‘Action Plan on Monitoring Hate Speech by the Racist Violence Recording Network’

For Signal – Researching and Confronting the Far Right:
Kostis Papaioannou, Ioanna Meitani, Daphne Sfetsa

For the Racist Violence Recording Network:
Garyfallia Anastasopoulou, Daphne Kapetanaki, Katerina Pantou

About Signal – Researching and Confronting the Far Right

Signal – Researching and Confronting the Far Right is an independent citizens’ initiative focused on countering the rise of far-right phenomena. It is committed to providing robust counter-narratives to combat the spread of racism, nationalism, sexism, homophobia, anti-Semitism, Islamophobia, and other ideologies that undermine human dignity. Upholding the principles of an open society, democracy, and human rights; Signal’s work is dedicated to combating right-wing extremism, encompassing scholarly research, publications, educational programs, and awareness campaigns. It engages with the public discourse, aiming to shape the strategies that address far-right extremism, racism, and nationalism.

Signal – Researching and Confronting the Far-Right is a Non-Profit Civil Partnership which adheres to the principles of transparency, accountability, and independence. Collaboration with similar initiatives in Greece and in Europe is key to our vision mandate.

Acknowledgements

This action plan was developed in response to the Racist Violence Recording Network’s (RVRN) longstanding aim to expand its scope and include monitoring hate speech in the public sphere. Therefore, the realization of this action plan and its specific outcomes would not have been possible without the valuable insights and input from the Network’s coordinators. Special thanks go to Assistant Coordinator Garyfallia Anastasopoulou, whose dedicated and significant contributions were pivotal to the planning and implementation of this project.

Building on these efforts, it is evident that the project’s success could not have been achieved without the active involvement of representatives from various organizations, who participated in the focus groups specifically organized for this initiative, including: DIOTIMA, Hellenic League for Human Rights, HIAS, Solidarity Now, Human Rights 360, Greek Forum of Migrants, Greek Council for Refugees, Refugee Support Aegean (RSA), Positive Voice, Babel Day Center, Lesvos Project, Signal – Researching and Confronting the Far Right, Colour Youth – Athens LGBTQ Community, Generation 2.0 for Rights Equality and Diversity, PRAKSIS, Trans Support Association, Arsis, and a delegate from the National Commission for Human Rights.

We would also like to extend our heartfelt thanks to Ferry Griffin, Joanna Perry, Daniel Heller, Maik Fielitz, Melissa Sonnino, Yves Renuixen-Mencel, Dani O, Irina Corobcenco, Valentina Scioneri, Simona Cruciani, Maria Westergen, for their invaluable assistance and input in earlier discussions. Our special gratitude goes to the Bodossaki Foundation, along with Beatrice Giannoglou and Sophocles Danassis, for their support, which has been instrumental in bringing this project to completion and, we hope, into effective action.

Stevi Kitsou

Ph.D researcher, Maastricht University

Research Associate, Signal – Researching and Confronting the Far Right

I. Foreword

This action plan offers a proposal to the Racist Violence Recording Network addressing the potential expansion of its mandate to monitor hate speech against minorities and target groups, in the public sphere. The Network’s initiative stems from the recognition of a correlation often observed between hate speech and hate crimes – a connection substantiated by service–providing organizations and evidenced in the Networks’ own annual reports.

The 2019 report specifically has pointed out the increasing prevalence of racist rhetoric¹, which has been used to legitimize certain institutional policies and has contributed to a rise in violent xenophobic and racist attacks observed since early 2020². it is highlighted that hate speech in the public sphere is becoming more prevalent, frequently targeting human rights’ defenders so as to discredit and obstruct their efforts³. This development, in turn, broadens the range of individuals targeted and increases the risks faced by groups already vulnerable to racist violence.

This trend is further confirmed at a broader institutional level by the Committee for the Elimination of Racial Discrimination (CERD), which states that those who propagate hate speech “exploit the principles and mechanisms of democracy to legitimize racist and xenophobic actions but also hate rhetoric itself”⁴.Furthermore, the Office for Democratic Institutions and Human Rights within the Organisation for Security and Cooperation and Development (OSCE/ODIHR) highlights the link between hate speech and hate crimes, underlining the importance of documenting instances of racist rhetoric¹. Aligning with these perspectives, the United Nations’ guidelines advocate for a coordinated action strategy ².

Consistent with this framework, the General Recommendation No. 15 from the European Commission against Racism and Intolerance (ECRI) of the Council of Europe, stresses the importance of systematically recording and analysing hate speech where it appears³. This process aims to achieve several key objectives: identifying hate speech as a reflection of societal attitudes, recording and evaluating the harm experienced by its victims, investigating the potential connection between this rhetoric and hate crimes, and providing insights to both institutional and educational organisations to enhance their understanding and response to the phenomenon. These actions are fundamental to the development of supportive, empowering, and inclusive programmes, which are crucial not only for the broader community but also within educational settings⁴.

