CAREER: Promoting Equal Opportunities through Measurement, Simulation, and Education

这项研究将提高我们利用在线社交媒体上的大规模数据来理解美国对种族不平等的态度及其与社会结构关键维度的相关性的能力。虽然有大量文献研究了社交媒体上的种族偏见，但关于社交媒体用户对种族不平等的存在及其原因的看法的研究却很少。虽然社会科学对这些信仰有很多了解，但该项目将开发新的方法，以前所未有的规模衡量这些信仰，从而对种族意识形态的分布及其与地点、身份和网络结构的关系有新的认识。长期研究目标是扩展关于如何将计算方法与社会理论和大型数据集一起建设性地使用以帮助解决社会不平等问题的知识。这项工作将包括开发和发布工具，使定性学者能够在没有机器学习专业知识的情况下利用类似的数据。此外，本提案旨在通过开发新课程和本科生研究机会来推进计算机科学教育。本项目有四个基本研究目标。首先，它将开发新的混合方法，将现代基于图形和自然语言的机器学习方法与定性分析相结合，以识别来自Twitter和Parler的大型数据集中表达的关于种族不平等的主要信念。其次，它将收集和分析一个创新的新数据集，将社交媒体和有关种族不平等态度的调查数据联系起来，为这两种测量来源之间的差异提供见解。第三，它将把社交媒体用户对种族不平等的看法与他们居住的地方和社交网络结构的数据结合起来，研究地点、身份和网络结构如何与种族意识形态相关联。第四，它将开发一个新的基于主体的模拟框架，该框架将实证研究结果与当代种族不平等理论相结合，可用于评估不同拟议干预措施的优缺点。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This research will advance our ability to draw on large-scale data from online social media to understand attitudes towards racial inequality in the United States and their correlations with critical dimensions of social structure. While a large literature has studied racial bias on social media, there has been little research on what social media users believe about the existence and causes of racial inequality. While much is understood about these beliefs in the social sciences, this project will develop new methods to measure these beliefs at an unprecedented scale, allowing a new understanding of the distribution of racial ideologies and their correlates with place, identity, and network structure. The long-term research goal is to expand knowledge on how computational methods can be constructively used alongside social theory and large datasets to help address social inequality. This work will include the development and release of tools that will allow qualitative scholars to leverage similar data without expertise in machine learning. In addition, this proposal aims to advance computer science education via the development of new curriculum and opportunities for undergraduate research. This project has four fundamental research objectives. First, it will develop novel mixed methods approaches that combine modern graph-based and natural language machine learning methods with qualitative analysis to identify dominant beliefs about racial inequality expressed in large datasets from Twitter and Parler. Second, it will collect and analyze an innovative new dataset linking social media and survey data on attitudes about racial inequality, providing insight into discrepancies between these two sources of measurement. Third, it will combine measurements of social media users' views on racial inequality with data on where they live and the structure of their social networks to study how place, identity, and network structure are correlated with racial ideology. Fourth, it will develop a novel agent-based simulation framework that combines empirical findings with contemporary theory on racial inequality, that can be used to evaluate the strengths and weaknesses of different proposed interventions.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.