For a new seminar of the DAML Pole of LIX, we are happy to host Dr. Benjamin Ocampo, a Postdoctoral researcher at the Human-Centered Data Analytics research group of CWI in the Netherlands. Benjamin will present his work on “When Hate Meets Facts: Exploring the Intersection of Hate Speech and Check-Worthiness”.
The seminar will take place on Thursday, 16th of October, 10:00 – 11:00 , at the Henri Poincaré room (RDC) of the Alan Turing building. For those unable to join us in person, the seminar can be attended online, here.
Title
When Hate Meets Facts: Exploring the Intersection of Hate Speech and Check-Worthines
Abstract
We study how hate speech relates to factual information and check-worthiness. Our goal is to understand when hateful messages also contain claims that could be verified. We use large language models (LLMs) to identify sentences that are likely to be check-worthy and compare their results with human annotations. When the models and humans disagree, we use another LLM as a judge to decide which answer is more accurate. With the resulting labels, we explore how often hate speech includes factual claims and whether there is a clear connection between hateful language and factual content. We also extend this analysis to other datasets to see if the patterns we find are consistent. Our work aims to better understand the overlap between harmful and factual language and to evaluate how well LLMs can support fact-checking and moderation tasks.
