What is agent-based modelling and how can it be useful for public health research?

NCRM news
Sophie Jones and Leandro Garcia, Centre for Public Health at Queen’s University Belfast
People jogging on a roadPeople jogging on a road

As public health researchers, we are accustomed to wrestling with multifaceted challenges that surpass simple cause-and-effect relationships. Whether it be disease transmission, obesity, or addressing health inequalities, our work often demands a nuanced understanding of the complex systems that underpin these issues.

Research methods traditionally used in public health and epidemiology have undoubtably contributed valuable insights to our understanding of such public health problems. Yet, the nature of these problems many times demand an approach able to account for the non-linear and dynamic interactions that occur between individual factors, social determinants, and environments for improving our understanding of public health challenges.

Agent-based modelling has the potential to serve as a valuable tool to understand these interactions, which could be integral to enhancing our understanding of complex phenomena and our ability to make informed public health decisions. Therefore, this blog describes what agent-based modelling is, and also highlights how the method could be useful for public health research.

What is agent-based modelling?

Agent-based modelling is a method used to investigate the emergence of population-level patterns and behaviours, by simulating autonomous agents (such as individuals, organisations, or places) that make decisions and take actions based on their characteristics and on the interactions with each other and their environments. The method adopts an ‘agent-centric’ perspective, whereby decisions by the agents in the model are made through the lens of the agent themselves, guided by a set of programmable rules.

These rules, along with each agent’s own characteristics, dictate how agents interact and make decisions as time progresses. By simulating these processes, we are able to demonstrate how individual-level decisions and interactions aggregate over time and generate population-level patterns and behaviours.

Through these simulations, agent-based models provide insights into the dynamics underpinning a system’s behaviour and offers a nuanced understanding of the dynamic factors and processes involved.

How can agent-based modelling be useful for public health research?

In public health, we have already seen the application of agent-based models for modelling disease transmission and health behaviours such as physical activity, diet, or alcohol consumption. In particular, these models have the ability to simulate interactions at the individual level and help us understand how they aggregate over time, providing a deeper understanding of the dynamic ways structural and environmental factors shape the population-level outcomes we observe.

For instance, when investigating obesity, it is widely acknowledged that population trends of obesity are a complex problem due to the multiple factors across different levels (such as an individual’s psychology, social context and norms, and the physical environment) which dynamically interact with each other to shape trends of obesity levels. Agent-based models help us to explore and interrogate this problem by modelling individual-level decisions and interactions (for example, an individual’s BMI influenced by their diet, physical activity, social interaction, or physical environments), and observing how they shape the population trends over time (such as obesity levels). By modelling the decisions and interactions at the individual level and observing how and why these aggregate to become population patterns, we are able to develop a deeper understanding of the relationships and processes involved, consequently allowing for more informed decisions about how to intervene.

Further to this above point, agent-based models can also act as virtual labs, enabling modellers to explore hypothetical scenarios, and forecast the potential (intended and unintended consequences) of interventions. This ‘what if’ capability is especially useful to discuss with stakeholders or decision makers when planning strategies to tackle public health challenges.

Therefore, the ability of agent-based models to enhance our understanding of public health challenges, and consequently improve our efforts to plan effective interventions, are two key contributions of the method to the public health field.

Notwithstanding of their potential for public health research, agent-based modelling is very much still in its infancy in this area. For this reason, the next section will highlight what to expect when developing agent-based models.

What should I expect if I want to start developing agent-based models for my research?

The first point to be emphasised is that although the behaviour may be complex, the computational model itself does not have to be. It is very easy to get lost with modelling if one does not have clear boundaries about what the model’s purpose is. That brings us to our second point, which is that building agent-based models can be a time demanding process. From conceptualising the model (for example, formulating mental models about how the processes happen in real life) to programming the model, the development of these models requires perseverance, patience, and openness to learn a skillset and mindset that normally is outside what public health and epidemiology researchers and practitioners are trained to us. Finally, agent-based models are useful tools for understanding the system of interest, however they are often not suitable for exact prediction of how the system will behave in the future.

In summary, agent-based models offer an innovative avenue for public health research and practice, facilitating nuanced insights of complex problems, which alongside methods more conventionally used in public health and epidemiology, could be useful for the exploration and informed, effective planning for public health strategies.


Training on agent-based modelling

On 1-4 July 2024, NCRM is running a course on agent-based modelling and how it can be used in public health research. Participants will learn how to build agent-based models and use them to run experiments.

Register for the course, Introduction to Agent-based Modelling for Public Health Research


Online tutorials

NCRM also has a series of three online video tutorials about agent-based modelling and they can be used in social research. The resources provide an introduction to the technique, show you how to analyse models and, finally, explore the documentation of agent-based models and provenance for simulation studies. They can be accessed using the following links:

  1. Introduction to Modelling, Introduction to Agent-based models, Time in Agent-based models
  2. Data Quality, Model Analysis and Psychological Experiments
  3. Documentation of Agent-based models and Provenance for Simulation Studies