Decision Centralization and Learning from Experience in Groups

Abstract

Organizational learning frequently involves groups that learn from feedback on decisions over time. While it is frequently assumed that this process resembles that of individual learning by doing, there is limited evidence to validate the assumption. Further, groups in organizations often have centralized rather than decentralized decision rights but we know little about how they differ when learning by doing. We compare individuals, decentralized and centralized groups undertaking learning by using experiments and computational models. We find that centralized groups behave like hyper-individuals: they update more rapidly and explore more than individuals (who in turn update and explore more than decentralized groups). Our evidence shows that not only do groups differ from individuals because of aggregation processes, but also individuals change how they behave simply by virtue of being in a group (a context effect). Implications are drawn for how this might alter the way we conceptualize and model organizational learning.