Following the same rationale, the Recommendation CM/Rec(2022)16 of the Committee of Ministers to Member States on Combating Hate Speech⁵states that Member States have an obligation to collect data on instances of hate speech within the public discourse. This obligation extends to speech that may not be subject to criminal charges but still has a harmful impact, according to European human rights and protection standards⁶.

Within this institutional framework, the need for systematically documenting, monitoring, and analyzing hate speech incidents in the public sphere becomes increasingly evident. Furthermore, this effort not only illuminates the connection between hate speech and racist violence, as evidenced in the Network’s existing reports through qualitative analysis of documented incidents⁷, but also has the potential to motivate civil society organizations to develop targeted advocacy strategies, research-driven programs, inclusive awareness-raising practices, and educational tools aimed at addressing and combating hate speech. Moreover, the evolving role of the Network as a central coordinating body is reinforced, enabling it to organize and oversee partnerships at local, national, and supranational levels. This development expands the Network’s operational reach, strengthens its sustainability, and broadens the engagement of diverse participants in addressing hate speech.

II. Methodology

The methodology adopted to develop the action plan combines two principal approaches:

(a) researching best practices at the European level for mapping and highlighting hate speech through civil society initiatives, and

(b) conducting focus groups with representatives from the Network for Recording Racist Violence incidents.

The research on best practices was deliberately framed within the constraints of time and resources available for this project. Consequently, the review offers an illustrative rather than exhaustive account of the various initiatives, programs, and actions undertaken by civil society organizations to monitor hate speech. The focus on the European context was a necessary limitation, driven by the absence of a universally accepted definition of hate speech. This contextual focus allowed for a more effective analysis aligned with the shared legal traditions and coordinated efforts of the European Union and the Council of Europe in addressing hate speech.

To gain a more nuanced understanding of the different methodological approaches employed in researching and recording hate speech, a series of individual online meetings were conducted with representatives from several key initiatives, organizations, and networks active in this field. These meetings provided an opportunity to explore a diverse range of good practices in hate speech monitoring across Europe. The consultations involved both individuals and organizations, including Joanna Perry from Facing Facts; Maik Fielitz from Das-Nettz in Germany; Yves Renuixen-Mencel and Dani O representing the Community Security Trust; Irina Corobcenco of Promo-LEX in Moldova; Valentina Scioneri from the Resource Center on Security and Crime in Italy; Simona Cruciani and Maria Westergen from the United Nations Office on Genocide Prevention and the Responsibility to Protect in New York. These consultations highlighted existing methods and approaches, which are analyzed further in the following sections.

The second phase involved conducting focus group discussions with representatives from organizations that are part of the Racist Violence Recording Network. These discussions were designed with a dual purpose: first, to present and evaluate best practices in hate speech recording, along with the specific challenges faced in this process; and second, to encourage an open and constructive dialogue on the future role and potential development of the Network.

The insights derived from both the research and the focus group discussions were crucial in shaping the final proposal presented in this action plan. Specifically, based on these findings, the proposal outlines a strategic role for civil society organizations in documenting hate speech. By integrating best practices from comparable European initiatives and taking into account the feedback from Network members, the recommendations are designed to be both comprehensive and practical. Ultimately, this approach seeks to balance theoretical insights with practical applications, strengthening the Network’s ability to effectively address hate speech in the future.

III. Overview of Best Practices

At the European level, several programmes have been developed to record and analyse hate speech incidents on social media. Some indicative examples are: Hate meter (2018-2020¹), E-more project (2016-2018²), C.O.N.T.A.C.T project (2015-2017³), MANDOLA project (2015-2017⁴), WE CAN for human rights speech (2020-2022⁵). The initiatives illustrate the diverse approaches taken to address hate speech and highlight the evolving strategies used across different European contexts.

Currently, the ‘European Observatory for Online Hate’⁶ project plays a central role in monitoring hate speech online using AI tools. This observatory is implemented through a partnership of four organisations: ‘Textgain’⁷ from the University of Antwerp, ‘Dare to be Grey’⁸ in the Netherlands, ‘PDCS’⁹ in Slovakia, and the Department of Applied Sciences at Utrecht University¹⁰. Funded by the European Commission under the 24-month ‘Rights, Equality and Citizenship Programme Call: REC-AG-2020’¹¹, the observatory records hate speech across 20 online platforms.

Furthermore, the International ‘Network Against Cyber Hate’¹² (INACH), based in Amsterdam, holds a prominent position in documenting and countering online hate speech. This extensive network consists of 32 organizations from 27 countries, all committed to systematically monitoring and facilitating the removal of illegal¹³ hate speech across online platforms. As part of its efforts, INACH coordinates monitoring activities through the SafeNet project¹⁴ which focuses on identifying and addressing hate speech in various partner countries, in line with national legislation. Additionally, INACH’s mission includes educational initiatives, digital tools, and the development of advocacy strategies to address hate speech on social media. The network is also actively involved in assessing social media platforms’ compliance with the ‘EU Code of Conduct on countering illegal hate speech online’¹– an agreement signed in 2016 between leading social media platforms² and the European Union, as a self-regulation model, where IT entities are tasked with identifying and removing effectively³ illegal content from their platforms.

