Collaborative Research: Computational Methods for the Analysis of Political Event Data

合作研究：政治事件数据分析的计算方法

• 批准号: 0455158
• 负责人: Philip Schrodt
• 金额: $ 7.69万
• 依托单位: University of Kansas Center for Research Inc
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2005
• 资助国家: 美国
• 起止时间: 2005-02-01 至 2006-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0455158&HistoricalAwards=false
• 关键词: Collaborative; Research; Computational; Methods; Analysis

Formal models of political behavior have generally followed the lead of the natural sciences and focused on methods that use continuous-variable mathematics. Stephen Wolfram has recently produced an extended critique of that approach in the natural sciences, and suggested that a great deal of natural behavior can be accounted for using rules that involve discrete patterns. Given the similarity between many of the models used in the natural and social sciences, Wolfram's critique can readily be applied to models of social behavior. Pattern-based models are especially relevant to modeling human behavior because human cognitive abilities are far more developed in the domain of pattern recognition than in the domain of continuous-variable mathematics. International event data---categorical data on who did what to whom at what time---are one of the most common forms of information available for the analysis of international behavior. Contemporary automated methods combined with inexpensive full-text news sources such as FBIS, NEXIS, and Factiva allow event data to be generated in near-real-time with relatively little effort. For example, it is now possible for a graduate student to generate a customized data set containing tens of thousands of events as part of a dissertation. However, consistent with Wolfram's critique, the methods for analyzing, or even visualizing, event data have generally not kept up with the advances in the generation of the data, largely because categorical time series have no analogue in other quantitatively-oriented social sciences such as economics and demography. In these circumstances, the nearly universal tendency has been to convert the data to interval-level measures using scales prior to analyzing it. This has at least three disadvantages: first, the scales are somewhat arbitrary; second information is lost when multiple events .cancel out. and third, most psychological evidence indicates that decision-makers respond to patterns of discrete events rather than numerical aggregations. This project develops and test several pattern- and rule-based analytical tools for the analysis of event data. These involve the development of visualization methods, methods for specifying patterns and rules at various levels of complexity, methods for statistically assessing and comparing patterns and rules, and methods for rule induction. These tools will be deployed on a publicly-accessible dedicated web server. That site will also provide a large number of event data sets in an easily accessible form with basic downloading and subsetting tools. All of the software developed on the project will be open-source under the GNU General Public License; the event data sets and analytical tools will be made available as they are produced rather than being embargoed. The project will experiment with patterns applied to the Israel-Palestinian and Israel-Lebanon conflicts for 1979-2005 and the conflict in the former Yugoslavia for 1991-2001. It will start with the classic tit-for-tat pattern, consider patterns designed to detect attempts at de-escalation, and finally consider complex "meta-rules" that look at the relationship between prior conflict and the propensity of the actors to engage in reciprocal behavior. The principal investigators anticipate conducting half-day workshops at the APSA and ISA to introduce researchers to the use of these tools. Broader Impacts: First, it will utilize undergraduate students, typically honors students in international studies, extensively in the research effort. Both Kansas and Utah are geographical regions under-represented in scientific research so these opportunities are particularly critical. Second, the project will produce new data and analytical tools that can be used by the research community. Third, this approach is likely to be very attractive for the policy community and may be used for forecasting and other policy-relevant tasks.

政治行为的正式模型通常遵循自然科学的领导，并专注于使用连续变量数学的方法。 斯蒂芬·沃尔夫勒姆（Stephen Wolfram）最近对自然科学中的这种方法进行了扩展性的批评，并提出大量的自然行为可以使用涉及离散模式的规则来解释。 鉴于自然科学和社会科学中使用的许多模型之间的相似性，Wolfram的批评可以很容易地应用于社会行为模型。 基于模式的模型与建模人类行为特别相关，因为人类的认知能力在模式识别领域比在连续变量数学领域发展得更快。 国际事件数据--关于谁在什么时间对谁做了什么的分类数据--是分析国际行为的最常见的信息形式之一。当代的自动化方法与FBIS、NEXIS和Factiva等廉价的全文新闻源相结合，可以以相对较少的工作量近实时地生成事件数据。 例如，现在研究生可以生成包含数万个事件的定制数据集作为论文的一部分。 然而，与Wolfram的批评一致，分析甚至可视化事件数据的方法通常没有跟上数据生成的进步，主要是因为分类时间序列在其他定量导向的社会科学中没有类似物，如经济学和人口学。 在这些情况下，几乎普遍的趋势是在分析数据之前使用尺度将数据转换为区间水平的度量，这至少有三个缺点：第一，尺度有点随意;第二，当多个事件相互抵消时，信息会丢失。第三，大多数心理学证据表明，决策者对离散事件的模式而不是数字集合做出反应。该项目开发和测试了几个基于模式和规则的分析工具，用于分析事件数据。 这些涉及可视化方法的发展，在各种复杂程度上指定模式和规则的方法，统计评估和比较模式和规则的方法，以及规则归纳的方法。 这些工具将部署在一个可公开访问的专用网络服务器上。 该网址还将以易于获取的形式提供大量事件数据集，并提供基本的下载和子集工具。 该项目开发的所有软件都将在GNU通用公共许可证下开放源码;活动数据集和分析工具将在制作时提供，而不是被封锁。 该项目将试验1979-2005年以色列-巴勒斯坦和以色列-黎巴嫩冲突以及1991-2001年前南斯拉夫冲突的模式。 它将从经典的达特模式开始，考虑旨在检测降级尝试的模式，最后考虑复杂的“元规则”，这些规则着眼于先前冲突与行为者参与互惠行为的倾向之间的关系。行为。主要研究人员预计将在APSA和伊萨举办为期半天的研讨会，向研究人员介绍这些工具的使用。更广泛的影响：首先，它将利用本科生，通常是国际研究的荣誉学生，广泛参与研究工作。 堪萨斯和犹他州都是科学研究代表性不足的地理区域，因此这些机会特别重要。 第二，该项目将产生可供研究界使用的新数据和分析工具。 第三，这一方法可能对政策界很有吸引力，可用于预测和其他与政策有关的任务。