Monitoring · LGBTIQ+ · Hate speech-related risks · Evidence-based approach
Not Just Counting Offensive Words
How NGO Human Rights Bureau “We Are!” develops monitoring of hate speech-related risks concerning LGBTIQ+ people as an evidence-based, proportionate and human rights-based tool.
Monitoring hate speech-related risks is not about collecting as many alarming examples as possible. For us, the key task is not only to document problematic content, but to understand its context, level of risk, potential harm and whether a responsible institutional response exists.
Offensive words matter. But on their own, they do not explain what is happening in public discourse.
One piece of content may contain inaccurate terminology. Another may use a stigmatising frame. A third may reproduce a disinformation narrative. A fourth may include a direct call to violence. These are different levels of risk and they require different responses.
This is why the monitoring approach of NGO Human Rights Bureau “We Are!” is built around precision rather than volume: we distinguish between types of problems, assess levels of severity, minimise repeated harm and clearly define the limits of our conclusions.
Monitoring is not...
- a list of the “worst examples”;
- a public display of hateful content;
- a list of offenders;
- a substitute for judicial assessment;
- a call for censorship;
- a claim about the full prevalence of hate speech in Ukraine.
Our approach is...
- structured discourse analysis;
- a scale of risk levels;
- assessment of potential harm;
- tracking institutional response;
- aggregated indicators without reproducing harmful content;
- an evidence base for responsible action.
Six principles behind the monitoring approach
Distinguishing levels of severity
Not every problematic text is equally harmful. Inaccurate language, stigmatisation, disinformation narratives, threats and calls to violence require different assessment and different responses.
Risk scaleNot treating every mistake as hate speech
A professional or ethical mistake may be harmful, but it does not always reach the threshold of hate speech in the narrower sense. We therefore distinguish biased reporting, stereotyping, humiliation, incitement, threats and other types of risk.
Precision of categoriesEmbedding standards into the methodology
International standards matter not as a formal reference, but as a practical logic of analysis: context, protected characteristic, severity, potential harm and proportionality of response.
Council of Europe standardsMinimising repeated harm
A public dashboard should not become an archive of offensive phrases or aggressive visuals. We use neutral descriptions, internal labels, partial masking and aggregated indicators.
Do no harmWorking with verifiable data
Each case has a structure: date, platform, source type, format, description, target group, narrative code, severity level, harm indicators, link status and response status.
ReproducibilityDefining methodological limits clearly
The monitoring corpus does not show the full prevalence of hate speech in Ukraine. It shows recurring patterns, visible narratives, types of risk and response gaps within the collected data.
Methodological clarityThe approach is informed by Council of Europe standards on combating hate speech, non-discrimination, freedom of expression and proportionate institutional response.
Three streams of evidence-based analysis
We analyse more than the content itself. Human rights-based monitoring also needs to understand potential harm and whether a response by a responsible actor has taken place.
Content stream
What was said, where, by whom, in which format and concerning which group.
Harm and risk stream
What potential impact this may have for LGBTIQ+ people, other vulnerable groups and the wider social environment.
Institutional response stream
Whether there was a response by a media outlet, platform, regulator, public authority or another responsible actor.
Different statements, different levels of risk
The scale is used for analytical risk assessment. It is not a legal qualification of specific content and it does not replace decisions by competent authorities.
Monitoring should not amplify what it records
Hate speech-related monitoring involves a difficult ethical choice: to show a problem, it has to be described. Yet unnecessary repetition of offensive phrases, screenshots, memes or aggressive visuals can turn monitoring itself into another channel for circulating harmful content.
- we do not reproduce harmful content without a clear need;
- we use neutral descriptions;
- we mask offensive words where appropriate;
- we show aggregated indicators;
- full evidence is kept only in a closed research dataset when needed for verification.
What stands behind a single monitoring record
Behind each aggregated indicator there is a structured record. This makes it possible to verify data, compare cases and avoid conclusions broader than the corpus can support.
- date
- platform
- source type
- content format
- neutral description
- target group
- narrative code
- hate speech code
- severity level
- harm indicators
- confidence level
- link status
- archiving
- response status
- verification notes
- Date
- 2026-04-12
- Platform
- web
- Format
- article
- Target
- LGBTIQ+
- Description
- material with a stigmatising frame concerning equality of rights
- Severity
- 2
- Harm
- discrimination_risk
- Response
- no_response_found
What monitoring shows
- recurring patterns;
- visible narratives;
- types of risk;
- levels of severity;
- gaps in institutional response;
- dynamics within the collected corpus.
What monitoring does not claim
- the full prevalence of hate speech in Ukraine;
- legal guilt of individual persons;
- automatic illegality of every recorded material;
- a need for censorship;
- a final assessment of the entire media or social environment.
For us, monitoring hate speech-related risks is a tool of precision, not volume. It should help distinguish between a mistake, bias, stigmatisation, incitement and a real threat.
This distinction is what makes a human rights-based response serious: without exaggeration, without censorship and without repeated harm.
We work with data so that response can be responsible
Explore the public dashboard, report a potential case or read the monitoring methodology.