Agent-based model

An agent-based model (ABM) is a type of computer simulation that shows how individual things act and interact to see how they affect the whole system.^[1] It can be thought of as a video game where lots of characters are moving around and doing their own thing. Each character follows a few simple rules like “walk toward food,” “avoid danger,” or “follow the leader.” These characters are called agents, and they can be anything depending on what you are studying: people, animals, cars, robots, or even tiny cells in the body.^[2] An agent-based model (ABM) is like running a giant computer simulation of these agents to see what happens when they all interact. Even though each one is just following simple rules, together they can create surprising and complicated patterns. This is called emergence, which means big patterns form from lots of small actions without anyone planning it.^[3]

Schelling's model of segregation is an example of an agent-based model (ABM)

Scientists use ABMs to study all kinds of real-world situations. In nature, they can model how animals migrate, fight for resources, or react to changes in the environment, like guessing how a fish population might recover if fishing rules change.^[4] In economics, ABMs can show how shoppers, businesses, and banks interact, sometimes leading to things like financial bubbles or inequality.^[5] In epidemiology, they can help predict how diseases spread through a community, factoring in things like different social circles and immunity levels.^[6] Even less obvious uses exist, like simulating traffic to prevent jams,^[7] studying how people move in a crowd to design safer buildings,^[8] or figuring out how ancient civilizations traded and grew.^[9]

One of the most interesting things about ABMs is that not every agent has to be the same.^[2] A forest fire simulation, for example, might have some trees that are dry and catch fire easily, while others are wet and harder to burn. The fire can spread tree-to-tree, while weather affects it on a larger scale. Sometimes these models include feedback loops, where agents change the system, and then the changed system affects the agents’ next actions.^[10] This can cause unexpected events or “tipping points” where everything changes suddenly.^[11]

Building an ABM takes a lot of careful planning. You have to decide what your agents are, what rules they follow, how they interact with each other and their surroundings, and how time works in the simulation.^[12] Scientists often use special computer programs like NetLogo, Repast, or AnyLogic to run these models and watch what happens.^[13] ABMs can be powerful tools for testing “what if” questions and spotting patterns that might be hard to see in the real world. Even though they can be tricky to set up and need a lot of data, they help researchers understand how simple actions can lead to complex, unpredictable results.^[14]