“`html
A study conducted by the University of Washington regarding X indicated that entries with Community Notes included were less likely to achieve virality and received reduced engagement. Following the addition of a Community Note, on average, reposts decreased by 46% and likes declined by 44%.iStock
In 2022, subsequent to Elon Musk’s acquisition of what is now X, the organization terminated 80% of its content moderation workforce and established Community Notes as the primary method of fact-checking on the platform. Previously a trial initiative at Twitter, Community Notes enables users to propose appending a comment to a specific entry — generally to provide context or rectify an erroneous fact. If other users with varied perspectives vote that the comment is beneficial, as determined by X’s algorithm, the note is then attached to the post. Other social media sites, including Meta and YouTube, have since adopted similar strategies.
A University of Washington-led examination of X revealed that entries with Community Notes were less susceptible to virality and experienced diminished engagement. After the incorporation of a Community Note, reposts fell by an average of 46% and likes by 44%.
“Our research indicates that Community Notes are effective when attached, particularly in decreasing engagement that signals endorsement for the content, such as reposts and likes,” stated senior author Martin Saveski, an assistant professor at UW’s Information School. “However, the proliferation of misinformation on social media is intricate and multifaceted, necessitating multiple strategies working in unison to effectively mitigate it. Tools like Community Notes are an essential addition to the platforms’ arsenal.”
The research group released its results on Sept. 18 in the Proceedings of the National Academy of Sciences of the United States of America.
Between March and June of 2023, investigators monitored 40,000 posts for which a note was recommended. Out of these, 6,757 notes were considered helpful and were added. The team observed posts for 48 hours post note attachment and contrasted posts with notes to those without on two significant measures: engagement, such as likes and reposts, and diffusion.
Diffusion represents how a post circulates through the social network — essentially its virality. For instance, do only individuals following an account engage with a post?
“We understand from other research that inaccurate information typically disseminates faster, wider, and more virally than accurate information,” explained lead author Isaac Slaughter, a doctoral candidate at UW’s Information School. “Our findings suggest that Community Notes considerably alter the manner in which information propagates through a network. Individuals who are farther away in the social network from the originator of the misinformation are significantly less inclined to engage with the post. Yet those closest to the source — followers, for example — tend to be less influenced by the note.”
On average, the team discovered that after notes were added, engagement descended by 46% for reposts, 44% for likes, 22% for replies, and 14% for views. Over the entire lifespan of posts, incorporating engagement prior to note attachment, the declines were 12% for reposts, 13% for likes, 7% for replies, and 6% for views.
“We believe views were less impacted because what users observe is predominantly determined by X’s feed algorithm,” Saveski stated. “From the public outline of the algorithm, we comprehend that X does not explicitly prioritize or deprioritize posts with notes attached, but this could change in the future.”
The investigation was also able to gather detailed data regarding what influenced the dissemination of posts. Notes added to modified media, such as deceptive images and videos, influenced those entries more than text-based posts. Notes affixed to highly popular posts resulted in greater decreases in engagement. Additionally, swift incorporation of notes was critical.
“Content propagates at a rapid pace across X, and if a note arrives too late, few users will have the opportunity to see it,” Slaughter remarked. “Notes that take approximately 48 hours to appear have nearly no impact.”
Saveski’s laboratory at UW is currently formulating potential instruments to accelerate the speed at which notes can be attached to posts to enhance their effectiveness.
The authors exclusively examined posts that had notes proposed in early 2023, and X has considerably updated its Community Notes procedures since that time. However, it has also terminated free access to its API, rendering further academic investigations impractical. The manuscript also focused solely on X, not on other social media platforms.
“The sustainability of this type of moderation as many distinct systems across various platforms, as it is currently being utilized, remains a significant question,” Saveski noted. “If someone is appending notes on X, does that reduce their likelihood of doing so on TikTok or Instagram? There’s also the issue of how much platforms should collaborate and exchange data, which could facilitate scaling. X has made its code and data accessible, but no other platforms have committed to making their systems available yet.”
Co-authors include Axel Peytavin from Stanford University and Johan Ugander from Yale University. This research was partially supported by a UW Information School Strategic Research Fund grant and an Army Research Office Multidisciplinary University Research Initiative grant.
For additional details, reach out to Saveski at [email protected] and Slaughter at [email protected].
“`