Impact of Social Reference Cues on Misinformation Sharing on Social Media: Series of Experimental Studies

Abstract

Background Health-related misinformation on social media is a key challenge to effective and timely public health responses. Existing mitigation measures include flagging misinformation or providing links to correct information, but they have not yet targeted social processes. Current approaches focus on increasing scrutiny, providing corrections to misinformation (debunking), or alerting users prospectively about future misinformation (prebunking and inoculation). Here, we provide a test of a complementary strategy that focuses on the social processes inherent in social media use, in particular, social reinforcement, social identity, and injunctive norms. Objective This study aimed to examine whether providing balanced social reference cues (ie, cues that provide information on users sharing and, more importantly, not sharing specific content) in addition to flagging COVID-19–related misinformation leads to reductions in sharing behavior and improvement in overall sharing quality. Methods A total of 3 field experiments were conducted on Twitter’s native social media feed (via a newly developed browser extension). Participants’ feed was augmented to include misleading and control information, resulting in 4 groups: no-information control, Twitter’s own misinformation warning (misinformation flag), social cue only, and combined misinformation flag and social cue. We tracked the content shared or liked by participants. Participants were provided with social information by referencing either their personal network on Twitter or all Twitter users. Results A total of 1424 Twitter users participated in 3 studies (n=824, n=322, and n=278). Across all 3 studies, we found that social cues that reference users’ personal network combined with a misinformation flag reduced the sharing of misleading but not control information and improved overall sharing quality. We show that this improvement could be driven by a change in injunctive social norms (study 2) but not social identity (study 3). Conclusions Social reference cues combined with misinformation flags can significantly and meaningfully reduce the amount of COVID-19–related misinformation shared and improve overall sharing quality. They are a feasible and scalable way to effectively curb the sharing of COVID-19–related misinformation on social media.

Mis- and disinformation in social media: mitigating risks in digital health communication

“…Likes aus dem eigenen sozialen Netzwerk) können das Verhalten, Inhalte zu teilen, zusätzlich verstärken. Analog können aber soziale Informationen, dass im eigenen Netzwerk bestimmte Informationen nicht geteilt oder „gelikt“ werden, auch dazu führen, dass weniger Falschinformationen geteilt werden [ 39 ]. Darüber hinaus können eigentlich positive soziale Motive wie der Wunsch, andere in seinem sozialen Netzwerk auf mögliche Risiken hinzuweisen, dazu führen, dass mehr Falschinformationen verbreitet werden.…”

“…Interventionen auf sozialer Ebene zielen vor allem darauf ab, durch die Veränderung von sozialen Prozessen das Teilen von Falschinformationen zu beeinflussen. So konnte in einer Interventionsstudie [ 39 ] gezeigt werden, dass durch Veränderungen in den sozialen Rückmeldungen Veränderungen im Teilen verursacht wurden. Statt Likes und Shares wurde angezeigt, wie viele Mitglieder des eigenen Netzwerks (Followers) oder des gesamten sozialen Netzwerks eine Falschinformation zwar gesehen, aber nicht gelikt oder weitergeleitet hatten.…”

Others (dis‐)endorse this so it must (not) be true: High relative endorsement increases perceived misinformation veracity but not correction effectiveness

“…Understanding how people appraise both endorsements and disendorsements may also be useful given their potential application as a social‐norm proxy. Specifically, explicit normative information can reduce the negative impacts of misinformation at least temporarily (Jones et al, 2023). However, adding normative information to posts is highly resource intensive and practically unviable at scale.…”