从自利专家那里获得预测:决策者的评分规则
Eliciting Forecasts from Self-interested Experts: Scoring Rules for Decision Makers
摘要:
用于引出随机变量的专家预测的评分规则通常是假设专家仅从其预测的质量中获得效用(例如,由规则奖励的分数,或预测市场中的收益)。我们研究一个更现实的设置,其中(a)的主要是一个决策者,并会根据专家的预测作出决定;(b)专家有一个内在的利益的决定。例如,在公司决策市场中,专家可能从其经理采取的行动中获得不同程度的效用。因此,如果使用典型的评分规则,专家通常会有动机误报她的预测,以影响决策者的选择。我们开发了一个通用的模型,这种设置,并介绍了补偿规则的概念。当与专家的决策的内在效用相结合时,补偿规则诱导了表现得像正常评分规则的净评分规则。假设充分了解专家效用,我们提供了一个完整的表征所有(严格)适当的补偿规则。然后,我们分析的情况下,专家的效用函数是不完全知道的决策者。我们展示了以下方面的界限:(a)专家误报的动机;(b)专家误报的程度;(c)决策者因这种不确定性而失去效用。这些界限以自然的方式取决于不确定性的程度,净评分函数的局部凸度,以及决策者效用函数的自然属性。他们还提出了优化程序的补偿规则的设计。最后,我们简要地讨论了使用补偿规则作为市场评分规则的自利专家在预测市场。
英文摘要:
Scoring rules for eliciting expert predictions of random variables are usually developed assuming that experts derive utility only from the quality of their predictions (e.g., score awarded by the rule, or payoff in a prediction market). We study a more realistic setting in which (a) the principal is a decision maker and will take a decision based on the expert's prediction; and (b) the expert has an inherent interest in the decision. For example, in a corporate decision market, the expert may derive different levels of utility from the actions taken by her manager. As a consequence the expert will usually have an incentive to misreport her forecast to influence the choice of the decision maker if typical scoring rules are used. We develop a general model for this setting and introduce the concept of a compensation rule. When combined with the expert's inherent utility for decisions, a compensation rule induces a net scoring rule that behaves like a normal scoring rule. Assuming full knowledge of expert utility, we provide a complete characterization of all (strictly) proper compensation rules. We then analyze the situation where the expert's utility function is not fully known to the decision maker. We show bounds on: (a) expert incentive to misreport; (b) the degree to which an expert will misreport; and (c) decision maker loss in utility due to such uncertainty. These bounds depend in natural ways on the degree of uncertainty, the local degree of convexity of net scoring function, and natural properties of the decision maker's utility function. They also suggest optimization procedures for the design of compensation rules. Finally, we briefly discuss the use of compensation rules as market scoring rules for self-interested experts in a prediction market.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。