This article uses survey experiments to assess whether administrative procedures fix cognitive bias. We focus on two procedural requirements: qualitative reason-giving and quantitative cost-benefit analysis (“CBA”). Both requirements are now firmly entrenched in U.S. federal regulation-making. Multilateral organizations such as the World Bank, OECD, and EU have encouraged their broad diffusion across many national contexts. Yet CBA, in particular, remains controversial. Supporters of CBA claim it leads to more rational regulation, with Sunstein (2000) explicitly proposing that CBA can reduce cognitive biases. By contrast, we argue that procedures should be conceptualized as imperfect substitutes subject to diminishing marginal benefits. To test and illustrate this argument, we examine how each procedure individually and cumulatively modulates the effects of gain-loss framing, partisan motivated reasoning, and scope insensitivity in a nationally representative sample. We find that one or both procedures decrease each cognitive bias. CBA is most helpful against partisan reasoning, where reason-giving does little. Both procedures are comparably effective for combatting the other biases, although in each case only one procedure produces cognitive benefits distinguishable from zero. We only find substantial synergies between the two procedures with respect to gain-loss framing. Layering on the less-useful procedure does not significantly reduce the other two cognitive biases. We hypothesize that procedures will only fix cognitive biases if they disrupt bias-inducing mental processes, and we reconcile this proposition with our findings. We conclude by relating this work to debates about the design of administrative procedures and describe a research agenda based upon rationality-improving procedures.

中文翻译：

---

行政程序能解决认知偏差吗？

本文使用调查实验来评估行政程序是否修复了认知偏差。我们关注两个程序要求：定性给出原因和定量成本效益分析（“CBA”）。这两项要求现已牢牢地纳入美国联邦法规制定中。世界银行、经合组织和欧盟等多边组织鼓励它们在许多国家背景下广泛传播。然而，尤其是 CBA，仍然存在争议。CBA 的支持者声称它会导致更合理的监管，Sunstein (2000) 明确提出 CBA 可以减少认知偏差。相比之下，我们认为程序应该被概念化为受边际收益递减影响的不完美替代品。为了检验和说明这个论点，我们研究了每个程序如何在具有全国代表性的样本中单独和累积地调节得失框架、党派动机推理和范围不敏感的影响。我们发现一个或两个程序减少了每个认知偏差。CBA 最有助于反对党派推理，在这种情况下，给出理由的作用很小。这两种程序在对抗其他偏见方面都相当有效，尽管在每种情况下，只有一种程序产生的认知益处与零不同。我们只发现这两个程序在增益-损失框架方面存在实质性的协同作用。在不太有用的程序上分层并不会显着减少其他两种认知偏差。我们假设程序只有在扰乱导致偏见的心理过程时才能修复认知偏见，我们将这一命题与我们的发现相协调。最后，我们将这项工作与关于行政程序设计的辩论联系起来，并描述了一个基于合理性改进程序的研究议程。