Agent-based modelling: A tool for addressing the complexity of environment and development policy issues

Submitted by Julia Barrott 3rd July 2017 19:02


Integrated policy problems call for integrated analysis. However, micro and macro approaches imply differences of perspective that conventionally have not been easy to unify. This paper introduces agent-based modelling (ABM) as one potentially useful tool for linking these aspects. ABM describes the system at the level of the social actors within it – that is, the individual entities, each with their own goals, values, rules, information, knowledge, strategies and social context. By doing this, ABM can help address the complexity of modern policy problems, particularly when used alongside other methods.

The purpose of this paper is to explain ABM and its applications, to help model users determine whether this approach could be useful in their own work. Motivated by the observation that there is inadequate briefing material on the method, the authors explain ABM and then address four of the most common questions raised when appraising it for research on sustainable development.

They draw on examples of SEI research using ABM for generating insights into a range of policy problems: the climate resilience of agroforestry livelihoods in Cameroon, energy policy and biofuels in Malaysia, sustainable livelihoods in small-scale fisheries in Kenya, and natural hazard disaster preparedness. By linking the example studies to the common questions, they further illustrate key lessons and findings in order to better inform readers.

Download the full paper from the right-hand column. A brief explanation of ABM and the key messages of this paper (taken from the conclusions) are provided below.

What is Agent-Based Modelling?

Agent-based modelling concentrates on describing a social system at the micro-level of the actors within it. This is usually done using a computer model (program). The description for each agent includes a set of instructions or “rules”. Agents also have goals and other internal information (knowledge, beliefs, values, etc.) which uniquely shape their actions. This bundling of data with instructions for agents allows them to be, in practice, coded as autonomous units representing different social entities. The agent descriptions are used as a template to create many copies and thereby populate a model (hence, ABMs are sometimes also known as multi-agent systems or multi-agent models).

In ABM, there is a focus on the micro-behavioural level, but models can include many or multiple types of agency at different levels of action, e.g. households, firms or local authorities. There is also a focus on interactions with other agents and interaction with the environment: ABMs have been used quite extensively to understand management and use of environmental resources, as well as adaptation processes under environmental change.

Agents are, first, endowed with some initial data and rules, and then simulations are made to investigate the results of their interactions, such as patterns of risky behaviours, or shifts in socio-technical regimes. In other words, ABM is an experimental approach for understanding the consequences of modelled assumptions. This can help to generate new knowledge or novel hypotheses. It can be particularly useful for looking, experimentally, at possible future evolutions of the situation (i.e. for producing a simulation).

Considerable detail can be included in ABM because it models low-level behaviours of actors, their decisions and (inter-)actions. The complexity of the situation can be explored, with different/alternative rule-sets, and with populations of heterogeneous agents. 

Key Messages

  • ABMs often need to be built – and understood – from scratch. They produce findings that are not easily transferable, and outputs that can appear to be quite difficult to understand. However, modern policy-making is much more complex than it appears and scientific assessments may be able to greatly benefit from using ABM.
  • Solutions tested in an ABM can be quite complex: adaptive actions such as policy withdrawal or policy-corrective strategies for “wicked policy problems” can be simulated.
  • A key strength of ABM is that it can include more detail and context than other modelling approaches. Models can be used to explore different trade-offs and interdependencies among policies, and different scales of decision-making, also including important micro-level priorities. Moreover, in ABM, the time dimension is explicitly acknowledged, which is crucial when assessing sustainability.
  • ABMs can allow us to conduct a deeper investigation of different scenarios for sustainability, by simulating the consequences of actions or measures taken and analysing conditions under which they may do well. They allow us to better understand how people may adapt differently to different types of interventions, in the long and the short run, and how their vulnerability and resilience may change. 
  • Examples of ABMs that have actually been used to support policy decision-making or for other purposes “in the real world” are increasing. Though it is still not a mainstream method, it is fair to say that ABM is now quite well regarded as a relevant and valid scientific method, although still lacking rigour in how models are tested and results communicated.
  • The extent to which ABM will be adopted outside of academia also depends on whether it will be viewed as a trusted, legitimate and practical method by those applying it. In an environment and development context there are often very significant knowledge gaps and complexities to come to terms with. We have found that ABM is a powerful tool to address complexity when it is used along with other methods.
  • One cannot address complex problems with simple solutions. ABM should be evaluated in comparison to other modelling approaches. ABM provides results and dynamics that are not possible to produce with standard modelling approaches. In particular, ABMs may be preferred to more mathematical, aggregate models for they bring greater opportunities for interaction and co-creation. Such modelling helps to structure people’s understanding of the situation. Both the modelling process and model outputs can help to clarify and to communicate that understanding

Further resources