The obvious question of when to stop the nested levels of modeling has two answers. First, theoretically, an agent should model others until it runs out of modeling knowledge. Thus, if an agent has information about how another agent models a third agent modeling the original agent, etc., to, say 10th level, then the modeling should go on all the way to the tenth level to include everything that is known. But, if the agent has no information beyond, say, the second level (i.e., agent has information on how to model other agents, but has no clue about the models the others use to coordinate), then the nesting stops just below the second level with, what we call, a no-information model.