Fanny Dufour October 25, 2021 at 7:34 p.m. © US Congress

The first articles about the Facebook Papers, as the documents leaked by whistleblower Frances Haugen are now referred to, have come out and depict a company monopolized by profit and overwhelmed by the hateful and violent content present on its site.

These articles are the result of a collaboration between several editorial staff around the world, who had access to the whistleblower's documents.

Facebook can't control the effects of its own features

Facebook's most recognizable features — the Like button and post sharing — aren't as innocent as they might lead one to believe. Following several searches, some Facebook employees have sounded the alarm, as this New York Times article shows. They report the harmful effects of the Like button, in particular on teenagers who use it to compare themselves with each other and on whom a lack of recognition of their publications can have a harmful effect. After an inconclusive trial of removing it for a few users in 2019, the company finally concluded that it was "pretty low on the long list of issues we need to fix".

The sharing system has also been the subject of several internal searches. The researchers concluded that it allowed users to share misinformation very quickly, amplified by recommendation systems and highlighting content shared by users' friends. The group and page recommendation system was also denounced by a researcher, who accused it of leading users down the path of conspiracy theories. Since then, some actions have been taken by Facebook, to allow users to hide posts or to no longer receive recommendations from political groups. Nothing fundamentally changed the operation of the site, which was however what was questioned by the employees.

Facebook fails to protect its non-English speaking users

Articles from Reuters and Le Monde focus on Facebook's inability to properly moderate non-English speaking content, particularly in the most fragile countries. Used in more than 190 countries, the social network does not have the means to moderate correctly in most of them, even though the role of the platform in the violence exerted in some of these countries is recognized. However, Facebook compiles a list every six months of countries considered to be at risk. But whether at the level of its human moderators or its artificial intelligences, the company has repeatedly failed to control hate speech and misinformation.

The documents recovered by the editorial staff show that the company did not have the necessary tools in 2020 for automatic detection of posts containing disinformation or hate speech made in Burmese for Burma, or in Oromo and Amharic for Ethiopia, two countries considered at risk. However, as early as 2018, the United Nations had already highlighted the role of the social network in the spread of hateful content against the Rohingyas in Burma. For its part, Reuters notes that it has found numerous publications in Amharic on the site, designating ethnic groups as enemies and containing several death threats.

The documents point to the same issues in other countries, such as India where several moderators in 2021 expressed concern over the rise of posts calling for violence against Muslims. They warn that the algorithms are not equipped to detect hateful or violent terms in Hindi or Bengali. The same concerns about the inability to automatically detect hateful or violent posts have been reported for Pakistan, Iran or Afghanistan.

Le Monde focuses on the recurring problems in Arabic-speaking countries, where the company combines a lack of human moderators capable of moderating content and an artificial intelligence that very regularly makes mistakes and deletes completely legitimate content, which leads to recurrent user censorship. The employees have however launched the alert, underlining the risk of violence in these countries amplified by the social network, and the importance of a solid moderation system. In vain it would seem, a Facebook spokeswoman having admitted that there was still work to be done in this area.

Mark Zuckerberg's decisions put users at risk

The Washington Post article also focuses on the decisions made personally by Facebook CEO Mark Zuckerberg. It would be him in particular who would have decided in the end to respond positively to the requests of the Vietnamese government which demanded more censorship of the publications which criticized him. The alternative being the prohibition for Facebook to operate in the country, the CEO preferred to comply, Vietnam representing more than one billion annual revenues for the firm, according to an estimate by Amnesty International.

He also allegedly knowingly lied to the US Congress by stating that the company removed 94% of hateful content, when internal research, to which he had access, puts the figure at 5% instead. When WhatsApp employees offered to make a Spanish-language version of Facebook's Election Information Center for the 2020 US elections, it was the boss himself who objected, saying it wasn't. was not "politically neutral enough". He also rejected proposals from company researchers to reduce misinformation about the coronavirus in the early months of the pandemic.

Following several decisions made by the founder of the social network, members of the team responsible for mitigating the harm caused by the platform left their role in 2019. “Our very existence is fundamentally opposed to the objectives of the company, to the goals of Mark Zuckerberg,” said one of those employees. A view corroborated by the New York Times article, in which several company executives confirm that several proposed changes to limit misinformation have been blocked in the name of growth.

Between the repeated bugs and the scandals that regularly splatter the most used social network in the world, many members have taken a radical decision: delete their Facebook account. Do you want to do the same? Never mind, we have prepared a tutorial for you on how to delete your Facebook account in minutes! Read more

Source: Engadget