The issues raised by data gathering on Twitter due to the opacity of its moderation
Abstract
My thesis deals with the perceptions of migrants and migration visible on Twitter.
According to an EU project (Bricks, 2016), we live in a context of increasing hate speech targeting minorities and migrants. Migrants are made responsible for the economic crisis, insecurity, and they serve the criticism of political elites. These latter are accused of favoring migrants' interests at the expense of nationals.
Ordinary users are not the only ones to express hatred, to share fake news, conspiracy theories, to spread misinformation or desinformation. Politicians and polemicists are also arousing the fears around migration for electoral or economic purposes.
In this context, social media are used as an echo chamber to relay controversial and polemic opinions. Twitter is particularly blamed for hosting verbal violence and having a lax moderation. Several events: the terrorist attacks of 2015, Cambridge Analytica, disinformation about Covid-19 and mail-in ballots, have evidenced the harmful consequences the online sphere could have on the offline one. As a result, states like Germany and France, and Europe, try to regulate social media and exert pressure to make them adopt and implement stricter moderation policies. Accordingly, Twitter has toughened its policies and added new labels to fight hateful or deceptive content.
Although Twitter's moderation policies are detailed on the platform, the material and human means dedicated to moderation and its effectiveness, remain unclear. This opacity constitutes an obstacle to the analysis of hate speech on Twitter. Indeed, the researcher has no means to know if sensitive content has been filtered and if the data he gathers is representative of the users' generated content. This lack of transparency prevents the evaluation of the missing data, as well as the role of moderation in the circulation and visibility of hate speech.
My presentation aims at questioning the following points: at this stage, what do we know about Twitter moderation and what remains opaque? Where is research headed on the subject of social media moderation? What are the leads to better comprehend the possible bias of moderation in social media data gathering?
I had the opportunity to have an interview with a Youtube moderator, which gives the
beginnings of an answer to these questions. I hope this workshop will provide the opportunity to discuss the relevance of coupling discourse analysis method with interviews of moderators to take stock of the moderators' interference in data gathering.