Assisted Counterspeech Writing at the Crossroads of Hate Speech and Misinformation

2026-05-21

AI Summary

This study addresses the intertwined prevalence of hate speech and misinformation in online environments, which jointly exacerbate social bias and polarization. It presents the first unified approach to jointly counter both phenomena by introducing a hybrid knowledge-guided strategy that integrates fact-checking resources with guidelines from non-governmental organizations. Leveraging large language models, the method generates rebuttals that simultaneously correct factual inaccuracies, mitigate stereotypes, and express empathy. Through expert revision and crowdsourced evaluation, the generated responses demonstrate significant improvements in naturalness, completeness, and adherence to normative standards, with 40% of initial outputs meeting usability criteria. The work also releases the first dataset comprising expert-validated rebuttals alongside their supporting knowledge sources.

Abstract

Hate speech and misinformation frequently co-occur online, amplifying prejudice and polarization. Given their scale, using Large Language Models (LLMs) to assist expert counterspeech (CS) writing has gained interest, yet prior work has addressed these phenomena separately. We bridge this gap by studying CS generation in contexts where both hate and misinformation co-occur. We test three knowledge-driven generation strategies: first we prompt an LLM with fact-checkers' guidelines and fact-checking articles; secondly, with NGOs' guidelines and reports; thirdly, we create a mixed strategy that combines guidelines and documents from both. 23 experts revise the generated CS, which are assessed via human and automatic metrics. While LLMs produce adequate CS in 40% of cases, expert edits substantially improve naturalness, exhaustiveness, and adherence to guidelines. Based on the post-edited CS, the mixed strategy proves to be the most effective in crowdsourcing evaluation, pairing strong factual correction with stereotype mitigation and empathetic engagement. We release a dataset of hateful and misinformed claims with expert-verified CS and supporting knowledge.

Problem

• hate speech
• misinformation
• counterspeech
• Large Language Models
• co-occurrence

Innovation

• counterspeech generation
• hate speech and misinformation co-occurrence
• knowledge-driven LLM prompting
• mixed strategy
• expert-verified dataset