2.1.2: Why is ABM important?#
2.1.2.1: Why is ABM relevant?#
Agent-based modelling (ABM) has become increasingly relevant as we face complex challenges such as pandemics, climate change, financial crises, and urban planning. In each case, people or organisations differ, interact locally, respond to changing conditions, and may adapt after an intervention. Aggregating them into one average can remove exactly the variation and feedback responsible for the outcome. ABM preserves those differences and asks how population-level patterns grow from them.
That does not make ABM the automatic choice for every difficult problem. Its extra detail is justified when individual heterogeneity, interaction, adaptation, or spatial structure is central to the research question and cannot be represented adequately at an aggregate level. When those features do not affect the conclusion, a simpler modelling approach may be clearer and cheaper. The relevance of ABM therefore comes from matching its distinctive representation to the structure of the problem, not from complexity alone.
2.1.2.1.1: What is a complex adaptive system?#
A complex adaptive system is one in which many parts interact and adjust to one another, so that the behaviour of the whole is not something any single part decides or intends. A great deal turns on telling these systems apart from simple ones, because it is the complex adaptive systems, and only those, that reward the extra effort of agent-based modelling; a simple system that follows one clear rule can be captured with a single equation.
Modern society is full of such systems, in which individual behaviours create surprising collective outcomes:
Financial markets: Individual trading decisions create market bubbles and crashes
Urban traffic: Individual route choices lead to system-wide congestion patterns
Social movements: Individual opinions generate viral trends or social movements
Ecosystem dynamics: Individual species behaviours create ecosystem stability or collapse
Learning to spot the difference between these and simple, rule-bound systems is therefore the first step in judging whether a model of this kind is worth building at all. The activity below presents a mix of systems to sort into complex adaptive or simple and predictable; in each case, ask yourself whether the parts are merely following a fixed rule or truly reacting to each other.
2.1.2.1.2: Why traditional models fall short#
Having recognised what makes a system complex and adaptive, we can see why the usual modelling toolkit struggles with it. Traditional models often assume rational, homogeneous actors, focus on equilibrium states, and miss how micro-level variation affects macro-level outcomes. That is precisely the gap agent-based modelling is built to fill.
The three cards below make those assumptions concrete: each states one, then shows why it fails for complex social systems and how agent-based modelling addresses it.
Taken together, these are the shortcomings that justify modelling a system at the level of its individuals.
2.1.2.1.3: What ABM adds: six capabilities#
Agent-based modelling closes that gap through six distinct capabilities: capturing emergent behaviour, modelling heterogeneity, representing adaptation, exploring dynamic feedback, testing interventions safely, and bridging disciplines. Select each capability to see what it means and an example:
ABM studies how collective phenomena emerge from individual interactions without pre-assuming outcomes.
Example: How individual health compliance creates population-level disease control.
ABM recognises that individuals differ in attributes and behaviours, and that this heterogeneity matters for system outcomes.
Example: Designing vaccination strategies for different risk groups.
ABM models how agents learn and evolve strategies over time as systems adapt to changing conditions.
Example: Understanding how people adapt their behaviour during crises.
ABM captures how agents influence their environment, which influences other agents, creating complex feedback loops.
Example: How individual energy consumption affects the climate and future behavioural choices.
ABM provides a “virtual laboratory” to test policies before real-world implementation, without risk and at low cost.
Example: Testing evacuation procedures or economic policies.
ABM enables interdisciplinary collaboration by providing a common framework combining insights from multiple fields.
Example: Pandemic response integrating epidemiological, economic, and social insights.
Taken together, these six capabilities are what let ABM see a complex adaptive system as it actually is: a population of diverse, interacting, adapting individuals rather than an aggregate average.
Important
Key insight: ABM is relevant not only as a modelling technique, but as a way of thinking about complex systems that recognises individual diversity, local interactions, and emergent phenomena.
Traditional approaches often assume homogeneity or aggregate behaviour, missing individual differences and local interactions. By filling this gap, ABM becomes a powerful tool for modern science and policy.
2.1.2.1.4: ABM as a virtual laboratory#
The fifth of those capabilities, testing interventions safely, is worth seeing in action. One of the most valuable things a model gives you is a safe place to test an intervention before doing it for real, and to discover that the obvious choice is not always the best. In the activity below, a disease is spreading through a town and you have a limited budget to split between two policies. Before you begin, predict whether splitting the budget between the two will beat spending it all on the obvious choice.
