Outcome-Constrained Large Language Models for Countering Hate Speech

arXiv:2403.17146v1 Announce Type: new
Abstract: Counterspeech that challenges or responds to hate speech has been seen as an alternative to mitigate the negative impact of hate speech and foster productive online communications. Research endeavors have been directed to using language models for the automatic generation of counterspeech to assist efforts in combating online hate. Existing research focuses on the generation of counterspeech with certain linguistic attributes, such as being polite, informative, and intent-driven. However, it remains unclear what impact the …

abstract arxiv challenges communications cs.cl hate speech impact language language models large language large language models negative productive research speech type