Expeditions without Maps: Why Faulty Initial Representations May Be Useful in Join Discovery Problems

When interdependent agents face structural uncertainty - i.e. they do not know ex ante how their choices interact, they must obtain such knowledge through shared trial-and-error learning experiences that build off their individual initial representations. We use an agent based model to formally analyse how joint learning and initial individual knowledge - and their interactions - influence the process of achieving coordination under conditions of structural uncertainty.