This website protects your privacy by adhering to the European Union General Data Protection Regulation (GDPR). We will not use your data for any purpose that you do not consent to and only to the extent not exceeding data which is necessary in relation to a specific purpose(s) of processing. You can grant your consent(s) to use your data for specific purposes below or by clicking “Agree to all”.
Findings from our Data-driven Approach to Countering Hate Speech (DACHS) project
Hate speech against journalists is at an all-time high, especially against women: one in two have been facing online attacks and abuse.
How can we protect privacy and ensure technical and emotional safety? What counter-action methods should we use? These are questions journalists need to tackle to do their job online effectively and safely.
To help journalists do that, and to understand the nature of online hate, the European Journalism Centre together with the media monitoring technology company DataScouting, embarked on the EU-funded Data-driven Approach to Countering Hate Speech (DACHS) project. Over eighteen months, we mapped hateful content across social platforms and developed machine learning models to detect hate speech. We also trained journalists on how to counter hateful attacks, how to stay safe online, and how to guard their emotional well-being.
Here are the main lessons we have learned about hate speech and the support that journalists need to stay protected.
When trying to create a tool to protect journalists from online hate, we first had to explore what qualifies as hate speech. The first lesson we learned was how hard it is to define what hate speech is while fully respecting freedom of speech and divergence of opinions.
We analysed hundreds of thousands of social media posts in five different languages, using both the DACHS machine learning models and human annotation. What we found was that national and cultural specifics heavily influence the content of hate speech. For example, domestic politics would trigger hate in the Spanish language, migration in German and the “yellow vests” crisis in French.
But it’s not enough to understand what hate speech is to be able to detect it online. Writing in slang, heavy dialect, using emojis and hashtags as well as misspelling are just some of the things hate posters use to bypass existing filters, especially automated ones. These tricks, combined with the importance of context, made our task even more complicated.
In the course of the project, we also directly witnessed how online hate might affect emotional balance. Even our experienced project annotators felt uncomfortable at times, after browsing through hundreds of examples with threats and incitement to violence. Thus, we realised how detrimental the effect of hate on journalists is and how it may force them to silence and self-censor themselves. As Finnish journalist Johanna Vehkoo said in an interview for DACHS, the real harm of hate speech is that “our societies are at risk of losing many voices from public debate”.
As part of the project, we visited seven newsrooms in four European countries to give face-to-face training on countering hate speech. During this process, we learned a lot about the ways journalists can be supported — and what role their newsrooms need to play.
Both journalists and newsroom managers agreed that it is important to talk about online hate. How that conversation goes, however, depends on how much their working environment is prepared to deal with the issue. In news organisations where the problem is “institutionally” recognised and there are procedures and protocols in place, individual journalists are more aware of the phenomenon and appreciate training on the topic. Where the conversation around hate speech has just started at an organisational level or has not been tackled before, there are more doubts and frustration.
Apart from leading the conversation, newsrooms need to acknowledge the issue of hate speech and support their staff, especially freelancers. This support has to include training, development of ethical standards and reporting procedures, as well as technological tools. Hate campaigns — What you should do, for example, is a great resource developed by the Union of Journalists in Finland that can be used as a stepping stone in that direction.
Ensuring diversity in newsrooms is also very important to make a strong argument for social inclusion and countering social aggression.
In addition to the support from the newsroom, journalists themselves need to be aware of their rights as professionals, individuals and social media users. They need to be able to report content to platforms (like Twitter, Facebook, Instagram reporting guidelines) and relevant authorities, and to be aware of the software tools that can help them flag, counter or mentally prepare for hateful attacks.
The ‘Doing It Yourself Step-by-Step Guide’ includes a toolbox of tips, tools and tactics for countering hate speech. The Counter-Narrative Toolkit provides resources for planning, creating content and promoting counter-narratives.
Such ready-made strategies and tools to apply in their daily work are crucial to allow journalists to fulfil their professional tasks while also maintaining emotional well-being and not surrendering to the pressure and self-censorship.
As for social platforms, they still need to tackle the issue of hate at a fundamental level. To achieve tangible results, it is worth considering giving more and easier access to their data for scientific and research purposes.
Our recently-launched website hatedetection.com is a place where journalists, scientists, researchers and other interested parties can access and further analyse the results of our project.
As the main target group of DACHS, journalists can benefit from the project in multiple ways, ranging from following a free online “Countering hate speech” course, to registering for the DACHS alert monitoring service and annotation extension.
The data sets created under DACHS have considerable potential for the scientific and research communities. The five data sets contain verified hate speech examples in English, German, French, Spanish and Greek. Together with the DACHS API, they can be accessed through the website.
The European Journalism Centre and DataScouting aim to continue the work on creating tools and training that empower newsrooms and journalists to counter online hate. To achieve that, we will maintain and improve hatedetection.com and we will also work with newsrooms and social platforms to find further solutions.
If you’re interested in the project, we’d love to hear from you. To stay up to date with DACHS and other projects the EJC is carrying out, sign up for our newsletters.