: Balint, T., Lamperti, F., Mandel, A., Napoletano, M., Roventini, A., & Sapio, A. (2017). Complexity and the economics of climate change: A survey and a look forward. *Ecological Economics*, 138, 252–265. https://doi.org/10.1016/j.ecolecon.2017.03.032 [^3]: Epstein, J. M. (2012). *Generative social science: Studies in agent-based computational modeling*. Princeton University Press. https://doi.org/10.23943/princeton/9780691158884.003.0003 [^4]: Filatova, T., Verburg, P. H., Parker, D. C., & Stannard, C. A. (2013). Spatial agent-based models for socio-ecological systems: Challenges and prospects. *Environmental Modelling & Software*, 45, 1–7. https://doi.org/10.1016/j.envsoft.2013.03.017 [^5]: Gallagher, C. A., Chudzinska, M., Larsen-Gray, A., Pollock, C. J., Sells, S. N., White, P. J. C., & Berger, U. (2021). From theory to practice in pattern-oriented modelling: Identifying and using empirical patterns in predictive models. *Biological Reviews*, 96(5), 1868–1888. https://doi.org/10.1111/brv.12729 [^6]: Schulze, J., Müller, B., Groeneveld, J., & Grimm, V. (2017). Agent-based modelling of social-ecological systems: Achievements, challenges, and a way forward. *Journal of Artificial Societies and Social Simulation*, 20(2). https://doi.org/10.18564/jasss.3423 [^7]: Schlüter, M., Wijermans, N., González-Mon, B., Lindkvist, E., Orach, K., Prawitz, H., Martin, R., Martínez-Peña, R., Pellowe, K. E., & Sanga, U. (2025). Navigating the space between empirics and theory – Empirically stylized modelling for theorising social-ecological phenomena. *Environmental Modelling & Software*, 189, 106444. https://doi.org/10.1016/j.envsoft.2025.106444