发表机构
Toulouse School of Economics, University of Toulouse Capitole（图卢兹经济学院，图卢兹第一大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文研究决策者间的建议传递，发现动机偏差会导致鼓励无信息价值，使社会因无人尝试或过度努力而停止学习，影响长期福利。
AI 中文摘要
建议不仅能决定对过往证据的解读方式，还能决定是否会产生新证据。本文研究一系列决策者，他们观察到公开的成功或失败，但无法看到彼此的努力程度，且会将无成本的因果评估传递给继任者。由于发送者重视成功，却低估其继任者的努力成本，因此有理由推荐要求更高的行动。这种动机可能使鼓励变得无信息价值。在模糊失败后，所有均衡都可能导致弃权（不执行），从而实验和学习停止。当最大努力使结果对基础技术不敏感时，同样的力量会产生看似相反的陷阱：每个人都努力工作，但没人测试更少的努力是否足够。崩溃随动机偏差单调变化，而适度的动机会保留因果分歧和识别实验。在阈值应用中，更强的动机会提高观察到的成功，同时降低长期福利。因此，动机选择的是代价高昂的行动，而非有信息价值的实验：一个社会可能停止学习，要么因为没人尝试，要么因为每个人都过于努力。
英文摘要
Success under a demanding method does not show that its additional demands are worthwhile. I study this distinction in a model of intergenerational learning with hidden effort. Moderate effort is informative about the value of a less demanding method; high effort produces more successes but no information about that question. I first characterize efficient experimentation with a complete record. I then show that parents who underweight descendants' effort costs can induce earlier retirement from experimentation. On an explicit region of priors, a paternalistic equilibrium stops after one failed experiment, whereas an aligned equilibrium allows two. Both transmit the relevant record truthfully and withstand communication deviations: every sender already obtains its preferred full-record continuation. Nevertheless, paternalism lowers discounted material welfare, raises long-run effort and success, and leaves less informative public evidence. Under stronger paternalism, parents may instead conceal knowledge to induce experimentation. I characterize a pooling continuation, establish its local communication stability, and show that it can occur in a full equilibrium. The results distinguish the efficient limits to learning from the additional loss of evidence caused by disagreement over effort costs.
Comments30 pages, 1 figure; substantially revised. Exact verification code is included as ancillary files