Together, these examples highlight the systematic and coordinated efforts by civil society organizations across Europe to effectively record and address hate speech. The review below examines these practices in detail, emphasizing their impact and the insights they provide for future approaches.

Security Community Trust

The Community Security Trust⁴ (CST) is a London-based organization dedicated to protecting the Jewish community across England and abroad. The scope of its activities is not confined to documenting anti-Semitic hate speech; it also includes initiatives to raise awareness and combat anti-Semitism in all its forms and manifestations.

The Community Security Trust (CST) systematically compiles reports of anti-Semitic incidents. These reports encompass a range of offences, including hate crimes like assaults and property damage, threats, and the propagation of anti-Semitic content in literature and slogans. Furthermore, CST documents hate speech across digital platforms and the public sphere, distinguishing between illegal and non-criminal yet offensive speech. For the incidents it records, CST does not require the personal details of those involved but does mandate that each report be attributed to a specific source (reporter)⁵.

The anti-Semitic incidents reported to CST are included in its biannual and annual reports, and they are received through various channels. These include reporting by phone, email, through the CST website or social media profiles, or in-person to CST staff and volunteers. Incidents can be reported by the victim, a witness, or an individual or organisation acting on the victim’s behalf, using CST’s specialised reporting forms.

CST has signed a memorandum of cooperation with the National Police Chiefs Council (NPCC) as well as similar agreements with regional police forces, enabling CST to share reports of anti-Semitic incidents with the Police while maintaining full anonymity and data protection compliance. In this manner, the Police gain a broader picture of the number and nature of incidents while this allows cross-checking for potential duplication with their own records.

The Community Security Trust does not proactively gather data on anti-Semitic hate speech through monitoring. Rather, CST only documents hate speech that is directly reported to the organisation or otherwise brought to staff and volunteers’ attention. However, CST does employ artificial intelligence tools to track social media platforms, with a dedicated team recording and evaluating online environments that show hostility toward the Jewish community (information collection on potential hostile reconnaissance and suspicious behaviour¹). While not incorporated into CST’s published incident reports, this social media monitoring remains an integral component of their work, as it offers a vital understanding of possible dangers to the Jewish community.

All anti-Semitic incidents are documented and presented in a consistent format across CST’s publications, regardless of whether they occurred online or elsewhere. The reports categorise each incident by offence type, victim demographics like gender and age, perpetrator identity when possible, suspected ideological motivations, medium/platform, and geographic location recorded². This qualitative analysis is supplemented with specific examples (including screenshots of online incidents) and graphs showing trends and patterns compared to previous months and years.

Promo-LEX Association³

Since 2017, the Promo-LEX Association has been systematically and consistently documenting hate speech incidents in the Republic of Moldova.

As a non-governmental organization committed to promoting democracy and human rights, Promo-LEX aims to strengthen civil society and monitor democratic processes in Moldova. The organization has developed a comprehensive system that deploys teams of human rights experts to track hate speech across both traditional media, including television and radio, as well as social media platforms, focusing on specific websites and public profiles tied to political and societal discourse.

Promo-LEX conducts its hate speech documentation in cycles that span 3 to 6 months, depending on the current sociopolitical conditions. Prior to each monitoring phase, specific public profiles and websites are selected based on their potential for hate speech or discriminatory content. Special attention is given to politicians and public figures, with a focus on their social media activity, including Telegram since 2022. For political figures, the team thoroughly assesses the public communications of these figures across all channels used to disseminate their political narratives, extending to traditional media platforms.

During the data collection phase, Promo-LEX’s experts conduct a distinct monitoring process specifically for election periods. For classifying a case as hate speech, they depend on an internally designed questionnaire, which helps to collect disaggregated data and categorize instances of hate speech by examining the speaker, the context, and the medium, with a special focus on journalistic discourse, following the typology suggested by the UN Rabat Plan of Action¹.

As Moldova has not yet become an EU member state, its national law is not subject to the binding requirements of the Council Framework Decision 2008/913/JHA on combating racism and xenophobia². Instead, Moldovan legislation outlines provisions for addressing bias-based and hate-motivated crimes, discrimination, incitement to discrimination, anti-Semitism, the dissemination of propaganda, genocide, and crimes against humanity. In response to this legislative context, Promo-LEX employs a comprehensive approach to defining hate speech, seeking to capture its various manifestations, including sexist or homophobic statements made by political figures³. The team also assesses comments that may contain subtle hate undertones and examines public reactions to such remarks to better understand their societal impact. Promo-LEX, therefore, adopts a comprehensive approach to defining hate speech aiming to capture its various manifestations, including political figures’ sexist or homophobic statements. The team also evaluates comments that could potentially have undertones of hate, alongside public reactions to such comments.

