404 Hate not found.

Filtering online hate speech
on social networks in Europe.

The dark side of free speech

Free speech is one of the most important humans rights but it has a difficult equilibrium when people want to contain and prevent hate. Indeed, hate speech cannot be easily defined, but by agreement it expresses loathing and intolerance: the consequences can range from mere bullying or intimidation to actual violent events.

Africa Asia Middle East United States Europe Latin America

of people feel they can express their opinion
Source

Where should we draw the line
between free speech and hate speech?

Filtering on social networks

EU Steps to fight online hate speech

Many EU countries believe that hate speech might be prevented by restricting free speech. One of the hardest battlegrounds is online, because of its characteristics of permanence, transnationality, unpredictability and anonymity.

According to the European Union average published by the European Commission, 29% of the population was a conscious witness of online hate speech phenomena. The social networks are among the main actors in the environment that are required to manage and dominate this kind of problems, since 70% of the European population has been on their platforms at least once and 44% uses them daily.

Run into Hate Speech online Use of social network Daily A few times per week A few times per month A few times per year Never EU Country - % Malta — 55% 63% 14% 3% 2% 18% Czech Republic — 53% 39% 16% 9% 4% 31% ======= >>>>>>> e9c09c9594a7fd2501a5de2e25793024dcb36275 Bulgaria — 52% 50% 18% 7% 3% 21% Poland — 50% 35% 17% 12% 7% 29% Denmark — 48% 63% 12% 5% 2% 18% Slovakia — 42% 46% 15% 9% 4% 26% Croatia — 41% 50% 14% 7% 2% 27% Romania — 37% 46% 18% 8% 3% 24% Portugal — 36% 47% 17% 5% 3% 28% Finland — 36% 53% 12% 5% 5% 24% Greece — 36% 44% 18% 7% 3% 27% Netherlands — 34% <<<<<<< HEAD 53% 17% 6% 4% 20% ======= >>>>>>> e9c09c9594a7fd2501a5de2e25793024dcb36275 Slovenia — 32% 40% 16% 6% 3% 35% Spain — 31% 48% 16% 8% 3% 25% Latvia — 30% 47% 17% 7% 4% 24% Belgium — 29% 48% 15% 7% 3% 27% Sweden — 29% 60% 13% 5% 2% 19% United Kingdom — 28% <<<<<<< HEAD 48% 14% 7% 2% 29% ======= >>>>>>> e9c09c9594a7fd2501a5de2e25793024dcb36275 France — 27% 38% 10% 9% 7% 36% Ireland — 27% 49% 16% 7% 3% 25% Austria — 25% 37% 15% 9% 5% 33% Luxembourg — 22% 39% 22% 8% 4% 27% Hungary — 21% 51% 22% 7% 2% 17% Cypro — 21% 55% 15% 5% 2% 23% Germany — 20% 37% 12% 7% 4% 39% Lituania — 19% 47% 15% 5% 2% 27% Italy — 17% 47% 17% 6% 2% 28% Estonia — 15% 50% 15% 5% 2% 27%
Source

The Code of Conduct

The Code of Conduct

The IT Companies have their own role models—called policies—that are of course influenced by laws in Europe. These policies regulate the filtering process or the removal of improper contents from each platform. The most relevant in the EU is the Code of Conduct on countering illegal hate speech online”, that since its establishment has done 3 surveys and has published them in order to understand how IT Companies react to notifications of illegal hate speech.

5 months 6 months 10 th October–18 th November 1 st monitoring 600 notifications submitted 2.575 notifications submitted 20 th March–5 th May Second monitoring 2.982 notifications submitted 6 th November–15 th Dicember Third monitoring Next steps Make a monitoring every six months January 2017 January 2016 January 2018 9 th January Google+ and Instagram join the Code of Conduct 31 st May Code of Conduct NetzDG law was passed in Germany The NetzDG law came into force in Germany European Council actions Legislative actions Period of monitoring N° of notifications
Source

What are they filtering for?

Twitter

"We do not tolerate behavior that harasses, intimidates, or uses fear to silence another person’s voice. If you see something on Twitter that violates these rules, please report it to us."

Source

Facebook

"Hate Speech is not allowed on Facebook because it causes intimidation, exclusion and in some cases it can generate actual violence."

Source

YouTube

"We encourage free speech and try to defend your right to express unpopular points of view, but we don't permit hate speech. There is a fine line between what is and what is not considered to be hate speech."

Source

The journey of a notification

There are two main important elements of effective filtering: the time of assessment of the notifications and the effective decision which establishes if a content is hateful or not.

How fast can it be?

Time is important because the longer the hateful content stays available, the more damage it can inflict on the victims and empower the perpetrators. The third monitoring highlights the fact that IT Companies have different timings when it comes to assess notifications.

The notifications provided during the third monitoring, established by the European Council, are conveyed and analyzed with different timings by the three social networks adhering to the Code of Conduct during the survey period.

Source

How does it end?

During the three monitorings, as the problem became more and more visible, there was a relatively increasing number of notifications that at the end of the process were actually removed.

1st Monitoring 2nd Monitoring 3rd Monitoring Removed notifications Notifications 270 1273 1124 847 63 1408 33,5% 20,2% 76,7% 23,3% 66,5% 79,8% 163 644 363 241 31 794 62,6% 54,3% 81% 19% 37,4% 45,7% 123 658 585 434 60 780 34,4% 25% 51,2% 65,6% 75% 48,8%
Source[1]
Source[2]
Source[3]

All you can hate

Hateful Categories

Since hate is generated in different contexts, it doesn’t have a unique target. Social Networks have only been a means of proliferation for different kinds of hatred, but the different sympathies and policies of the companies have given more importance to some groups despite others.

The following chart shows the categories of speeches that are held to be Hate Speech from Facebook, Twitter, YouTube.
Over time, the amount of affluence of hatred towards categories may change.

Hateful Categories in the years

The alluvial diagram shows the development of Hate Speech categories during the three surveys carried out by the European Commission.

10 Oct–18 Nov 1 st monitoring 20 Mar–5 May 2 nd monitoring 6 Nov–15 Dec 3 rd monitoring Anti-Muslim hatred Antisemitism Ethnic origin Gender identity National origin Other Race Religion Sexual orientation Xenophobia
Source[1]
Source[2]
Source[3]

Hate words

Hatebase is a tool that aims to map the way hatred spreads: the words it contains help people recognizing hate episodes both online and in real life. Moreover it gives a better understanding of how hate words cause moral damage and how their censorship is in many contexts justifiable.

Source