3IA PhD/Postdoc Seminar #41

Detecting and Countering Hate Speech Messages on Social Media

Abstract: In today’s digital age, the huge amount of abusive content and hate speech on social media platforms presents a significant challenge. Natural Language Processing (NLP) methods have focused on detecting explicit forms of hate speech, often overlooking more nuanced and implicit instances. To address this gap, we aim to enhance the detection and understanding of implicit and subtle hate speech. Once hate speech has been detected, the most promising method to counter-act it is to use counter-speech. The potential effectiveness of counter-speech as a hate speech mitigation strategy is attracting increasing interest in the NaturalLanguage Generation research community, particularly towards the task of automatically producing it. However, automatically generated responses often lack the argumentative richness which characterises expert-produced counter-speech. For this reason, we focus on two aspects of counter-speech generation to produce more cogent responses. First, by investigating the tension between helpfulness and harmlessness of LLMs, we test whether the presence of safety guardrails hinders the quality of the generations. Secondly, we assess whether attacking a specific component of the hate speech results in a more effective argumentative strategy to fight online hate.

Domain-agnostic segmentation of brain MRI scans

Open discussion about the two contributions