Multi-Scenario Searches: Implementing Uncertainty Management in Integrated Assessment

人们已经收集了大量的信息来帮助解决与全球气候变化有关的问题，但这些信息还不足以回答一些最基本的问题，例如气候变化问题的严重性以及预防气候变化可能需要多少费用。这些问题的答案基本上是不可预测的，而且取决于许多人不同的价值判断。因此，从事气候变化综合评估的人所面临的挑战是，如何在极端不确定和社会价值高度变化的情况下，最好地利用现有信息。综合评估的方法和模型研究项目将继续开发与探索性建模相关的方法来应对这一挑战。模拟模型将创建一个大型数据库，其中包含可能的未来情景。然后，搜索引擎和使用先进可视化技术的人类决策者将从数据库中提取有用的信息，以区分政策选择。研究设计将寻找稳健的策略，以及那些在广泛的合理未来范围内取得合理成功的策略。贝叶斯决策理论用于捕获模型结构、先验和损失函数中的不确定性。这些方法还将考察基于场景的规划的一些最佳特征，包括结合定量和定性信息的能力，使用多角度来传递和接收有关风险的信息，以及通过使用路标、对冲和塑造行动的灵活策略来管理极端不确定性的语言。计算机搜索和可视化技术的新功能极大地促进了这种方法。本研究将有助于建立综合评估中管理不确定性的一般框架，使用稳健的策略作为处理风险和不确定性的框架，并使用多种场景的数据库搜索作为寻找这些策略和组合不同类型信息的手段。该项目将使研究人员能够推广他们以前工作中开发的方法；实现此应用程序所需的特殊搜索算法；并演示这些方法与利益相关者群体的使用，特别是作为一种手段来考虑对多种社会价值的稳健性。因此，该项目将有助于开发具有广泛用途的方法和分析工具，用于处理不确定性，并在各种其他综合评估框架和模型中组合不同类型的信息。

Tremendous amounts of information have been gathered to assist in addressing questions associated with global climatic change, but this information has not been sufficient for answering some of the most basic questions, such as the seriousness of the climate change problem and how much it might cost to prevent it. Answers to such questions are fundamentally unpredictable and depend on the divergent value judgments of many individuals. The challenge facing those engaged in integrated assessments of climate change therefore is to best make use of the available information in the context of extreme uncertainty and highly variable social values. This Methods and Models for Integrated Assessment research project will continue the development of methods associated with exploratory modeling to address this challenge. Simulation models will create a large database of plausible future scenarios. Search engines and human decision makers employing advanced visualization techniques then will extract from the database information that is useful to distinguish among policy choices. The research design will look for robust strategies and those that are reasonably successful over a wide range of plausible futures. Bayesian decision theory is used to capture uncertainties in model structure, priors, and loss functions. These methods also will examine some of the best features of scenario-based planning, including the ability to combine quantitative and qualitative information, the use of multiple perspectives for transmitting and receiving information about risk, and the language of managing extreme uncertainty through flexible strategies using signposts and hedging and shaping actions. New capabilities in computer search and visualization techniques have greatly facilitated this type of approach. This study will contribute toward a general framework for managing uncertainty in integrated assessment, using robust strategies as a framework for handling risk and uncertainty and database searches of multiple scenarios as a means for finding these strategies and for combining diverse types of information. This project will enable the investigators to generalize the methods developed in their previous work; to implement the special search algorithms required for this application; and to demonstrate the use of these methods with stakeholder groups, in particular as a means to consider robustness against multiple social values. The project therefore will help develop methods and analytic tools with broad utility for treating uncertainty and combining different types of information in a wide variety of other integrated assessment frameworks and models.