RAPID: Detecting and Deterring Harmful Online Speech Directed at American Election Officials

在2020年的选举周期中，美国的选举官员受到人身攻击、仇恨言论，在某些情况下还受到身体威胁。其中许多攻击都是在社交媒体上发起的，这可能放大了它们的毒性和影响范围。该项目使用机器学习和人工智能方法筛选社交媒体帖子，以识别针对选举和其他公职人员的仇恨言论和在线攻击。该项目的新奇之处在于，使用机器学习和人工智能方法，近乎实时地发现针对公职人员的攻击和仇恨言论，并研究这些攻击和有毒言论的来源网络。该项目的更广泛的意义和重要性在于，当这些攻击针对管理选举的公职人员时，开发出快速准确地检测社交媒体对话中的毒性的方法。该项目对计算机科学和社会科学做出了贡献。在社交媒体数据串流中识别针对特定公职人员的网络攻击很复杂。该项目的第一个贡献是构建工具，以确定要收集和分析的流传输社交媒体数据的域，因为有数千名州、县和地方选举官员。该项目的第二个贡献是使用人工智能方法，这些方法自适应地寻找仇恨言论和在线攻击，特别是在对手试图通过改变在线身份和他们使用的语言来避免被发现的情况下。从实质上说，该项目有助于加强对社交媒体错误信息和骚扰活动的研究，特别是查明攻击选举官员工作并使其失去合法性的有组织的努力。最后，该项目的软件和代码将可供其他研究人员使用。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

In the 2020 election cycle, election officials in the United States were subjected to personal attacks, hate speech, and in some instances physical threats. Many of these attacks were launched on social media, which may have amplified their toxicity and reach. This project uses machine learning and artificial intelligence methods to sift through social media posts to identify hate speech and online attacks directed at election and other public officials. The project’s novelties are the use of machine learning and artificial intelligence methods to find attacks and hate speech directed at public officials in near real-time, and in the study of the networks where these attacks and toxic speech originate. The project’s broader significance and importance are the development of methodologies that detect toxicity in social media conversations quickly and accurately, when these attacks are directed at public officials administering elections.The project contributes to computer science and social science. Identifying online attacks directed at specific public officials in streaming social media data is complex. The first contribution of the project is building tools to determine the domain of streaming social media data to collect and analyze, as there are thousands of state, county, and local election officials. The second contribution of this project is the use of artificial intelligence methods that adaptively seek hate speech and online attacks, in particular in situations where the adversaries try to avoid detection by shifting online identities and the language they use. Substantively, the project contributes to the growing research on social media misinformation and harassment activities, in particular the identification of organized efforts to attack and delegitimize the work of election officials. Finally, the project’s software and code will be available for other researchers to use.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.