case study

HateLab (Cardiff University)

AI Monitoring and Preventing Anti-Immigrant Hate on Social Media After Brexit Call

Toxicity Radar for Brands
1 December 2023
reading time: 6 minutes

Project Information

HateLab is a global hub for data and insight into hate speech and crime. They use data science methods, including ethical forms of AI, to measure and counter the problem of hate both online and offline.

Academics and policy partners have created the Online Hate Speech Dashboard to provide overall trends over time and space. The UK Home Secretary announced the National Online Hate Crime Hub, which piloted the Dashboard. HateLab is funded by grants from the Economic and Social Research Council (ESRC), part of UK Research and Innovation (UKRI) and the US Department of Justice.

Company type

Time on market




The police, victim assistance organizations, and lawmakers all have severe concerns about hate crimes. The biggest increase in hate crimes that the police have ever seen was in the year after the Brexit vote, up 57% from the year before. In England and Wales, the police registered 94,098 hate crime offenses in 2017–18, an increase of 17% over the previous year. Detailed data is available on the HateLab website here.

More than 1,000,000 Poles reside in the United Kingdom, making them the largest national minority, according to the Office of National Statistics in the UK.

In anticipation of a rise in violence against historically persecuted groups as a result of the political climate, HateLab researched the effects of hate speech and violence before and after Brexit. Poles, as the largest minority there, were the study’s primary focal group. Samurai Labs’ pilot investigations reveal that up to 5% of content about this group posted on UK social media has a derogatory or offensive tone, which taking into account the scale of posts published on social media, is a very serious concern.

Samurai was engaged to develop AI models and analyze posts on Twitter and YouTube to detect violent speech acts against Poles. At the latter stage of the project Samurai’s data aim to power up HateLab custom dashboard for better measuring violence before and after Brexit, alerting authorities to trends and immediate threats.

“As the United Kingdom prepares to leave the European Union, using the most advanced methods of artificial intelligence is going to be vital in helping the authorities quickly recognize warning signs and provide reassurance and security to the Polish community living here.” Matthew Williams, Director at HateLab


The ambiguity between the words “poles” and “polish” provides a linguistic problem for automated machinery. The words may refer to the native population or to common nouns that describe ordinary things. Samurai’s artificial intelligence enabled precise analysis of online violence trends, accurate detection of violence directed at Poles, and the ability to react in real-time to offline threats of violence.

Only Samurai’s neuro-symbolic approach to AI could correctly distinguish violence against Poles from ordinary, non-threatening speech acts.

As a former UK expat I was thrilled to collaborate with Hatelab in addressing hate against the Polish people during the Brexit era. Through Samurai’s AI and rigorous research we identified ten categories of hate speech on Twitter. Our collaborative efforts shed light on the magnitude of the issue, empowered stakeholders, and inspired conversations to foster a more inclusive and respectful digital environment. This partnership shows the transformative power of neuro-symbolic technology in keeping online communities safe and promoting social awareness.” Kamil Soliwoda, Head of Symbolic AI at Samurai Labs

What did our cooperation look like?

We started our cooperation in June 2019 with a combined workshop between the HateLab and Samurai Labs teams, which took place in-person in Cardiff, Wales. During our cooperation we conducted more online workshops to make sure both sides were satisfied with outcomes after each phase.

The Samurai Team was provided with an Enterprise Twitter API access to collect and analyze tweets in terms of hateful messages. Having access to the whole platform, our first task was to create preliminary filters to catch only the tweets that could potentially contain anti-immigrant content. Based on the collected messages, the Samurai Team was able to recognize and propose 10 hate categories to be developed within the project. Apart from Twitter, we also analyzed comments published under YouTube videos uploaded by Brits:

  • A combined workshop between HateLab and Samurai Labs in the United Kingdom
  • Processing tweets and comments in YouTube videos related to immigrants
  • Text analysis in terms of hate for national minorities in Great Britain
  • Development of 10 hate categories and categorizing messages for selected groups

We focused on categorizing messages and creating 10 different categories of hate:

  • Insults and stereotypes
  • Requesting to leave the country
  • Accusing prejudice and discrimination
  • Accusing of stealing jobs and claiming benefits 
  • Using specific derogatory terms
  • Threats to health or lives
  • Facts, distortions, and manipulations
  • Accusing crimes (generalizations)
  • Abusive word formations
  • Polish jokes

Thanks to these categories, we are able to check not only where hate speech against national minorities occurs but also to specify the nature of the messages, locate their sources and adjust preventive measures to the nature of the hate speech.


Our collaboration has resulted in an analysis of online hate speech directed against immigrants. Due to the high level of linguistic similarity and the abundance of homonyms, there was a high risk of mismatching the national group and common language expressions, like verbs. This implies that a straightforward keyword search is insufficient. With a precision of up to 82.3%, Samurai developed software that successfully identified publications that contained hate speech against national minorities.

In order to inform and foresee potential dangers to national minorities, state services and other organizations will be able to use an interactive dashboard that will be developed as the project progresses. Then, HateLab will be able to modify communication in areas where certain types of hate are on the rise so that the actual issue is addressed. It allows for better resource allocation, budgeting, and ultimately improved efficiency.

Precision of up to 82.3%

With a precision of up to 82.3%, Samurai developed software that successfully identified publications that contained hate speech against national minorities.

Better resource and budget allocation

HateLab will be able to modify communication in areas where certain types of hate are on the rise so that the actual issue is addressed. It allows for better resource allocation, budgeting, and ultimately improved efficiency.

Do you want to achieve such
results with us?

Case Studies

We support various businesses in their efforts to moderate online speech. Read our recent case studies and find out how we help our clients all over the world.

Developing AI system that meets the goals of the National Strategy for Preventing Veteran Suicide

Suicide rates have been historically high among Veterans, with the estimated risk being 57% higher than that of the general population. In order to investigate the suicidal tendencies among this group, we collected over 41,000 posts from VA Disability Claims Community Forums –

Keeping schools and online learning safe by monitoring spaces and 1:1 chats to detect hostile or inappropriate conversations.

The purpose of Samurai is to detect and prevent violence, making it an essential asset for educational institutions that use platforms such as Webex for remote learning and communication.

Moreover, it has a positive impact on the schools’ reputation. By showcasing a reduced incidence of aggressive behavior, institutions can attract parents; preference and potentially  enhance students’ performance in international educational rankings.

Keeping a top gaming community safe from toxicity and cyberbullying by identifying 30% more cases of community guidelines violations

● Over 130 violations of Community Guidelines were detected by Samurai Guardian each day

● 30% more Community Guideline violations were detected by Samurai Guardian and would be automatically removed when compared to human moderators.

● Less than 3% of Community Guideline violations were removed by moderators without being detected by Samurai Guardian.

    Schedule a free consultation
    with our expert

    Take the first step to make the change in your company with just a little effort. Our representative will contact you within 24h after submitting your request.

    Chief Marketing Officer

    Chief Growth Officer