Evidano is an AI-powered qualitative data analysis platform that helps researchers analyze interviews, open-ended surveys, and documents. This post explains how AI-enabled qualitative analysis can surface patterns in anti-First Nations online hate, using the Tackling Hate Lab study reported in The Conversation Africa on 6 August 2026 as the worked example. The primary keyword for this piece is ai qualitative analysis of online hate, and the following sections translate the study's methods and statistics into practical steps qualitative teams can take.
Key Takeaways
According to The Conversation Africa (6 August 2026), researchers at the Tackling Hate Lab analysed more than three million posts and news items to map anti-First Nations online hate and its links to political events.
- The Tackling Hate Lab dataset covered more than three million posts and news items collected across platforms, according to The Conversation Africa on 6 August 2026.
- During the 2023 Voice campaign daily detected hateful posts rose from 200–400 in early 2023 to regularly above 2, 000 in September 2023 and peaked at about 3, 600 per day in early October 2023, according to The Conversation Africa.
- Around Anzac Day and the March for Australia rallies in 2025, more than 96% of detected hate directly targeted First Nations identity and white supremacist or dehumanising language made up almost 90% of hateful posts, according to The Conversation Africa.
- Call It Out receives roughly 500 reports per year and in 2023–24, 28% of reports referenced the Voice campaign, according to Call It Out and The Conversation Africa.
What Happened: how the study measured online hate
Answer: The Tackling Hate Lab combined automated toxicity scoring with community-labelled training data to identify anti-First Nations hate across platforms.
According to The Conversation Africa (6 August 2026), the researchers captured content from X, news media, Reddit, YouTube, Facebook, forums and other online platforms and processed more than three million posts and items.
According to The Conversation Africa, the study used Google’s Perspective API to flag potentially harmful language and "we used a bespoke AI model trained using judgements from First Nations community annotators to identify anti-First Nations hate, " a phrase used by the authors to describe their human-in-the-loop approach.
According to The Conversation Africa, the authors emphasise that community annotation matters because deciding what counts as racism must reflect the experiences of those targeted.
Findings snapshot
| Date or Event | Metric | Value | Implication |
|---|---|---|---|
| Dataset (study overview, reported 6 Aug 2026) | Total posts and news items analysed | More than 3, 000, 000 | Enables large-scale pattern detection across platforms, according to The Conversation Africa |
| Voice campaign (2023) | Typical daily hateful posts early 2023 | 200–400 posts per day | Baseline hostility before campaign spikes, according to The Conversation Africa |
| Voice campaign (Sept–Oct 2023) | Daily hateful posts at peak | Regularly >2, 000 in Sept 2023; ~3, 600 in early Oct 2023 | Political moments amplify hate, according to The Conversation Africa |
| March for Australia (2025) | Share of detected hate that was white supremacist or dehumanising | ≈69% (March for Australia); almost 90% around Anzac Day/March period | Event-linked mobilisation of white supremacist narratives, according to The Conversation Africa |
| Call It Out reporting (2023–24) | Reports referencing the Voice campaign | 28% of reports | Online debate translates into increased reporting and harm, according to Call It Out and The Conversation Africa |
Implications for qualitative researchers
Answer: Qualitative researchers should combine community-grounded annotation, event-aware sampling, and mixed AI-human workflows to reliably study online hate.
According to The Conversation Africa, the Tackling Hate Lab found hate patterns cluster around political events, so researchers should implement time-based sampling aligned to campaign milestones rather than uniform random sampling.
According to The Conversation Africa, the study’s use of First Nations annotators demonstrates that localised, community-labelled training data improves validity and reduces misclassification of everyday political disagreement as hate.
According to The Conversation Africa and Call It Out, qualitative analysis must also track downstream harms because online narratives spilled into offline attacks and threats in 2025, implying ethical risk mitigation and researcher duty of care are required.
How Evidano Helps: AI-enabled qualitative research for online hate studies
Problem: large noisy datasets and event-driven spikes
Solution: automated ingestion and event-aware filtering speeds discovery, allowing researchers to focus on substantive coding rather than collection.
