Expeditions Without Maps: When are faulty initial representations useful in coupled learning?

Abstract

Coupled learning processes, in which specialists from different domains learn how to make interdependent choices, are common in organizations. We analyse when and why faulty initial representations can be useful in such learning processes. We show that when learning is rapid but communication is restricted, faulty initial representations held by all agents can suppress superstitious learning and promote valid learning, relative to situations where no agent has a representation, or a mix of correct and incorrect representations exists among the agents. Our results have implications for the design of coupled learning processes.