This study determined the antecedents of diffusion scope (total audience), speed (number of adopters/time), and shape (broadcast vs. person-to-person transmission) for true vs. fake news about a falsely claimed stolen 2020 US Presidential election across clusters of users that responded to one another’s tweets (“user clusters”). We examined 31,128 tweets with links to fake vs. true news by 20,179 users to identify 1,069 user clusters via clustering analysis. We tested whether attributes of authors (experience, followers, following, total tweets), time (date), or tweets (link to fake [vs. true] news, retweets) affected diffusion scope, speed, or shape, across user clusters via multilevel diffusion analysis. These tweets showed no overall diffusion pattern; instead, specific explanatory variables determined their scope, speed, and shape. Compared to true news tweets, fake news tweets started earlier and showed greater broadcast influence (greater diffusion speed), scope, and person-to-person influence. Authors with more experience and smaller user clusters both showed greater speed but less scope and less person-to-person influence. Likewise, later tweets showed slightly more broadcast influence, less scope, and more person-to-person influence. By contrast, users with more followers showed less broadcast influence but greater scope and slightly more person-to-person influence. These results highlight the earlier instances of fake news and the greater diffusion speed of fake news in smaller user clusters and by users with fewer followers, so they suggest that monitors can detect fake news earlier by focusing on earlier tweets, smaller user clusters, and users with fewer followers.

中文翻译：

---

Twitter 上的选举欺诈和错误信息：作者、集群和消息前因

这项研究确定了关于虚假声称被盗的 2020 年美国总统大选的真假新闻的传播范围（总受众）、速度（采用者数量/时间）和形状（广播与人际传播）的前因跨对彼此的推文做出响应的用户集群（“用户集群”）。我们检查了 20,179 名用户的 31,128 条带有虚假新闻与真实新闻链接的推文，通过聚类分析识别了 1,069 个用户集群。我们测试了作者属性（经验、追随者、关注、总推文）、时间（日期）或推文（链接到虚假 [vs. 真实] 新闻、转发）是否影响跨用户集群的传播范围、速度或形状多级扩散分析。这些推文没有显示出整体的传播模式；相反，具体的解释变量决定了它们的范围、速度和形状。与真新闻推文相比，假新闻推文起步较早，传播影响力（传播速度更快）、范围和人与人之间的影响力更大。具有更多经验和较小用户群的作者都表现出更快的速度，但范围更小，人与人之间的影响更小。同样，后来的推文显示出更多的广播影响力、更小的范围和更多的人与人之间的影响力。相比之下，拥有更多关注者的用户表现出的广播影响力较小，但范围更大，人与人之间的影响力略大。这些结果突出了较早的假新闻实例以及较小用户群和关注者较少的用户中假新闻的更快传播速度，因此他们建议监控器可以通过关注较早的推文、较小的用户群和用户来更早地检测假新闻追随者较少。