The micro-foundations of coupled learning processes: Organizational learning and allocation of attention

We examine how attention allocation influences a process of coupled learning. We obtain experimental evidence from a design where a boss, or principal, must learn to identify rewarding tasks based on the feedback provided by an agent. The agent, in turn, is engaged in classical regression learning with the added complication that her performance depends on the task the principal gives to her. We combine laboratory experiments with the development of a multi-level learning model that closely replicates and predicts experimental evidence. Our experiments provide evidence for large performance heterogeneity across the principal-agent pairs, and show that allocation of attention is a major determinant of long-term performance. Our model suggests that early interaction of attentional processes involving both the principal and the agent decisively affects long-term performance, generating self-reinforcing dynamics that lock the principal-agent pairs in either vicious or virtuous coupled learning processes.