Upon concluding the monitoring periods, Promo-LEX engages in both qualitative and quantitative analyses of the gathered data, providing thorough justifications for classifying specific incidents as hate speech. The gathered data is subsequently incorporated into the organization’s annual reports, which are made available to both the public as well as governmental and local officials. Alongside publishing their reports, they submit policy recommendations to the government informed by their findings, such as the recently adopted parliamentary provision prohibiting hate speech and discrimination during elections.

Germany stands out as a leading example of effective strategies for monitoring, analyzing, and combating hate speech, supported by numerous initiatives and programs that often receive federal backing. A prominent project in this field is “Das-NETTZ,” which plays a central role in tackling hate speech on social media.

i. Das NETTZ

Das NETTZ¹ is an umbrella network that coordinates, connects, and enables the work of civil society groups tackling hate and, especially, online hate speech in German-language contexts. The Network consists of diverse organizations and experts across Europe, ranging from grassroots activists directly engaged with hate issues to prominent NGOs, researchers, communication specialists, and digital data experts involved in investigating, examining, and combating online hate.

Das NETTZ has initiated a federal action group called ‘BAG gegen Hass im Netz’² (Against Online Hate) to tackle hate speech on the Internet. This group operates as a research centre and is part of the ‘Living Democracy!’ programme, receiving support from the Federal Ministry of Family Affairs and the Robert Bosch Foundation.

The group’s efforts are twofold: it investigates hate speech online, particularly within German-speaking social media, and it runs a network system to distribute its research findings to civil society, fostering an interactive exchange. Organizations within this network use this data to inform their strategic initiatives while they provide feedback to the research team regarding the needs of civil society in order to define future research directions.

The process at the research centre is outlined as follows: It captures real-time data from social media platforms and organizes it into datasets that will later become part of comprehensive reports for ongoing hate speech monitoring. The data is gathered using artificial intelligence that targets 1000 to 2000 public accounts, at intervals of every four hours. After gathering the data, the centre conducts a qualitative analysis and shares the findings with the community, other researchers, and journalists through the organization’s network. Meanwhile, NGOs participating in the initiative suggest improvements to the information technology methods, creating a dynamic feedback relationship with the research team responsible for organising and managing the data collection. The NGOs use the collected data to support their educational initiatives, advocacy work, and the development of counter-narratives to hate speech identified online, along with the regular publication of the online magazine Machine Against the Rage³. Additionally, BAG coordinates and supports NGO initiatives that monitor hate speech online, and it has partnered with 20 NGOs—a number that has been growing since the group’s activities began in June 2022.

ii. Kompetenznetzwerk gegen Hass im Netz

The ‘Kompetenznetzwerk gegen Hass im Netz⁴’ (Competence Network against Online Hate) is a collaboration between four organisations that are active in developing initiatives to counter online hate. These organisations include the: ‘Das NETTZ’⁵, ‘Hate Aid’⁶, ‘jugendschutz.net’¹, and ‘Neue Deutsche Medienmacher*innen’². This network aspires to be a reference point for hate speech research within Germany. In the case of the competence network, emphasis is placed on interventions. The network consolidates expertise in various areas, including media education, counseling on online hate, child and youth protection, support for affected individuals, and guidance for media professionals, while also engaging in overarching advocacy work. Although the initiative does not currently document hate speech online in an independent manner, it is expected to make this monitoring process a major part of its work, particularly in terms of analysing media content and discourse.

iii. Mechanisms and Systems for Recording Hate Speech Incidents

Regarding the reporting and recording of hate speech online in Germany, individuals have various options to report relevant incidents. Beyond the standard processes of flagging illegal content on platforms or contacting authorities directly, Germany requires social media platforms to offer a specific reporting feature (known as ‘impressum’) in accordance with the NetzDG law, aimed at combating hate speech on social networks. This requires platforms to include a reporting option compliant with German regulations, not just platform community standards.

A prime example of an online hate speech reporting tool in Germany is ‘Respect! Gegen hetze im netz’³, an initiative supported by the Youth Foundation of Baden-Württemberg, the Democracy Centre of Baden-Württemberg, and the Bavarian government. The initiative consists of an interdisciplinary team that tries to collect reports of online hate, filter them, and guide victims towards further actions, while simultaneously gathering statistics for researchers. In rare cases, it forwards reported incidents to the Police. The Democracy Centre is funded by the Ministry of Social Affairs, Health, and Integration of Baden-Württemberg through funds from the Federal Ministry of Family Affairs, Seniors, Women, and Youth (BMFSFJ). This funding comes from the federal ‘Living Democracy!’ programme, which also provides support to organizations such as ‘Das NETTZ’. The initiative receives additional support from the Ministry of Family Affairs, Labor, and Social Affairs of Bavaria.

