Agent-Based Model
- Agent-Based Model Basics: A Guidebook & Checklist for Policy Researchers (Johnson, Liz)
The time has come to examine complexity in the world and its impact on policy under a new lens of opportunity. Everything that happens in the world is new and the world does not repeat itself (Bar-Yam 2010). Agent-based modeling represents and describes unique ways to think about mimicking and learning about the world. Bar-Yam offers, “We want to generalize ideas that can be used to take the past to the future” (Bar-Yam 2010). The goal is to learn with simulation models while capturing more and more levels of system details.
The key to translating complexity mechanisms into useful policy and research is learning how parts of systems interact and give rise to patterns. The next step is to observe the dynamic relationships of the whole and interdependent sub-systems.
People are represented as agents in simulations. Agents themselves can choose their level of connectedness, interdependence, and responsiveness. Attention should be focused on evolutionary learning and dynamic connections in politics and policy instead of designing static institutions, laws, regulations, and other traditional policy instruments (OECD 2008).
The modeler sets the parameters of how variable values are to be represented. Some software programs like NetLogo allow for sliders of variables where the programmer has the flexibility to move the slider to various values. Building in randomness in relation to agents and interactions is a prime way to abstract complexity in the real social world (Gilbert 2008). The model can provide insights, trends, and tipping points.
The modeler establishes simple rules to guide agent actions and environment interactions. Model rules follow an “if then” questioning step-by-step process. For example “if” the agent meets a like agent “then,” the agent makes a decision and takes a specified action. The consequent actions create interactive effects and externalities. The rules direct discreet agent actions by agents like movement, emotions, strategy, and/or decisions and are threshold-based.