Agent-based modeling (ABM) is a computational modeling technique where individual AI agents, programmed with specific behavior rules, interact with and change their environment within the model — including by interacting with other agents. ABM is a type of simulation that helps make sense of complex systems from the ground up.

Scientists frequently use modeling programs to see how agents do different things and make different choices. Simplified as such, modeling helps us understand complex systems like how traffic flows in a city, how a disease spreads through a population that is or isn’t wearing masks, or even how a colony of ants or a box full of toy people could work together.

In his seminal paper on creating computer models of complex social interaction, Justin E. Lane writes extensively about ABM and its subset, multi-agent AI (MAAI).

You can think about it this way: Imagine a toy box filled with lots of little toy people, and each toy person is an agent. Each agent can do different things and make different choices. When you take all the agents out of the toy box and let them play together, they can form different groups and even do different things with each other, like build a tower or play a game.