Replication Data for: You’ve been shadowbanned: Has Facebook’s strategy to suppress rather than remove COVID-19 vaccine misinformation actually slowed the spread?

Abstract: In March 2020, shortly after the World Health Organisation declared COVID-19 a global pandemic, Facebook (the company is now rebranded as Meta) announced steps to stop the spread of COVID-19 and vaccine-related misinformation. This entailed identifying and removing false and misleading content that could contribute to “imminent physical harm”. For other types of misinformation the company’s fact-checking network was mobilised and automated moderation systems ramped up to “reduce its distribution”. In this paper we ask how effective this approach has been in stopping the spread of COVID-19 vaccine misinformation in the Australian social media landscape? To address this question we analyse the performance of 18 Australian right-wing and anti-vaccination Facebook pages, posts and commenting sections collected over 2 years until July 2021. We use CrowdTangle’s engagement metrics and time series analysis to map key policy announcements (between Jan 2020 and July 2021) against page performance. This is combined with content analysis of comments parsed from 2 pages, and a selection of posts that continued to overperform during this timeframe. The results showed that the suppression strategy was partially effective, in that the performance of many previously high performing pages showed steady decline between 2019 and 2021. Nonetheless, some pages not only slipped through the net but overperformed, proving this strategy to be light-touch, selective and inconsistent. The content analysis shows that labelling and fact-checking of content and shadowbanning responses were resisted by the user community who employed a range of avoidance tactics to stay engaged on the platform, while also migrating some conversations to less moderated platforms.

摘要：2020年3月，在世界卫生组织（World Health Organisation）将新型冠状病毒肺炎（COVID-19）宣布为全球大流行病后不久，脸书（Facebook，现更名为Meta）宣布采取措施遏制新冠疫情及疫苗相关错误信息的传播。该举措涵盖识别并移除可能引发"迫在眉睫的人身伤害"的虚假及误导性内容；针对其他类型的错误信息，该公司启动了事实核查网络（fact-checking network），并升级自动化审核系统（automated moderation systems）以"降低其传播范围"。本文旨在探究该策略在澳大利亚社交媒体语境中遏制新冠疫苗错误信息传播的实际成效。为解答这一研究问题，我们分析了2019年至2021年7月两年间收集的18个澳大利亚右翼及反疫苗脸书主页、帖子及评论区数据。我们借助CrowdTangle的互动指标与时间序列分析（time series analysis）方法，将2020年1月至2021年7月间的关键政策公告与主页运营表现进行关联映射分析。该分析结合了对2个主页提取评论的内容分析，以及对该时段内持续获得超额互动的精选帖子的分析。研究结果显示，该压制策略仅取得部分成效：此前互动量较高的多数主页在2019年至2021年间呈现稳步下滑趋势。尽管如此，仍有部分主页不仅未被管控，反而获得了更高的互动量，这表明该策略存在管控宽松、选择性强且前后不一致的问题。内容分析结果显示，用户群体对内容标注、事实核查以及影子屏蔽（shadowbanning）等措施存在抵制行为，他们采用了多种规避策略以维持在该平台的活跃度，同时将部分讨论转移至审核强度更低的平台。