Synthesis produced by the Fondation Descartes of the following research paper:
Allcott, H., Gentzkow, M., & Yu, C. (2019). Trends in the diffusion of misinformation on social media. Research & Politics, 6(2), 2053168019848554.
Since 2016, interactions with fake news have been going down on Facebook, whereas they have been going up on Twitter. Overall, Facebook still remains a key network for the diffusion of misinformation, much more so than Twitter.
Facebook and Twitter, two of the main social networks, have developed tools to improve their algorithms faced with the rise in misinformation since the 2016 US presidential election.
Facebook’s moves to include fact-checking and labels to warn users have been criticised for underperforming and being inadequate. Many commentators even believe that fact-checking in itself is too ineffective and that fake news is impossible to prevent.
This article aims to factually document the presence of misinformation on Facebook and Twitter from 2015 to July 2018.
Based on five different lists (three from fact-checking sites, two from scientific studies on fake news), the authors collected 569 fake news sites (after excluding some due to lack of data). This list is not exhaustive, the total number of online and social media fake news stories is very likely underestimated in this study.
9,540 URLs were extracted in the end: each URL links to a fake news story on one of the sites from the list. Each URL was manually checked to make sure that it did indeed contain a false story.
It is important to note that in order to measure the engagement with fake news on Twitter and Facebook, the authors used the tool Buzzsumo, which does not count the number of views (like on YouTube, for example). The authors define fake news interactions on Facebook by calculating the total engagements for each fake news site based on reactions (likes, etc.), shares and comments: the more reactions, the higher the number of engagements. On Twitter, the number of engagements is based on the number of shares. Once again, the results are very likely to have underestimated true exposure to fake news.
With the assistance of the online traffic measuring tool Alexa, the authors made a list of 170 sites, split between three categories (major news sites; small news sites; major business and culture news sites). This list provides a comparison between the popularity of fake news and other types of news.
On Facebook, engagements with fake news were highest between the end of 2016 and the beginning of 2017. They then dropped sharply – going from around 160 million engagements at the peak at the end of 2016 to around 60 million in July 2018 (a drop of more than 60%). By comparison, major news sites had 200 million engagements (three times more) in July 2018.
On Twitter, the number of shares greatly increased between 2015 and July 2018. It went from less than 2 million shares to around 5 million shares. By comparison, major news sites got around 25 million shares (five times more) on Twitter in July 2018.
To better understand the results, graphs are available in the open access article.
The two main takeaway messages from this study are the following:
It is possible, though more studies need to be undertaken on the subject, that Facebook’s anti-misinformation policy has led to fake news reducing in visibility. For Twitter, it is difficult to explain the growing trend of fake news on this site. The authors do not give any hypothesis and warn that comparisons should be made with caution, because the number of engagements were not counted in the same way on the two social networks.