In recent years, content moderation on social media has become a flashpoint in global conversations about free speech, misinformation, and platform accountability. Once seen as essential for curbing harmful content, from hate speech to false information, moderation practices are now under scrutiny. Some social media platforms, such as Alphabet, X and Meta, are stepping back from stringent oversight, citing challenges in balancing free expression and user safety, with political and legal pressures. This shift raises critical questions about the future of online discourse, the role of tech companies in shaping it, and what this means for users navigating what are still essentially unregulated digital spaces.
Complicating the issue is the uncertainty surrounding the efficacy of “soft” content moderation measures, such as fact-checked footnotes and blur filters. While these measures aim to inform users without outright censoring content, their impact is debated. Critics argue that these measures often go unnoticed or are dismissed by users, limiting their effectiveness in curbing the spread of misinformation. Others contend that they strike a more balanced approach to moderation, promoting awareness rather than restriction. This uncertainty further fuels the debate over whether social media platforms should continue to invest in such measures or move away from them altogether.
In a new study titled “To Share or Not to Share: Randomized Controlled Study of Misinformation Warning Labels on Social Media,” researchers Anatoliy Gruzd, Philip Mai, and Felipe Soares investigated whether soft moderation techniques can decrease engagement with misinformation. The study won the Best Paper Award at the 6th Multidisciplinary International Symposium on Disinformation in Open Online Media (MISDOOM 2024), held in Münster, Germany, from September 2–4, 2024, and was published by Springer in the conference preferred proceedings.
The study simulated a Facebook-like interface using ModSimulator, an open-source mock social media tool designed to test the effectiveness of two common soft moderation techniques—footnote warning labels and blur filters—in reducing the spread of misinformation and informing users about the accuracy of posts. In this controlled experiment, 1,500 users were exposed to false claims about the Russia-Ukraine war in an ecologically valid environment. The results showed that both interventions reduced engagement with posts containing false information, with no significant difference between the two approaches.
On average, the control group interacted 0.9 times with false claims, 1.5 more than either of the two intervention groups (0.62 for the footnote warning label group and 0.60 for the blur filter group).
Additionally, the study found that people who believed in pro-Kremlin claims and trusted partisan sources were more likely to engage with misinformation. In contrast, those who trusted fact-checking organizations and frequently commented on Facebook were less likely to engage.
These findings underscore the effectiveness of soft moderation interventions. It also shows that the less restrictive intervention of adding a warning footnote to a fact-checked post produced a similar response as a more restrictive blur filter in front of a fact-checked image or video.
However, since other factors, such as users’ prior beliefs and political leanings, impact their decision to engage with misinformation on social media, soft moderation should be coupled with other approaches, such as pre-bunking and information literacy campaigns.
See the paper for more details, which is available as open-access.
Citation:
Gruzd, A., Mai, P., Soares, F.B. (2024). To Share or Not to Share: Randomized Controlled Study of Misinformation Warning Labels on Social Media. In: Preuss, M., Leszkiewicz, A., Boucher, JC., Fridman, O., Stampe, L. (eds) Disinformation in Open Online Media. MISDOOM 2024. Lecture Notes in Computer Science, vol 15175. Springer, Cham. https://doi.org/10.1007/978-3-031-71210-4_4