San Francisco (AFP)

Facebook assured Tuesday that it has made progress in the detection of content inciting hatred, without succeeding in appeasing the associations at the origin of a large advertising boycott of the giant of social networks, accused of laxity in the management of this type of publications.

Despite the Covid-19 pandemic, which has disrupted and reduced the capacities of its content moderation teams, the platform estimates that its proactive detection rate for hate content has increased from 89% to 95%.

This artificial intelligence-based mechanism makes it possible to remove messages and images before they are reported and therefore before they are potentially seen by hundreds of thousands of users. On Instagram, that rate jumped from 45% to 84% in one quarter.

“We have made progress in fighting hate on our apps, but we know we need to do more to make sure everyone feels comfortable on our platforms,” notes Guy Rosen, one of the vice presidents of the Californian group, in charge of integrity.

In all, between April and June, Facebook says it banned 22.5 million publications that violated its regulations on content relating to hate, more than double the previous quarter (9.6 million). This leap is due to improved detection technology and the addition of three languages.

- Stereotypes -

Facebook will also add specific types of content to the banned list, such as "stereotypes about Jews who rule the world" or images related to "blackface" practices, when people make their faces black.

"Often when we talk about hateful content, we think of explicit attacks. But over the years we have identified images, memes (devious images) and generalizations that implicitly attack others," explained Monika Bickert, manager Facebook regulations at a press conference.

The network therefore adds items to the list after feedback and consultations from various partners around the world, knowing that human moderators will be responsible for judging nuances: a person claiming a "blackface" photo does not have them. the same intentions as someone who would denounce the racist past of a political figure, for example.

After the death of George Floyd at the end of May, an African-American killed by a white policeman, many movements and associations of civil society mobilized to fight discrimination and racism in all institutions in the United States, including on Facebook, the dominant social network.

- Boycott -

Civil rights organizations launched a boycott in June called #StopHateForProfit ("No to hate for profits"), which was joined by more than 1,100 companies in July, including Adidas, Levi's, Coca-Cola, Starbucks, etc.

Several hundred are currently continuing the boycott, according to an estimate of the collective, including Unilever (Lipton, Magnum, Dove, etc.).

Facebook responded by insisting that it did not derive any benefit from incitement to hatred, on the contrary, and by making several announcements about the dismantling of white supremacist or conspiracy networks.

Tuesday Guy Rosen also recalled that the group has set up teams specifically responsible for subjects such as diversity, equity and inclusion and lends itself to audits.

"The point is that it's still easy to find a lot of hate on Facebook," responded a spokesperson for Stop Hate For Profit, for AFP. "Facebook needs to show results, not talk about it."

At the beginning of July, a meeting between the organizers of the boycott and the executives of Facebook ended in the disappointment of the associations.

"They talked about + nuance + (...), they told us that they were + on the right track +, that they were improving, that they were almost there", lamented Jonathan Greenblatt, head of the Anti-Defamation League, at a press conference.

"But Starbucks would not say: + we are on the right track, 89% of our coffees do not contain toxins!".

© 2020 AFP