Similarly, the ‘Hassmelden’⁴ (hate reporting) allows individuals to anonymously report online hate speech. The team reviews these reports and, if necessary, passes potential legal violations to the authorities, keeping the reporter’s identity confidential. However, lately, Hassmelden has remained relatively inactive.

Rete Nazionale per il Contrasto ai discorsi e ai fenomeni d’odio⁵ (National Network for Combating Hate Speech and Hate Phenomena)

In Italy, numerous independent initiatives like ‘Contro l’odio’¹, ‘Barometro dell’odio’², and ‘eMORE’³ have been launched to address hate speech on social media.

Yet, for comprehensive, systematic efforts in mapping and countering hate speech, the ‘Rete Nazionale per il Contrasto ai discorsi e ai fenomeni d’odio’ is prominent in the field. This network includes NGOs, academics, research centres, the National Criminological Council, and the National Office against Racial Discrimination⁴. The Network was established in 2020 following efforts beginning in 2018 by the Italian section of Amnesty International to create a coalition of experts and organizations that could study, document, and address hate speech.

Within the Network, interdisciplinary work is carried out as research centres and activists jointly monitor hate speech phenomena from different perspectives and approaches. As stated by the Network itself, research conducted by its members spans a wide range of fields, including social sciences, psychology, law, discourse analysis, and computational linguistics. This comprehensive approach involves analysing a substantial volume of data from multiple angles to gain a deep understanding of the mechanisms governing the creation and dissemination of hate speech. This research not only helps in dismantling these mechanisms through direct actions but also contributes to legislative proposals, educational initiatives, professional training, and the development of counter-narratives to combat hate speech.

Under the Network’s framework, the ‘Vox-Osservatorio Italiano sui diritti’⁵ (Vox-Italian Observatory on Rights), in collaboration with the State University of Milan, the University of Bari Aldo Moro, Sapienza University of Rome, and the ‘ITSTIME’ team of the Catholic University of Milan⁶, created the ‘Mappa dell’Intolleranza’⁷ (Map of Intolerance). This annual report maps hate speech incidents on Twitter⁸ using AI and linguistic analysis tools. The data collection focuses on six main categories: misogyny, homophobia, xenophobia, ableism, Islamophobia, and anti-Semitism. The qualitative analysis geo-locates the incidents throughout Italy, producing a ‘heat map’ depicting the degree and type of prejudice by category across regions.

The research records hate speech incidents and trends solely on Twitter. Although not the most widely used platform in Europe, Twitter allows for direct reproduction of ‘tweets’, highlighting the influence of the medium and a user community that consistently engages with and spreads hateful expressions. The choice of ‘hashtags’ (#) also provides insight into the emotions experienced or supported by users. The results are compared to data from previous years, with an in-depth analysis of factors influencing observed trends.

The framework defining expressions as ‘hate speech’ was based on a proposal from the University of Milan’s Department of Public Law, taking as a general basis the prohibition of discrimination and the rights violated by hate speech.

The second phase of the recording process involves the identification of ‘sensitive’ words, which serve as keywords associated with emotions and are subjected to analysis. This specific aspect of research is conducted by experts from the Department of Clinical Psychology at Sapienza University of Rome, specialising in the study of gender identity and the collective emotions manifested on the Internet.

In the third stage, the actual mapping of the ‘tweets’ is performed using specialised software developed by the Department of Computer Science at the University of Bari. This Social Network Analytics & Sentiment Analysis platform employs AI algorithms to comprehend linguistic nuances, identifying and extracting specific content.

The collected data is then sociologically analysed by the research group ITSTIME¹. The software used in the project is trained to extract the most aggressive tweets, highlighting intensity levels based on the MOAS scale (Modified Overt Aggression Scale)². The degree of intensity regarding hate speech expression is also analyzed.

According to the project’s description³, the ‘Map of Hatred’ aims to prevent and interpret hate rhetoric, as well as combat hatred and bigotry more broadly, locally and nationally, in schools, agencies, and field organizations.

IV. Focus groups findings

In the second phase of the project, focus group discussions were held with representatives from member organizations of the Racist Violence Recording Network. The sessions aimed to present the best practices identified in prior research and to engage participants in discussions about the Network’s future development, potential challenges, and ways in which members could contribute more effectively.

The discussions addressed four main themes:

1. Framework for defining hate speech and hateful discourse

2. Recording framework (covering public spaces, social media, and formal reports)

3. Methodology

4. Data usage and application

Participants included representatives from the following organizations: DIOTIMA, Hellenic League for Human Rights, HIAS, Solidarity Now, Human Rights 360, Greek Forum of Migrants, Greek Council for Refugees, Refugee Support Aegean (RSA), Positive Voice, Babel Day Center, Lesvos Project, Signal – Researching and Confronting the Far Right, Colour Youth – Athens LGBTQ Community, Generation 2.0 for Rights, Equality and Diversity, PRAKSIS, Trans Support Association, Arsis, and a delegate from the National Commission for Human Rights.

