Information-Constrained Dynamic Models of Choice Behavior

本项目将探索经济背景下人类选择行为背后的认知过程模型。要探索的模型允许偏离完全知情的最佳选择。他们不会简单地假定行为是“非理性的”，而是将偏离最佳决策的原因解释为（通常是前意识的）选择算法的使用，这些算法非常适合于实现决策者的目标。选择算法也不需要更大的信息处理能力。如果成功，该研究将展示如何将心理学、视觉科学和神经科学等其他学科的研究成果整合到经济分析中，从而对经济理论的发展产生广泛的影响。如果对政策动机的准确认识能改变个人的行为，它也应该对经济政策的设计产生广泛的影响。该项目将探索提出的理论在各种情况下解释经济选择异常的能力。要考虑的应用包括作为观察选择随机性来源的处理错误；对不断变化的市场条件的行为调整的延迟以及对不同类型新闻的不同程度的延迟反应；部分交易者对当前市场状况的不关注导致资产价格过度波动；在“聚焦错觉”中，选择的某些属性对决定选择的影响不成比例；情境对消费者选择的影响，例如那些允许通过向消费者提供的一组额外选择来操纵购买行为的因素；以及与预期效用最大化理论不一致的风险结果方面的选择。所追求的方法不同于先前对这些行为方面的处理，它寻求对它们的更统一的解释，并为这些行为模式提供功能解释，而不仅仅是记录“偏见”。要探索的理论对“理性不注意”理论施加了额外的约束（Sims, 2003, 2010）。将特别注意由于积累了连续的证据而产生的动态决策过程，其中每一项证据对于最佳行动方案只提供很少的信息，并将特别注意在这种动态过程中对选择情况的不同方面的最佳选择。所提出的信息处理约束是涉及感知判断和基于价值的决策的认知过程的描述性（非规范性）模型的共同特征。这些经验观察不仅包括在特定决策问题中做出特定选择或判断的频率，还包括提供导致这些选择的认知过程证据的其他测量，如反应时间和神经激活的测量。拟议的研究将比较基于信息约束的规范模型的预测与用于测试决策的描述性模型的实验证据。

This project will explore models of the cognitive processes underlying human choice behavior in economic contexts. The models to be explored allow for departures from fully-informed optimal choice. Rather than positing that behavior is simply "irrational," they will explain departures from optimal decision-making as resulting from the (generally pre-conscious) use of choice algorithms that are well-suited to mostly achieving decision makers' objectives. The choice algoritms also would not require greater information-processing capacity. If successful, the research will have broad implications for the development of economic theory by showing how research findings in other disciplines, such as psychology, vision science and neuroscience, can be integrated into economic analysis. It should also have broad implications for the design of economic policies, if accurate perception of the incentives of policies changes individuals' behavior.The project will explore the ability of the proposed theory to explain anomalies of economic choice in a variety of contexts. Applications to be considered include processing errors as a source of randomness in observed choices; delays in the adjustment of behavior to changing market conditions and differential degrees of delay in the response to differing types of news; excess volatility of asset prices as a result of inattentiveness on the part of some traders to current market conditions; "focusing illusions" in which certain attributes of alternatives are disproportionately influential in determining choice; context effects in consumer choice, such as those that allow manipulation of purchasing behavior by the addition of additional alternatives to the set with which consumers are presented; and aspects of choice with regard to risky outcomes that are inconsistent with the theory of expected utility maximization. The approach to be pursued differs from previous treatments of these aspects of behavior by seeking a more unified explanation for them, and providing functional explanations for such patterns of behavior rather than merely documenting "biases." The theory to beexplored imposes additional constraints on the theory of "rational inattention" (Sims, 2003, 2010). Particular attention will be given to the dynamic processes of decision-making as a result of accumulation of successive pieces of evidence, each of which is only slightly informative as to best course of action, and to the optimal selection of attention to different aspects of the choice situation in the course of such a dynamic process. The constraints on information processing that are proposed are ones that are common features of descriptive (non-normative) models of cognitive processes involved in both perceptual judgments and value-based decisionmaking. The empirical observations on which these are based include not only the frequencies with which particular choices or judgments are made in particular decision problems, but other measurements that provide evidence about the cognitive processes giving rise to these choices, such as reaction times and measures of neural activation. The proposed research will compare the predictions of normative models based on information constraints with experimental evidence of the kind used to test descriptive models of decision making.