Combining Prescriptive and Descriptive Decision Theory for Counter-terrorism Decisions

在这个项目中，首席研究员将为多个决策者开发决策分析方法，规定一个决策者的决策，同时对其他决策者的决策进行描述性建模。决策分析和博弈论已被广泛用于协助制定反恐决策。最近，已经提出了一些方法，允许在一个决策模型中处理防御者决策、攻击者决策和不确定性。然而，这些组合方法假设攻击者和防御者都是理性的决策者。理性假设适用于防守方，因为决策分析师是在一种规定性模式下工作，以帮助防守方做出理性的最优决策。描述性决策理论关注的是真实的人实际上是如何思考和行为的。评判这些模型的标准是它们在预测人们做出的实际选择和捕捉人们偏离理性、规范理想的许多方式方面的有效性。这些理论更适合于为攻击者的决策建模。这项研究将需要开发算法，这些算法可以使用针对防御者的说明性理论和针对攻击者的描述性理论来解决具有多个决策者的复杂影响图。我们还将开发用于多属性决策的描述性决策模型，并将其纳入规范/描述性方法中。就更广泛的影响而言，本研究将通过纳入更现实的攻击者行为模型，同时仍然提供防御者的最优、合理的规定行动，从而使国土安全领域受益。它也将有利于一般决策理论。该方法适用于有多个决策者的所有决策情况，分析员试图在不确定的情况下帮助其中一个决策者做出更好的决策，同时对其他决策者所做的决策进行建模。这种情况发生在竞争性议价和谈判、市场营销和战略以及其他一般商业决策中。此外，虽然这项研究的重点是帮助一方在面对另一方的描述性模型时做出更好的决策，但这种方法有可能被用于描述性科学，如经济学或政治学，在这些科学中，研究人员认为游戏中的一方最好被建模为理性，而另一方最好被建模为有限理性。这项研究的结果将传播给规定性和描述性决策研究人员，以表明这些方法的真正力量是在它们结合在一起时发现的。我们将开发iPad软件，通过直观的触摸界面图形化地构建和分析这些模型。这可能会大大增加接触到这项研究及其整合的方法的人数。

In this project, the Principal Investigator will develop decision analytic methods for multiple decision makers that prescribe one decision maker's decisions while descriptively modeling the decisions of the other decision makers. Decision analysis and game theory have been used extensively to assist in making counter-terrorism decisions. Recently, methods have been proposed to allow treatment of defender decisions, attacker decisions, and uncertainties in one decision model. However, these combined approaches assume that both the attacker and defender are rational decision makers. The rationality assumption is appropriate for the defender as the decision analyst is working in a prescriptive mode to help the defender make a rational and optimal decision. Descriptive decision theory focuses on how real people actually think and behave. Such models are judged by their validity in predicting the actual choices people make and capture many of the ways in which people deviate from the rational, prescriptive ideal. These theories are more appropriate for modeling the attacker's decision. This research will require the development of algorithms that can solve complex influence diagrams with multiple decision makers using prescriptive theories for the defender and descriptive theories for the attacker. We will also develop descriptive decision models for multi-attribute decisions and incorporate them in the prescriptive/descriptive approach. In terms of broader impacts, this research will benefit the field of homeland security by incorporating more realistic models of attacker behavior while still providing the optimal, rational prescribed action of the defender. It will also benefit general decision theory. The approach is applicable to all decision situations where there are multiple decision makers and an analyst is attempting to prescriptively help one of them make better decisions under uncertainty while modeling the decisions made by the other decision makers. Such situations occur in competitive bargaining and negotiation, marketing and strategy, as well as other general business decisions. Furthermore, while the focus of this research is to prescriptively help one side make better decisions faced with a descriptive model of the other side, there is the potential for this approach to be used in a descriptive science such as economics or political science, where the researcher believes one player in a game is best modeled as rational while the other is best modeled as boundedly rational. The results of this research will be disseminated to both prescriptive and descriptive decision researchers to show that the true power of these methods is found when they are brought together. We will develop iPad software to construct and analyze such models graphically through an intuitive touch interface. This could significantly increase the number of people exposed to this research and the methodologies it integrates.