Navigating Hate and Anti-Hate Speech: Bridging Large Language Model and Human Expertise in Public Officials’ Online Communication

The rise in hate speech targeting minority communities underscores the urgent need for effective tools to detect and address harmful content in digital communication. We examine over 3 million tweets posted by state legislators between 2020 and 2021, focusing on messages directed at Asian communities. To address the nuanced nature of hate speech, we develop three comprehensive definitions for identifying hate speech. With human-in-the-loop, our fine-tuned BERT-NLI model achieved improved classification performance. We find that while anti-Asian tweets comprise only a small portion of legislators’ total tweets, there are distinct geographic patterns across states. In addition, the frequency of posting Pro-Asian or Anti-Asian tweets is significantly influenced by legislators’ demographics. Women and Democrats post more Pro-Asian content while men and Republicans post more Anti-Asian content. By combining advanced computational methods with human oversight, this study advances efforts to address sensitive issues in digital discourse with greater precision and accountability.