2.1.2.2: Example domains for ABM#
Having seen what ABM can do, we now turn to where those capabilities get applied. ABM is used across diverse fields in which individual behaviours create complex system-level patterns. Select a domain to see example applications[1]:
Predator-prey dynamics: Understanding how hunting strategies and escape behaviours create population cycles and spatial distributions
Ecosystem resilience: Studying how individual species responses to disturbances affect overall ecosystem stability and recovery
Consumer behaviour: Understanding how individual purchasing decisions lead to demand patterns and market trends
Supply chain management: Analysing how individual firm decisions affect supply chain resilience and disruptions
Residential segregation: Understanding how individual housing preferences lead to neighbourhood demographic patterns (Schelling’s model)
Collective action: Studying how individual participation decisions affect social movements and protest dynamics
Social network formation: Analysing how individual relationship choices create community structures and information flow
Disease spread modelling: Simulating how individual contact patterns and behaviours affect epidemic trajectories
Vaccination strategies: Testing how individual vaccination decisions impact population-level herd immunity
Health behaviour adoption: Understanding how individual health choices spread through social networks
Traffic flow optimisation: Modelling how individual driving decisions create congestion patterns and travel times
Crowd dynamics: Studying pedestrian flows and evacuation patterns from individual movement behaviours
Public transportation usage: Analysing how individual mode choice affects system efficiency and ridership
Emergency evacuation: Modelling how individual evacuation decisions affect crowd flow and safety outcomes
Disaster response coordination: Understanding how individual responder actions create effective emergency response
What is striking about agent-based modelling is how far the same idea travels. Ecology, economics, social science, epidemiology, transport, and safety look like very different fields, yet each contains systems where individuals interact and a pattern emerges at the level of the whole, which is exactly the shape of problem this method is built for. Seeing the common thread matters because a technique you learn in one setting can be carried into another, and because it helps you recognise, in a new problem, whether an agent-based model might apply. In the activity below, match each application to the field it belongs to, and notice how similar the underlying descriptions are once the domain labels are stripped away.
The takeaway is that one pattern, individuals interacting to produce a system-level outcome, recurs across many disciplines, which is why a single modelling approach can serve so many of them.
Key takeaways#
Key takeaways
Many of the systems worth studying are complex adaptive systems, in which heterogeneous, interacting, adapting individuals produce collective behaviour that aggregate models cannot capture.
ABM matters because it represents these individuals directly and acts as a virtual laboratory for exploring scenarios and interventions that cannot be tested in reality.
Because a model’s assumptions can shape real decisions, this capability also carries a responsibility to build and communicate models transparently.
2.1.2.3: Social implications of ABM#
Being applicable across so many fields is not only a technical convenience: it gives ABM a reach into public life. Beyond the fields where it is applied, the use of ABM has important social and policy implications[1]. Rather than serving merely as theoretical exercises, agent-based models (ABMs) function as powerful tools to inform policy actors, coordinate interdisciplinary research, and engage the general public. Two of these three roles, safe intervention testing and bridging disciplines, appeared earlier as capabilities; here we view them through a policy and public lens, and add a third role, public engagement. The Virus on a Network model illustrates this threefold utility well. By simulating a virus spreading through a population, policymakers can safely test “what-if” intervention scenarios, such as school closures or targeted vaccinations, to predict real-world impacts without risking lives. Developing these pandemic models also requires collaborative research that bridges epidemiology, network science, and human behaviour into a single framework. Finally, when these interactive simulations are shared with the public, they transform complex systemic risks into accessible, visual tools, helping citizens grasp the necessity of policy trade-offs and collective behavioural shifts during a crisis.
The three cards below set out these roles, each illustrated by the pandemic example just described.
One and the same model can therefore serve science, policy and the public at once.
2.1.2.3.1: Worked example: virus on a network#
To make the Virus on a Network model just mentioned concrete, consider a virus spreading through a network of individuals. Each agent (person) can be susceptible, infected, or recovered. The network structure (who meets whom) and individual behaviours (mask-wearing, social distancing) determine how the disease spreads. ABM allows us to:
Simulate outbreaks on realistic social networks
Test the effect of interventions (e.g., closing schools, targeted vaccination)
Observe emergent patterns like super-spreader events or herd immunity
For example, the figure below shows several runs of a Virus on a Network model with varying network structures and policies. These runs highlight how small changes in individual behaviour or network connectivity can lead to markedly different epidemic outcomes, demonstrating the value of ABM for understanding and managing real epidemics.
Fig. 15 Figure 2.1 Virus On A Network model: runs with different network structures and individual behaviours#
The lesson to take away is that in a network with hubs, which few people you protect can matter more than how many, and a model like this lets you find that out safely before any real decision is made.
Now that we know why ABM matters and where it is used, the next section (2.1.3) compares ABM with the other main modelling paradigms to clarify when it is the right choice.