Systems Model Series: Agent-based Modeling

Agent-based Modeling, System, Complex

Adapted from the article originally written by Chunyu Lin

for the Cabrera's Cornell University, System Thinking in Public Affairs Course



Agent-based modeling, known as ABM, is an influential simulation modeling technique that has been widely used in cases for recent years. It combines a variety of elements including game theory, complex systems, emergence, computational sociology, multi-agent systems, and evolutionary programming. It is also well-known as an effective problem-solving tool in the business world. Instead of handling certain practical or engineering issues as the multi-agent system does, the objective of ABM is to focus on providing rationality for the collective behaviors of agents obeying simple rules, especially in natural systems. Under the scope of ABM, a system ‘is modeled as a collection of autonomous decision-making entities called agents’ (E Bonabeau). Each agent individually analyzes its situation and makes decisions based on a set of rules.

Core principles:

The three major ideas for ABM are objects, emergence, and complexity.

  • “objects” are defined as “computational entities that encapsulate some state, are able to perform actions, or methods, on this state, and communicate by message passing” (Wooldridge, 2009).
  • There are four features that relate to the ‘emergence’ in ABM. Emergence will not occur for a single independent Emergent Behavior is Inversely Proportional to the Degree of Bondage between Systems. Emergent Behavior is Non-linear. Emergent behavior is Self-organized (Manson, 2011).
  • The complexity refers to aggregate complexity. It concentrates on how complex systems derive from interactions over individual entities (Manson, 2011).Application:


Southwest Airlines used an agent-based model within their cargo touring operations to facilitate efficiency of their transfer of cargo (Seibel and Thomas, 2000).

Macy’s have used an agent-based model for their new store design. With ABM, Macy's had the chance to use visualization to monitor data in a way that becomes informational and directs to solutions. (Bonabeau, 2003).

Pacific Gas and Electric have used an agent based model to predict how energy flows through the power grid (Bonabeau, 2003a).

Pros and Cons


  • ABM performances well in discerning emergent phenomena
  • ABM offers a very natural description of a system
  • ABM offers a lot of flexibility and generality in its applications


Choosing enough number of parameters, characteristics, and behaviors to include in the model would be challenging not only the practical level but also the theoretical level. It would be impossible to interpret the model if there are too many variables (Manzo)

*CA & ABM :


Both agent-based modeling and cellular automata start with individual agents or cells and model a huge number of interacting agents causing the complex behavior.

CAS often binds to only a few rules to update state depending on its neighboring states, thus this has limited application. Whereas ABM will give more realistic simulations, especially when handling social phenomena and complex adaptive systems.


Reference and further reading

  1. Eric Bonabeau. PNAS May 14, 2002 99 (suppl 3) 7280-7287;
  2. Gianluca Manzo. Revue française de sociologie. DOI: 10.3917/rfs.554.0653. https://www.cairn- and-limitations-of.htm
  3. Michael Woodbridge. 2009. An introduction to Multi-Agent system. 2nd ed. John Wiley & Sons Ltd, ISBN 978-0-470-51946-2.
  4. Steven M. Manson. Agent-Based Models of Geographical Systems.