University of Notre Dame, Notre Dame, USA.
Pacific Northwest National Laboratory, Richland, USA.
Sci Rep. 2021 Apr 14;11(1):8137. doi: 10.1038/s41598-021-87473-8.
Many authoritarian regimes have taken to censoring internet access in order to stop the spread of misinformation, restrict citizens from discussing certain topics, and prevent mobilization, among other reasons. There are several theories about the effectiveness of censorship. Some suggest that censorship will effectively limit the flow of information, whereas others predict that a backlash will form, resulting in ultimately more discussion about the topic. In this work, we analyze the role of communities and gatekeepers during multiple internet outages in Venezuela in January 2019. First, we measure how critical information (e.g., entities and hashtags) spreads during outages focusing on information recurrence and burstiness within and across language and location communities. We discover that information bursts tend to cross both language and location community boundaries rather than being limited to a single community during several outages. Then we identify users who play central roles and propose a novel method to detect gatekeepers-users who prevent critical information from spreading across communities during outages. We show that bilingual and English-speaking users play more central roles compared to Spanish-speaking users, but users inside and outside Venezuela have similar distribution of centrality. Finally, we measure the differences in social network structure before and after each outage event and discuss its effect on how information spreads. We find that with each outage event social connections tend to get less connected with higher mean shortest path, indicating that the effect of censorship makes it harder for information to spread.
许多威权政权采取审查互联网访问的方式来阻止虚假信息的传播、限制公民讨论某些话题以及防止动员等。关于审查制度的有效性有几种理论。一些人认为审查将有效地限制信息的流动，而另一些人则预测会出现反弹，最终导致更多关于该主题的讨论。在这项工作中，我们分析了 2019 年 1 月委内瑞拉多次互联网中断期间社区和把关人所扮演的角色。首先，我们通过关注信息在语言和位置社区内以及跨社区的重复和突发，衡量了关键信息（例如实体和主题标签）在中断期间的传播方式。我们发现，信息突发往往跨越语言和位置社区边界，而不是在几次中断中仅局限于单个社区。然后，我们确定了发挥核心作用的用户，并提出了一种新方法来检测把关人——即在中断期间防止关键信息在社区间传播的用户。我们发现，与讲西班牙语的用户相比，双语和英语用户比讲西班牙语的用户发挥更核心的作用，但委内瑞拉境内外的用户的中心性分布相似。最后，我们测量了每次中断事件前后的社交网络结构差异，并讨论了它对信息传播方式的影响。我们发现，随着每次中断事件的发生，社交联系的连接性变得越来越差，平均最短路径也越来越长，这表明审查制度的影响使得信息更难传播。
