The hate speech expands through the Internet at full speed. In its attempt to mitigate this phenomenon, corrosive to democracy, the technological giants committed in 2016 to erase in less than 24 hours messages that incite racism, xenophobia or sexism. To carry out this task they have hired an army of vigilantes that filter an avalanche of content circulating on the Internet and social networks. Your task is to erase everything that encourages hatred in any of its aspects: ethnicity, religion or nationality.
The images of the relay on Facebook of attack on two mosques in New Zealand, which caused around fifty deaths, were circulating on the Internet until the police gave the alert. Facebook ensures that he quickly deleted the video and also the shooter's accounts. However, the copies had already multiplied. His artificial intelligence filters failed. The content was played 4,000 times until the company was able to access the link with the recording. In 24 hours they eliminated 1.2 million videos. YouTube, Twitter, Instagram and Reddit also removed the images as they appeared on these platforms.
Four major technology companies – Facebook, Twitter, YouTube and Microsoft – signed a code of conduct with the European Commission in 2016 to combat hate speech on the Internet. Other smaller companies such as Dailymotion or Snapchat joined this document. These corporations claim that they have reinforced their security and surveillance systems to eradicate malicious content. They use three methods: user complaints, artificial intelligence programs and human watchers.
Google explains that "the standards of what constitutes hate speech vary from one country to another, as does the language and jargon they use" and that "speed and accuracy" have improved. Between July and September of last year, YouTube eliminated more than seven million videos for violating its rules and blocked 224 million comments. At that time, Facebook deleted 2.5 million pieces that violated "the norms on hate speech". 53% were detected "proactively" through their own systems before receiving external notice.
The company created by Mark Zuckerberg has hired experts in areas related to minors, hate speech and terrorism. Their teams monitor in 80 different languages. It's hard work, as pointed out by the Justice Commissioner, Vera Jourová, after visiting, recently, a content control center in Dublin. "I can say that they have one of the least pleasant jobs in the world. They are heroes of the digital era. "
Artificial intelligence is not enough; you need the control of people
On Facebook, around 15,000 people work in control missions. One of the centers is in Barcelona and operates in collaboration with the Competence Call Center (CCC). In November, it already had 800 employees to "review all kinds of problematic and abusive content."
"It is impossible to avoid 100% that those messages are present. But it is possible to detect early, limit their virality and make their search on the Internet very complex, "says Javier Lesaca, a history doctor and visiting researcher at Columbia University. Considers that "new digital communication platforms should be aware that they are no longer simple technology companies", because "they have become the backbone of public opinion in society and must assume the responsibility that this new role implies".
The platforms know that the EU is vigilant. "They do not invest motu proprio. Act by the pressure of citizens and institutions. They are forced to adopt measures if they want to keep the goose that lays the golden eggs, "says Ferran Lalueza, professor of the Information and Communication Sciences Department of the UOC. The platforms also act because "they are aware that they are exposed to millions in fines," stresses Lalueza, according to which the mechanisms to combat hate speech are insufficient. "These contents continue to spread through social networks."
When it comes to executing these tasks, Facebook relies on artificial intelligence tools, but human surveillance is fundamental. Those who are dedicated to flooding the Network of hate messages are managed to circumvent the detection algorithms. Therefore, companies increasingly trust these tasks to controllers of flesh and blood. YouTube wants to have 10,000 people to review the videos hosted under their domains.
All these initiatives are an advance, but "there is still much to improve", says Lalueza. Systems that use artificial intelligence are an effective way, but in a multinational like Facebook, with 2,300 million active users, "any attempt at human control is doomed to fail or not have enough agility for the measures to take effect."
Platforms prefer self-regulation and, by the way, keep regulators away. In front of those who consider that eliminating content is a type of censorship, Lalueza maintains that the red line is not in the freedom of expression but in the freedom of aggression. "And at this point, the authorities should not shake their pulse. Hate speech can not be protected by freedom of expression. It must not have protection because it undermines the rule of law. "
For the European Commission, the code of conduct is "an important step", but Lesaca perceives that legislative regulations should be updated to toughen the publication of violent messages or incitement to hatred in the digital environment and require platforms much faster to detect and eliminate those contents. And it gives the warning voice: "Terrorist, violent and extremist groups are increasingly migrating their content to smaller platforms, but they are equally effective in viralizing and socializing their contents."
Saying the word "hate" is not enough to erase a message on the Internet or make a video disappear from a social network. To be considered dangerous, the text or images must be accompanied by a threat, incite violence or make reference to race, religion or sexual condition.
Through the Somos Más initiative, Google has designed a program to prevent and sensitize Internet users about hate speech and violent radicalization. The project is aimed at 30,000 teenagers from 250 Spanish schools. "Somos Más", Google points out, "has three objectives: to make the Internet a hate-free space, to prevent young people against radicalism and to sensitize teenagers through YouTube creators."
In Europe, last year companies linked to information technologies eliminated 70% of the messages with illegal incitement to hate reported by NGOs or public bodies. And in a great majority of cases (81%) they retire in less than 24 hours.
Facebook has applied filters outside the community environment. The NGO Business for Social Responsibility analyzed the impact of the social network in Burma and detected a worrying increase in hate messages against the Rohingyas, the Muslim minority. The company admitted that it had not done its homework in stopping the expansion of these messages and hired a hundred Burmese-speaking employees to identify the xenophobic messages and suppress them.