Models of Uncertainty Due to Lack of Knowledge

这项研究将调查人们在只有不完全知识的情况下如何判断陈述的真实性。从商业到公共政策再到个人事务，这个问题经常出现。例如，情报分析员可能会被要求对鲍里斯·叶利钦在未来6个月内是否会继续在俄罗斯执政作出概率判断，环境保护署官员可能不得不根据对某种物质对健康的影响所知不多的情况为其设定污染标准，或者医生可能不得不根据不确定的医疗数据选择治疗方案。这项工作将主要集中在判断和反应选择的认知过程。该研究将在前人工作的基础上发展出可以表达为数学模型的心理学理论，并对这些理论和模型进行实证检验，并对其进行相应的修正。这些模型将提供一个框架，用于区分一个人对一个命题是真的主观信心程度，以及一个人对另一个人做出或提供的公开决定或反应。一个人的自信程度取决于他对主题领域的了解程度，以及如何组织和搜索这些知识，而反应选择取决于动机和相关因素。本研究将探讨的一个问题是某些文献记载的现象之间的关系。一种是认为陈述是真实的隐性偏见。另一种是相对于一个人的隐蔽判断，认为陈述为真（当回应选项被限制为真或假时）的额外偏见，但不是分配陈述为真概率的偏见（当回应选项是从0到100的数字时）。另一个问题是，一个人搜索知识库的策略如何取决于他或她所面临的判断任务的具体形式和顺序。该研究将调查的其他问题包括判断过去和未来事件的差异，跨知识领域对陈述真实性的信心比较，判断和回应的回报效应，知识的程度和组织对判断的影响。这项工作的长期目标是提高判断的质量。应该有可能将在这项工作中发展的判断的数学模型与派生的性能测量相结合，以提高准确性或以一种可以施加适当修正的方式对其进行校准。

This research will investigate how people make judgments about the truth of statements when they have only incomplete knowledge. The problem arises very frequently in areas ranging from business to public policy to personal matters. As examples, an intelligence analyst may be asked for a probability judgment of whether Boris Yeltsin will remain in power in Russia for the next six months, an EPA official may have to set a pollution standard for a substance based on sparse knowledge of its health effects, or a physician may have to select a treatment plan given inconclusive medical data. The work will focus primarily on the cognitive processes of judgment and response selection. The research will develop psychological theories, based on prior work, that can be expressed as mathematical models, test these theories and models empirically, and modify them accordingly. The models will provide a framework for distinguishing the subjective degree of confidence one has that a proposition is true from the overt decision or response one makes or provides to someone else. One's degree of confidence depends on what one knows about the topic area, as well as on how that knowledge is organized and searched, whereas the response selection depends on motivational and related factors. One question this research will investigate concerns the relationships among certain documented phenomena. One is a covert bias to think statements are true. The other is an additional bias to call statements true relative to one's covert judgment (when the response options are limited to true or false), but not a bias to assign probabilities that statements are true (when the response options are numbers from 0 to 100). Another question concerns how a person's strategy for searching his or her knowledge base depends on the specific form and sequence of the judgment tasks that he or she faces. Additional issues the research will investigate include differences in judging past versus future events, the comparison of confidence in the truth of statements across knowledge domains, payoff effects on judgment and responding, and effects of degree and organization of knowledge on judgment. The long-range goal of the work is to improve the quality of judgments. It should become possible to combine the mathematical models of judgment to be developed in this work with derived measures of performance either to improve accuracy or to calibrate it in a manner that suitable corrections can be imposed.