Disinformation on Covid in Europe would have a fast track, at least on Facebook. To support this, the latest report by the non-governmental organization Avaaz which has often railed against the management of content by social networks. This time a clear disparity would emerge between the interventions to stop the hoaxes on the pandemic in the United States and those that took place in other countries such as Italy, Spain, Portugal, France. All this, and it should be borne in mind, while the European commissioners V? Ra Jourová and Thierry Breton, respectively Values and transparency and Internal Market, are working on the new guidelines of the code of good practices on disinformation that will be imposed on the giants of the Web.
Task force against fake news, Meloni: “Limit freedom”. Martella: “It is used to fight disinformation, not opinions”
Monica Rubino
“This is not a quantitative but a qualitative analysis,” he says Luca Nicotra, who led Avaaz’s investigation. “And the results are similar to what we saw last year, so the progress has been relative. Indeed, there has been a slight deterioration, especially in languages such as Italian ”.
In detail, 23 different completely bogus theses were analyzed in 135 posts that were viewed by millions of people. The boldest, which appeared on January 3, claimed that Bill Gates was convinced that the vaccines had already caused the deaths of 700,000 people. Another stated that several doctors were certain that vaccines altered the DNA, yet another that the obligation of the mask was due to the bulk purchase by the States in the previous months and that there was a need to dispose of stocks .
Mike Schroepfer, number one of Facebook technologies: “We were wrong, but the new Ai will stop hatred and bullying online”
by Jaime D’Alessandro
According to Avaaz, while in English only 29 percent of posts that spread similar theories managed to overcome Facebook’s barriers, both automatic ones based on the analysis of texts by artificial intelligence and that of moderators, for Italian it comes to 69 percent. Followed by the French with 58 percent, the Portuguese with 50 and the Spanish with 33.
But that artificial intelligence (Ai) is more proficient on English is not a discovery today. The algorithms, including those of Facebook, are trained primarily on that language for which there are many more texts. And this means that in the lesser-known Ai idioms it is much easier to circumvent the defense lines, perhaps by changing a few sentences, leaving certain terms implied and so on.
YouTube’s denial algorithm and climate change
Jaime D’Alessandro
“Of course, it is a problem of data availability, but the result in the end is that languages like Italian are penalized,” Nicotra points out. “And then there is also a difference in who controls the content: moderators in Europe are probably fewer”. The request is clear: publication of data on the interventions that are made. The reports communicated to date by Facebook are in fact judged to be too general. Nobody wants censorship, argue the non-governmental organization, but it is a question of establishing a gradual path made up of steps forward and with objective verification of the same. It is not a simple terrain, even from a technical point of view. It will now be up to the European Commission to decide how and when to intervene.
.