Evidano is an AI-powered qualitative data analysis platform that helps researchers analyze interviews, open-ended surveys, and documents, and it can ingest social data and news feeds to isolate time windows around political events.
Use the Evidano features page to learn about platform connectors and automated preprocessing.
Problem: context-sensitive definitions of hate
Solution: human-in-the-loop training and community annotation support produce bespoke classification models aligned to target communities.
According to The Conversation Africa, the Tackling Hate Lab trained a bespoke AI model with First Nations annotators, a practice Evidano supports through custom dictionaries and labeled training workflows that record annotator metadata and inter-annotator agreement.
Problem: verifying claims and producing quotable evidence
Solution: combined thematic, frequency and co-occurrence analyses produce extractable statistics and annotated exemplar quotes for reports.
Evidano’s analytics export includes frequency counts and coded exemplar extracts so qualitative teams can present dated, source-attributed evidence similar to the Tackling Hate Lab’s quoted statistics.
Problem: ethical handling of sensitive data
Solution: encrypted storage, PII redaction and controlled user access reduce research risk.
Evidano supports transcription with PII redaction and secure sharing features, aligning with good research ethics for sensitive content.
FAQ: ai qualitative analysis of online hate
How did the Tackling Hate Lab identify anti-First Nations hate?
Answer: They combined automated toxicity flags with a bespoke AI model trained on First Nations community annotations.
According to The Conversation Africa, the study used Google’s Perspective API to flag harmful language and then used a bespoke model trained with community judgements to classify anti-First Nations hate.
Can AI reliably detect contextual racism versus policy disagreement?
Answer: AI can help, but community annotation is essential to distinguish racism from legitimate policy critique.
According to The Conversation Africa, the Tackling Hate Lab explicitly did not classify all opposition to the Voice as hate and relied on First Nations annotators to identify language that denied colonisation or framed recognition as a threat.
What sampling strategy should researchers use for event-driven hate analysis?
Answer: Use event-aware sampling that increases capture density around key campaign milestones and media moments.
According to The Conversation Africa, the study detected a "staircase" pattern where each campaign moment increased hostility, so concentrated sampling around launches, debates and rallies improves sensitivity to spikes.
How should researchers present statistics to policymakers?
Answer: Present dated, source-attributed counts and clear definitions of hate alongside exemplar quotes.
According to The Conversation Africa, the Tackling Hate Lab reported absolute daily counts, percentage shares for categories like white supremacy (for example 69% during March for Australia) and direct quotes such as "Racism does not suddenly appear. It builds through repeated stories about who belongs, " which aid policymaker understanding.
Conclusion & Next Steps
The Tackling Hate Lab analysis reported in The Conversation Africa shows that AI-enabled qualitative analysis reveals event-linked patterns in anti-First Nations online hate and that community-grounded annotation improves validity.
Researchers should adopt event-aware collection, human-in-the-loop models, and clear dated statistics when translating findings into policy or moderation recommendations, according to The Conversation Africa and Call It Out.
If you are running qualitative research on online harm, consider tools that combine secure ingestion, community annotation and extractable analytics.
To explore these capabilities, Try Evidano for free.
Topics
- ai qualitative analysis of online hate
- qualitative analysis of online hate australia
- AI-enabled qualitative research
- online hate analytics
Keep reading
- Commentary on NewsAI Qualitative Analysis: Data Centers and Indigenous ResistanceHow to use AI qualitative analysis to study data-center-driven data colonialism, with methods and stats from Truthout on August 6, 2026. Practical workflows.
- Commentary on NewsAI Qualitative Analysis for Snakebite PreventionTurn 142 studies into actionable insight: AI qualitative analysis of snakebite circumstances extracts behaviours, environments, and occupations for targeted prevention.
- Commentary on NewsQualitative analysis of water insecurity in MaliPractical guide to qualitative analysis of water insecurity: what the PLOS Water Mali study measured, key statistics, and AI-enabled research workflows. Learn next steps.