Automated collection and coding of online campaign advertising.

摘要选举活动在我国民主政治中起着极其重要的作用。发送给选民的信息有助于确定选举的内容以及选民最终将如何选择。两项新的广告技术——对个人选民的大量数据的使用和YouTube等新的在线广告平台——有可能彻底改变竞选活动如何能够精心制作特定的信息来吸引个人选民。然而，到目前为止，还没有系统的证据证明通过在线视频广告定位的准确性或有效性。这个项目将是第一个测试竞选活动如何有效地通过在线广告瞄准选民的项目。pi将开发新的计算机科学工具来捕获和编码在线广告的内容。这将使pi能够测试一系列关于目标定位有效性的假设，在线广告中使用的策略，以及竞争对手竞选活动相互反应的能力。这个跨学科团队将向研究界免费提供其工具和数据，培训政治学和计算机科学方面的学生，并创建实时跟踪广告的机制，这些机制将与领先的媒体机构共享。技术摘要：在本提案中，pi将设计一项关于政治运动对受众微目标在线广告有效性的实证研究。他们开发了两套计算工具来测试在线广告微目标的有效性。第一个工具是一组“机器人”，它们具有被操纵的个人资料特征，可以模仿不同类型的选民。这些机器人允许团队跟踪显示给“选民”的广告，并根据机器人的个人资料特征测试差异。研究小组将利用YouTube作为在线平台进行研究，因为YouTube在最近的总统选举中被候选人广泛使用。YouTube拥有超过10亿的用户，每天有超过40亿的视频观看量。超过100万广告主利用谷歌广告平台，该平台为广告主提供了基于性别、年龄、婚姻状况和兴趣等因素选择目标用户的能力，这些因素与谷歌的兴趣类别相吻合。此外，该平台还接受第三方网站投放的广告。为了测试目标定位的有效性，该团队计划开发第二套工具，使他们能够在多个维度上自动对在线广告进行编码。这将使他们能够构建发送给选民的信息的实时度量，并评估有关竞选活动通过在线广告瞄准选民的能力的多种理论。

General AbstractElection campaigns serve an extremely important role in our democracy. The messages sent to voters help to define what elections will be about and how voters will ultimately choose. Two new advertising technologies -- the use of large amounts of data about individual voters and new online advertising platforms such as YouTube -- have the potential to revolutionize how campaigns are able to craft specific messages to appeal to individual voters. To date, however, there is no systematic evidence about the accuracy or effectiveness of targeting via online video advertising. This project will be the first to test how effectively campaigns can target voters with online advertising. The PIs will develop new computer science tools to capture and code the content of online advertising. This will allow the PIs to test a series of hypotheses about the effectiveness of targeting, the strategies used in online advertising, and the abilities of the rival campaigns to respond to one another. The interdisciplinary team will make its tools and data freely available to the research community, train students in both political science and computer science, and create mechanisms for real time tracking of advertising, which will be shared with leading media outlets.Technical AbstractIn this proposal, the PIs set devise an empirical study of the effectiveness of political campaigns to micro-target online advertising to audiences. They develop two sets of computational tools to test the effectiveness of online advertising micro-targeting. The first of tools are a set of 'bots' with manipulated profile characteristics that mimic different types of voters. These bots allow the team to track the advertisements shown to "voters" and test for differences based on the bots' profile characteristics. The team will rely on YouTube as the on-line platform for the study as it is very popular, and was utilized by candidates during the recent presidential election campaigns. YouTube boasts over 1 billion users, and over 4 billion video views a day. More than a million advertisers utilize the Google ad platform that provides advertisers ability to select target users based on factors such as gender, age, marital status, and interests that coincide with Google's categories of interest. Furthermore, the platform accepts ads delivered from third-party sites. To test hypotheses about the effectiveness of targeting, the team plan to develop a second set of tools that allow them to code automatically the online ads along multiple dimensions. This will allow them to construct real-time measures of the messages sent to voters and to assess multiple theories regarding the ability of campaigns to target voters via online advertising.