One of the central questions discussed was the framework for defining hate speech. Specifically, the group considered whether the Network should differentiate between speech that is legally defined as hate speech and speech that, while not illegal, is still harmful. The challenge, therefore, was how to implement this distinction in a way that would enhance the Network’s work without limiting its scope.

Participants agreed that adhering strictly to a legal definition, confined to the current criminal legal framework¹, would limit the Network’s ability to depict hate speech as a societal indicator and to comprehensively analyze the range and nuances of hatred, including its potential connection to documented hate crimes. Therefore, it was decided to adopt a broader definition consistent with the Council of Europe’s General Recommendation No. 15 on combating hate speech, while also differentiating between illegal speech and pursuing appropriate legal actions when necessary. Monitoring both illegal and harmful speech directed at marginalized groups is essential for comprehensively understanding and addressing hate speech as a societal issue. This approach aligns with the Racist Violence Recording Network’s focus on both operational and long-term, strategic objectives.

Regarding the strategy for documenting hate speech incidents, there was a consensus that social media platforms are key areas where hate speech is expressed and quickly spread. As such, any recording efforts should primarily focus on social media, while not entirely excluding traditional media, especially when they clearly broadcast hate speech. Specifically, Twitter (now ‘X’) was identified as having particularly high levels of toxic and aggressive public discourse. It was further reiterated that the use of hashtags (#) on this platform allows for a detailed analysis of trends over time and across various topics.

It was also suggested that the age of social media users and the platform they choose, could influence the type and severity of hate speech that occurs, depending on the speaker’s profile. A question was raised about the need to document hate speech from politicians or public figures separately, using Moldova’s approach as a potential model. This separate documentation could facilitate a more targeted analysis. The idea of focusing primarily on political and public figures was also suggested as a way to manage the potentially overwhelming volume of data from hate speech reports.

The methodology for documenting hate speech was a central point in the discussions about expanding the Network’s activities. Considering the best practices from various European countries and the Network’s limited resources, it was evident that the current methods used for recording racist violence are not suitable for monitoring hate speech, especially on social media. Given that Network members are already committed to their primary roles, expecting them to also voluntarily gather large amounts of data on documenting hate speech through conventional methods is not practical or feasible.

Consequently, the use of artificial intelligence tools, similar to Germany’s BAG model, emerged as the only realistic and sustainable method for identifying hate speech incidents. Additionally, the idea of a ‘self-referral’ system was discussed, enabling direct reporting by victims via a dedicated form on the Network members’ websites. This system would allow individuals to report hate speech incidents, whether they are victims or bystanders, with robust guarantees for the protection of personal data, similar to the reporting system used by the Community Security Trust in the UK.

Concerns about the new reporting system challenges were twofold: It requires individuals to be able to identify hate speech to report it effectively. Additionally, there’s a logistical challenge: determining the team responsible for processing the reports, filtering them, and classifying the incidents by severity. It was acknowledged that this workload should not be shouldered by the Network’s assistant coordinator alone since the Network’s members can only indirectly engage by facilitating the process through their websites.

Additionally, it was highlighted that there’s a need to reassess how Network members engage with the communities they represent, especially if these communities are willing to be a part of the process on a voluntary basis, in the collection and processing of reports, following the models of the UK and Moldova. Another important consideration revolved around the potential risk of overburdening or diverting attention from hate crime recording, as it was noted that the observed trend is not a reduction in racist violence itself but an underreporting of such incidents.

Notably, in this context, it is important to acknowledge the potential limitations of self-reporting, particularly from the perspective of victims. One concern is that self-reported data may seem to lack follow-up or visibility, leaving victims uncertain about whether their experiences will lead to any tangible outcomes. In contrast, the RVRN methodology provides a more supportive and impactful approach: victims are listened to, referred to appropriate support services when available, and reassured that their stories have purpose and direction, including the option of involving law enforcement. This sense of validation and potential for impact can be very significant for victims. While online self-referral systems do have their advantages, it is crucial to highlight their limitations and the extent to which they may not fully address victims’ needs for acknowledgment and support.

Furthermore, a practical proposal was put forward to designate specific focal points within each organization. Under this arrangement, one individual would be responsible for handling and participating in the plenary session dedicated to documenting hate crimes, while another would engage in a separate Network plenary session focused on recording hate speech incidents.

In any event, ensuring systematic documentation and long-term sustainability of the project to achieve desired outcomes was identified as a significant challenge. To address this, a proposal was made to implement a pilot plan aimed at uncovering potential practical obstacles, especially given that civil society has already engaged in similar efforts to compensate for the State’s shortcomings in this area.

