Part I. Introduction
I-3. Introduction to agent-based modeling
In this chapter, we briefly explain what agent-based modeling (ABM) is about, including a paradigmatic example. We also try to clarify the relationship between evolutionary game theory and agent-based modeling, and offer some arguments in favor of implementing evolutionary models using an agent-based approach. Finally, we provide some references to other books that can be used to learn more about how to implement agent-based models.
1. What is agent-based modeling?
Agent-based modeling (ABM) is a methodology used to build formal models of real-world systems that are made up by individual units (such as e.g. atoms, cells, animals, people or institutions) which repeatedly interact among themselves and/or with their environment.
The essence of agent-based modeling
The defining feature of the agent-based modeling approach is that it establishes a direct and explicit correspondence
- between the individual units in the target system to be modeled and the parts of the model that represent these units (i.e. the agents), and also
- between the interactions of the individual units in the target system and the interactions of the corresponding agents in the model (figure 1).
This approach contrasts with e.g. equation-based modeling, where entities of the target system may be represented via average properties or via single representative agents.
Thus, in an agent-based model, the individual units of the system and their repeated interactions are explicitly and individually represented in the model (Edmonds, 2001).[1] Beyond this, no further assumptions are made in agent-based modeling.
At this point, you may be wondering whether game theory is part of ABM, since in game theory players are indeed explicitly and individually represented in the models.[2] The key to answer that question is the last sentence in the box above, i.e. “Beyond this, no further assumptions are made in agent-based modeling“. There are certainly many disciplines (e.g. game theory and cellular automata theory) that analyze models where individuals and their interactions are represented explicitly. The key distinction is that these disciplines make further assumptions, i.e. impose additional structure on their models. These additional assumptions constrain the type of models that are analyzed and, by doing so, they often allow for more accurate predictions and/or greater understanding within their (somewhat more limited) scope. Thus, when one encounters a model that fits perfectly into the framework of a particular discipline (e.g. game theory), it seems more appropriate to use the more specific name of the particular discipline, and leave the term “agent-based” for those models which satisfy the defining feature of ABM mentioned above and they do not currently fit in any more specific area of study.
The last sentence in the box also uncovers a key feature of ABM: its flexibility. In principle, you can make your agent-based