New York (AFP)
Regularly vilified for letting through too many hateful and racist messages, Facebook is currently modifying its algorithms to more severely crack down on insults against minorities while stopping systematically withdrawing attacks against people of Caucasian origin, men or Americans.
This development contrasts with the usual policy of the social network, which generally does not wish to distinguish between genders or ethnicities.
But the group realized "that hate speech targeting underrepresented groups can be the most dangerous," a spokesperson for the social network told AFP on Thursday.
"This is why we have modified our technology to focus on the words that users and experts tell us are the most serious", such as those aimed at dehumanizing or despising a certain category of people, the spokesperson said. .
The group has also updated its practices to be more vigilant in the face of the implicit expression of hate speech, such as the messages on the practice of "blackface" which consists for Caucasians to blacken their faces in order to make fun of black people or on stereotypes suggesting that Jews control the world.
Twitter, which has its own policy of combating abusive and harmful content, has embarked on a similar path by announcing Wednesday that it wants to ban any publication seeking to dehumanize according to racial, ethnic or nationality criteria.
The micro-message platform, for example, plans to remove messages qualifying an ethnic, racial or national group as "rot", "parasites" or "cockroaches".
- "Men are fools" -
Along with the development of automatic moderation systems that are more sensitive to speech against minorities, Facebook has stopped using algorithms that proactively seek out certain attacks against Caucasians, Americans and men like "Americans are stupid" or "men are fools".
But the group can still remove them if they are reported to them.
Facebook just wants to rule out the idea that it more easily censors comments considered less harmful than racist attacks, such as derogatory comments about men after a breakup.
According to the Washington Post, which first reported on Facebook's policy developments on Thursday, the social network wants to focus on content that is malicious against black people, mestizos, Muslims, Jews, and members. of the LGBTQI community.
Mark Zuckerberg's company has for years mastered the automatic removal of certain content, linked to pornography or terrorism for example.
The spread of hate online is more complicated to manage, as machines find it difficult to distinguish between news, humor, parodies, rumors and insults.
As a result, Facebook tends to more easily remove problematic content directed at people of Caucasian descent and report visible minority posts more often, according to the Washington Post.
The platform claims to have initiated the evolution of its moderation policy from 2019.
Such an initiative could be badly received by some of the users, on the right in particular, who already consider that social networks disadvantage conservative messages.
But like other major social networks, Facebook is also regularly attacked by civil rights organizations who accuse it of not doing enough to fight against hate speech.
These grievances grew even more prominent this year after the death of African American George Floyd, and the group faced a boycott this summer from several major groups.
© 2020 AFP