In this context, the methodology of the ‘Stand-Up – Standing up against hate in the EU¹’ programme was also discussed. This initiative aims to develop a multi-service model for public authorities to tackle hate crimes. It entails developing technological tools to bolster monitoring and data exchange across services. The programme aspires to create an open-source intelligence (OSINT) database that will include comprehensive information on hate crimes and hate speech. This database is expected to detail the specifics of victims, perpetrators, and incidents, as well as compile reports and other relevant data. The project also focuses on developing detailed manuals to define hate crimes and hate speech, establishing clear guidelines for the reporting of hate crimes, and introducing a handbook dedicated to victim support services. Additionally, it involves training activities focused on public authorities, especially law enforcement, the judiciary, and civil servants, alongside awareness-raising initiatives¹.

A final point of consideration was how the Network plans to utilize the data on hate speech, particularly from online sources. In this context, it is crucial for the project to be both sustainable and clearly purpose-driven, ensuring the active support and engagement of civil society and the affected communities. Nonetheless, it was emphasized that non-governmental organizations should not be overburdened with responsibilities for societal reforms and changes that are fundamentally the duty of the State, especially those aimed at protecting vulnerable groups from hate speech and promoting social cohesion.

V. The Proposal

Upon examining established best practice models and the conclusions drawn from focus group discussions, it becomes evident that the current method the Network uses to document racist crimes is not suitable for the systematic tracking of hate speech, both online and in the wider public domain.

The distinction between hate speech and racist violence is critical. Racist violence refers to direct attacks against people based on their identity, which occur in specific contexts and can be precisely documented. This is currently achieved through victim interviews, a method that depends on the cooperation of Network members to facilitate victims’ testimonies. The specific members are already engaged in providing services to targeted groups; Consequently, victims approach these Network members seeking support following incidents of racist violence they have experienced. It is through this process that the incident is acknowledged and then reported as part of the support process.

In the case of hate speech, the conditions are different. Hate speech does not only target specific individuals but potentially also entire communities that represent and bear protected characteristics. As a result, hate speech can manifest at any time and in any public setting, and may not always be immediately noticed by the communities affected or support groups.

Therefore, reporting hate speech is equally as important as reporting racist crimes. However, capturing hate speech requires methodological tools that can accurately assess public discourse, looking specifically for indications of language that deliberately targets or disparages specific groups or individuals.

The proposed action plan lays out a strategy for the Network to assess the requirements for potentially extending its scope to include the monitoring of hate speech as a new operational focus. It also outlines the guiding principles upon which Network coordinators and members are already building the next steps to achieve their objectives. Central to this process were the focus group discussions, in which Network coordinators and members actively participated, providing valuable input for the planned strategy. These discussions not only covered research findings but also the Network’s commitment to establishing a clear definition, a unified methodology, and a specific monitoring framework for hate speech in the public sphere, in a sustainable and effective manner. Therefore, it is evident that a long-term planning horizon of 2-4 years is necessary to address the two main challenges posed by the potential expansion of the Network’s activities in documenting hate speech.

The first challenge concerns necessary processes to be initiated both within the Network and its members in relation to their redefined role, as well as potential collaborations with stakeholders beyond civil society. These endeavors will allow the implementation of an effective and systematic mechanism for recording and addressing the phenomenon. The steps and methodology recommended for the application of this pillar are analyzed in detail below.

The second challenge involves the cost assessment and securing the appropriate resources and/or personnel required to ensure that the project is carried out effectively and that the Network continues to operate smoothly in its enriched role.

These challenges strike at the heart of the Network’s operations and are beyond the research scope of this project. Consequently, the plan draws on the findings from best practices research and focus group discussions and outlines a practical framework. This, in turn, will enable coordinated action based on the Network’s current technical know-how and partnerships, aiming to effectively monitor hate speech and guarantee the initiative’s sustainability.

Regarding the definition of hate speech for the purposes of the Action Plan, discussions with the Network’s members and coordinators revealed a dual focus: legal and social aspects of the issue are both of interest. The groups concluded that the definition should reflect the harmful impact on affected groups, aligning with European human rights and protection standards¹. Moreover, the definition will incorporate criteria to distinguish incidents with potential legal consequences, such as incitement to violence, in accordance with the legal provisions in place.

As for the field on which the proposed methodology will focus, it is deemed essential that the proposed methodology be primarily applied to social media platforms, which have been identified as the central spaces for the articulation and proliferation of hate speech. Additionally, these platforms also accommodate the online presence of established traditional media organizations.

The action plan advocates for a combined recording and reporting system, drawing on practices from Germany and Italy for recording hate speech, and from England for reporting mechanisms. The aim is to develop AI tools for comprehensive monitoring of hate speech, while also enabling manual data collection by trained individuals (reporters). It should be noted that the Network has already trained a significant number of individuals to report incidents of racist violence. In this context, these individuals have been introduced to and have become familiar with the framework for recognizing hate speech during these training sessions. Additionally, the composition of the Network is significant for ensuring the detection and documentation of hate speech incidents. It includes civil society organisations that advocate for specific targeted groups, as well as the communities that bear protected characteristics and are thus subject to targeting. This diverse membership provides a robust assurance that potential hate speech incidents will be accurately identified and recorded by the Network’s members. Therefore, the Network members are inherently valuable as channels for reporting hate speech incidents- a practice that has already been established.

