Collaborative Research: How online foci shape conversation

这个项目调查了影响他们谈话质量的在线社区的特征。社交媒体网站并非同质化；它们由在线互动的“焦点”组成——在线社区，如群组、博客和讨论论坛，其特点是稳定和有界。在某些焦点中，对话是友好的和建设性的，在其他焦点中，对话是敌对的和无效的。本研究的目的是了解网络焦点的不同特征——他们的文化、节制政策、对局外人的态度和社会关系——如何影响他们的对话质量。研究结果将推动科学知识的发展，以改善民主社会，并帮助学者和实践者想象和设计一个更具包容性和民主的在线公共领域。这项研究利用了一个大规模的多层次数据集，该数据集涵盖了在一个流行平台上的1058个主题论坛的主要活动期间展开的所有公共对话。这些数据包括12亿条评论、82亿条帖子反应和26亿条评论反应，涉及3亿用户、750万条帖子。此外，该项目收集标记的评论来训练机器学习模型，这些模型根据两个核心维度来衡量和预测评论是否经过审议：对话的文明程度和有意义的跨意识形态互动。这些数据，结合准实验设计分析、社会网络分析、自然语言处理和多层次建模策略等方法学上的最新进展，用于识别促进跨意识形态互动和民间话语的在线焦点特征，并了解在线焦点在缓和分裂性外生社会事件的影响方面所起的作用。本项目由社会学、方法论、测量与统计学、人际网络与数据科学研究、安全与可信网络空间项目联合支持。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 project investigates the characteristics of online communities that shape the quality of their conversations. Social media sites are not homogeneous; they are made up of online “foci” of interaction -- online communities like groups, blogs, and discussion forums characterized by their stability and bounded nature. In certain foci, conversations are friendly and constructive, in others, hostile and unproductive. The goal of this study is to understand how the different characteristics of online foci — their cultures, moderation policies, attitudes towards outsiders, and social relationships — influence the quality of their conversations. Findings will advance scientific knowledge to improve a democratic society and help scholars and practitioners imagine and design a more inclusive and democratic online public sphere.This study leverages a large-scale multilevel dataset that covers all public conversations that unfolded during a major campaign across 1,058 topical forums on a popular platform. The data consist of 1.2 billion comments, 8.2 billion reactions to posts, and 2.6 billion reactions to comments among three hundred million users across 7.5 million posts. In addition, the project collects labeled comments to train machine learning models that measure and predict whether comments are deliberative according to two core dimensions: the civility of conversations and meaningful cross-ideological interactions. These data, in conjunction with recent methodological advances in the analysis of quasi-experimental designs, social network analysis, natural language processing, and multilevel modeling strategies, are used to identify the characteristics of online foci that foster cross-ideological interaction and civil discourse and to understand the role online foci play in moderating the impact of divisive, exogenous social events. This project is supported jointly by the Sociology, Methodology, Measurement, and Statistics, Human Networks and Data Science-Research, and Secure and Trustworthy Cyberspace Programs.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.