[FR] CSA publishes extensive study on the distribution of fake news on Twitter

IRIS 2021-1:1/33

Amélie Blocman


Under the Act on the fight against the manipulation of information of 22 December 2018, the French national audiovisual regulatory authority (Conseil Supérieur de l’Audiovisuel - CSA) was given new powers in this area. The CSA therefore wished to learn more about the mechanisms used to spread fake news, a problem exacerbated by the COVID-19 pandemic, by commissioning a specific study of the Twitter social network. The data required for the study was collected using Twitter’s open API (interface), which can be used to collect the Tweets posted from an account, statistics specific to each Tweet (the number of ReTweets, likes, etc.) and the identity of accounts that follow or are followed by a given account.

Twitter provided all the information that it considered relevant for the report, especially concerning the efforts it was making in this area. For the purposes of the study, news is considered fake if it has been designated as such by journalists specialising in fact-checking. In the same way, Twitter accounts are categorised as reliable or likely to spread fake news, as determined by certain fact-checkers (in particular the Décodeurs du Monde, as part of the Décodex initiative).

The first part of the study is devoted to the analysis of the least reliable news accounts, which have a much smaller number of Twitter followers than most of the reliable news accounts. Nevertheless, accounts known to spread fake news generate the same number of ReTweets as reliable accounts; this is because followers of unreliable accounts are much more likely to help spread the information posted on these accounts than followers of reliable accounts (posting 10 to 20 times the number of ReTweets per follower). In order to better understand the impact of the virality of these accounts, the study contains a qualitative analysis of their most shared Tweets. This shows that the accounts identified as unreliable mainly focus on topical themes linked to controversial issues (policies on immigration, health, religion, terrorism, etc.).

The first section concludes by analysing the Twitter accounts of fact-checkers affiliated to traditional media. The accounts of journalistic organisations specialising in fact-checking have, on average, more followers than accounts categorised as unreliable, but they generate less interaction than the latter. In terms of the themes most commonly dealt with, only politics and religion are covered by both types of account. Fact-checkers are more interested in topics linked to their profession, such as media-related news or content connected to media and information literacy.

The next part of the study analyses in greater detail certain pre-selected Tweets relating to fake news that was either spread or refuted by the Tweets concerned. All the fake news studied is contained in a heavy concentration of Tweets posted over a very short space of time. This observation underlines the difficulty facing fact-checkers, who have to react very quickly in order to have an impact on the spread of fake news. The analysis of their distribution chronology shows that, contrary to what might be expected or hoped, "real" news does not drive out the fake news. Furthermore, fake news is primarily used to fuel criticism of authorities or to express a form of panic, for example in relation to sensitive health issues. It relies on information that has not been cross-checked or identified as false by journalists who, for their part, post corrections in accordance with codes specific to the social networks. However, these corrections do not trigger such a high level of engagement (ReTweets, comments) as accounts that are used to spread fake news.


This article has been published in IRIS Legal Observations of the European Audiovisual Observatory.