The Naming Game is a simulation that shows how agents interacting can bootstrap small a shared vocabulary from nothing. For more details, see Baronchelli, Loreto and Steels (2008).

The agents are communicating about different colours (meanings). Agents start with no word for any of the colors (?), but through repeated communicative interaction, they evolve shared words. Just by choosing random strings and adjusting their vocabularies according to whether those strings are communicatively successful, the population evolves shared conventions for naming each colour.

The graph on the bottom shows success (on the y axis) over time (the x axis) for each color "meaning". Success is at first low for all colors, but increases over time, indicating that a shared name for the color is emerging. The large bar on the right indicates the proportion of successful interactions (green) for a sliding window of the last 100 interactions. The "eyes" of the agents indicate their individual success over the same window. When the population has reached consensus on a name for a color, it brightens. The time to consensus is shorter when there are fewer meanings and/or a smaller population. Note that in the naming game, one unit of time is generally defined as N interactions, where N is the population size.

Change some parameters using the controls on the top of the page: speed up time by changing the interaction rate. Change the number of agents in the population, or change the number of colors they have to name to see how this changes the rate at which consensus is reached among the agents. Note that each time the simulation is re-set, this constitutes a single run - investigations of the naming game generally average over hundreds of runs with the same parameters to determine the properties of a typical system.