Collaborative Research: Models of Unawareness and Ambiguity

这项关于无意识和歧义的研究有三个组成部分。首先，模型是开发，其中隐式信息嵌入在对行为的描述中。例如，如果向消费者提供了一个保证，如果她的轮胎破裂，将提供500美元，她就会意识到这种意外情况是可能发生的。鉴于这种认识，契约构建世界状态的方式就变得重要起来。该模型通过考虑状态框架来补充考虑结果框架的前景理论。模型的选择对象是或有事项和支付的列表。该模型为意识理论提供了两个技术创新。首先，决策者的意识是通过她对标准赌注的偏好来确定的。因此，无意识与可能性和概率的概念联系在一起，而不是与知识和真理联系在一起。这允许她部分意识到自己的无意识。明确的公理方法还允许分析人员根据她的偏好和其他意识水平来预测决策者在给定特定意识水平时的行为。其次，无意识是在整个状态空间的分区上定义的，而不是在特定状态上定义的。这样，决策者就可以知道一个事件，而不知道它的子事件。例如，她可以意识到她的传输可能会失败，而不必列出她的传输中可能会中断的所有特定组件。研究的第二个组成部分制定了歧义的表示，其中决策者对可能是非线性聚集的先验集具有二阶信念。歧义的存在由其二阶信念的传播来模拟，而对歧义的态度则由聚合器的曲率来模拟。该效用的现有特征假设状态空间上的外生产品结构。例如，在最近的一项研究中，Klibanoff、Marinacci和Mukerji假设不确定性的第二个维度超过了第一个维度的先验。这在提供基础方面尤其麻烦，因为先验的概念本身就是一个决定性的理论结构，就像折现因子或阿罗-普拉特测度一样。本项目试图将理论从这种外生结构中解脱出来，从对国家的直接赌注的偏好中识别表征。提出了一个有趣的过程，用于从一阶偏好中识别二阶信念和聚合器，这应该产生一个完整的表征。在研究的最后一部分，设计了一个实验来估计这种模糊表示的参数。它采用了一种新颖的图形选择理论设计，提供了足够的数据来自信地校准参数，而不是简单地检验假设。这样的估计在实证应用中将非常有用。这项研究更广泛的影响是，这两种理论都特别有利于应用。这里发展的无意识模型是在标准主观模型的精神下构建的，以方便其使用。它将标准行为和契约作为其原语。事实上，实际合同通常以或有事项和付款清单的形式传达。歧义的表现允许对歧义的存在和对歧义的厌恶进行单独的比较静态。此外，这些比较静态数据带有标准风险理论中的许多直觉。通过为这种表示提供基础，这些技术可以自信地使用。此外，这里概述的测量设备具有独立的意义，因为它提供了一种从稀疏的“本地”数据中推断歧义规避的通用方法。歧义厌恶的实验校准为识别包含歧义厌恶的其他经验模型提供了有用的估计，就像阿罗-普拉特测量的估计有助于识别包含风险厌恶的经验模型一样。

This study of unawareness and ambiguity has three components. First, a model is development, where implicit information is embedded in the description of an act. For example, if a consumer is offered a warranty that provides $500 in case her tires disintegrate, she is made aware that such a contingency is possible. Given this recognition, the way a contract frames the states of the world becomes important. The model complements prospect theory, which considers the framing of consequences, by considering the framing of states. The choice objects of the model are lists of contingencies and payments. This model provides two technical innovations to the theory of awareness. First, the decision maker's awareness is identified by her preferences over standard bets. Therefore, unawareness is associated with concepts of likelihood and probability rather than with knowledge and truth. This allows for partial awareness of her own unawareness. The explicit axiomatic approach also allows the analyst to predict how the decision maker will behave given a particular level of awareness from her preferences with other levels of awareness. Second, unawareness is defined over partitions of the entire state space, rather than over particular states. Then the decision maker can be aware of an event without being aware of its sub-events. For example, she can be aware that her transmission can fail, without listing all the specific components in her transmission that might break.The second component of the study formulates a representation of ambiguity where the decision maker has a second order belief on the set of priors that is possibly nonlinearly aggregated. The existence of ambiguity is modeled by the spread of her second order belief while the attitude towards ambiguity is modeled by the curvature of the aggregator. Existing characterizations of this utility assume an exogenous product structure on the state space. For example, in a resent study, Klibanoff, Marinacci, and Mukerji assume that the second dimension of uncertainty is over the priors on the first dimension. This is especially troubling for providing f oundations, because the very concept of a prior is a decisional theoretical construct, like a discount factor or an Arrow-Pratt measure. This project attempts to disentangle the theory from such exogenous structure, identifying the representation from preference over direct bets on the states. An interesting procedure is presented for identifying the second order belief and the aggregator from first order preferences, which should yield a full characterization.In the final part of the study, an experiment designed to estimate the parameters of this representation for ambiguity. It uses a novel graphical choice-theoretic design that provides enough data to confidently calibrate parameters, rather than simply test hypotheses. Such an estimate would be eminently useful in empirical applications.The broader impact of this study is that both theories are especially conducive for applications. The model of unawareness developed here is constructed in the spirit of standard subjective models to facilitate its use. It keeps standard acts and contracts as its primitives. In fact, actual contracts are often communicated as lists of contingencies and payments. The representation of ambiguity allows separate comparative statics for the existence and aversion to ambiguity. Moreover, these comparative statics carry much of the intuition in standard risk theory. By providing foundations for such a representation, these techniques can be used confidently. In addition, the measurement device outlined here is of independent interest because it provides a general method to infer ambiguity aversion from a sparse amount of "local" data. The experimental calibration of ambiguity aversion provides a useful estimate to identify other empirical models that incorporate ambiguity aversion, in the same way that an estimate of the Arrow-Pratt measure helps to identify empirical models that incorporate risk aversion.