While the Network’s long-established practice of reporting hate speech has been valuable, it is not adequate on its own for a comprehensive analysis of the issue. Hence, it is essential to complement this approach with advanced tools for systematic observation. This enhancement will involve the development of artificial intelligence technologies.

To comprehensively depict hate speech as a societal issue, collecting extensive and detailed data is crucial. This effort requires the use of artificial intelligence (AI) tools, a widely adopted method in large-scale projects globally, to manage and analyze the vast amount of information effectively. The objective of this monitoring is to present hate speech as a systemic problem, highlighting particularly the influence and impact of hate speech propagated by public figures, which underscores the importance of this targeted approach. Collaboration with universities is integral to this effort, bringing in essential expertise in legal, sociological, and political analysis, while computer science labs contribute their skills in data management and preliminary evaluation. Alongside these efforts, the Network will maintain its established recording mechanisms, enabling affected individuals and witnesses to report incidents through a standardized form, thereby ensuring the robustness and depth of the data collected.

To facilitate the adoption of the new monitoring framework for hate speech, member organisations of the Network will feature a specialised banner on their websites. This interactive element will serve as a portal for victims or witnesses of hate speech to report incidents directly into the monitoring system. It is also worth recalling that Network members, including focus group participants, have agreed to regularly communicate this new tool to the communities they support. This communication will occur through the described online approach and by incorporating the new monitoring system into relevant education and information initiatives addressing racism within these communities.

The core component of data collection involves establishing a database that will encompass two types of data: firstly, the incidents of hate speech captured using artificial intelligence tools, and secondly, the incidents reported directly to the Network’s organisations via their platforms. The process for recording and compiling this data can be visualised in the following diagram:

In the second and final stage, the Network’s team will gather data reported via the self-referral forms, along with the data captured by artificial intelligence tools. This step follows an initial assessment or ‘filtering’ conducted by the designated research team. Given the expected increase in data volume and complexity of procedures, this team may require additional resources to manage both monitoring mechanisms effectively. Additionally, it’s important to note that Universities or tech labs interested in contributing to the hate speech recording model should join the Network as members or observers in order for them to be bound by a common operational framework and a code of ethics.

The Network already includes among its members an institution affiliated with a Greek university, while over the past eleven years, the Network has cultivated strong partnerships with academic and research groups dedicated to addressing racism and hate speech. Going forward, the goal is to put these partnerships on a formal footing, which will enable a more organized and unified way of working together in this context.

Additionally, the Network’s assistant coordinator will be essential in analyzing the data on hate speech and hate crimes, identifying any emerging trends as well as where and when these incidents are most likely to occur. This detailed analysis aims to deepen our understanding of the issue, highlight any shortcomings in current institutional responses, and determine where focused strategic efforts are most needed.

The Network aims to not only develop and refine its tools for recording hate speech but also to broaden its monitoring scope beyond incidents of racist violence. By analysing data related to both racist violence and hate speech, the goal is to draw reliable conclusions which will inform specific recommendations for authorities to effectively tackle these manifestations of racism.

VI. Conclusion

The Action Plan at hand was developed for the Racist Violence Recording Network (RVRN), and it aspired to explore the potential of extending the network’s scope into the realm of hate speech monitoring. Drawing on a detailed analysis of European best practices and input from RVRN’s focus groups, this plan addresses the challenges of monitoring hate speech both in physical public spaces and online platforms. The strategy proposed here marries technological innovation with the critical human element to accurately track and document hate speech instances.

The plan adopts a comprehensive definition of hate speech, aligned with the Council of Europe’s General Recommendation No.15, covering both illegal and harmful forms of speech. While this broad definition is crucial for understanding the societal impact of hate speech and its potential to escalate into hate crimes, it also presents practical challenges. Given the complexity and overlapping nature of hate speech, affecting various groups, monitoring tools must be carefully designed. Specifically, forms should be structured to clearly and effectively capture these complexities, ensuring accurate documentation of a phenomenon that is both diverse and, at times, contested.

To address these challenges, the action plan combines automated and manual data collection methods. Drawing inspiration from systems in Germany, Italy, and the UK, it incorporates Artificial Intelligence tools for extensive, real-time monitoring alongside manual data gathering through the RVRN. To ensure data quality and reliability, the plan also includes a structured review process for both AI-driven and human-collected data. This review process will be essential in validating findings, minimizing biases, and maintaining consistency, making the monitoring efforts both comprehensive and methodologically sound.

In conclusion, this Action Plan presents a practical and sustainable framework for the RVRN to expand its activities into hate speech monitoring. It emphasizes the importance of an integrated approach that combines technological tools with human judgment to ensure comprehensive and accurate data collection. Additionally, the Plan calls for collaboration among civil society, academia, and government stakeholders, highlighting that only through coordinated efforts can the complex challenge of documenting and addressing hate speech be effectively managed.

Footnotes: