Thou shalt not hate: Countering Online Hate Speech (1808.04409v2)

Abstract: Hate content in social media is ever-increasing. While Facebook, Twitter, Google have attempted to take several steps to tackle the hateful content, they have mostly been unsuccessful. Counterspeech is seen as an effective way of tackling the online hate without any harm to the freedom of speech. Thus, an alternative strategy for these platforms could be to promote counterspeech as a defense against hate content. However, in order to have a successful promotion of such counterspeech, one has to have a deep understanding of its dynamics in the online world. Lack of carefully curated data largely inhibits such understanding. In this paper, we create and release the first ever dataset for counterspeech using comments from YouTube. The data contains 13,924 manually annotated comments where the labels indicate whether a comment is a counterspeech or not. This data allows us to perform a rigorous measurement study characterizing the linguistic structure of counterspeech for the first time. This analysis results in various interesting insights such as: the counterspeech comments receive much more likes as compared to the non-counterspeech comments, for certain communities majority of the non-counterspeech comments tend to be hate speech, the different types of counterspeech are not all equally effective and the language choice of users posting counterspeech is largely different from those posting non-counterspeech as revealed by a detailed psycholinguistic analysis. Finally, we build a set of machine learning models that are able to automatically detect counterspeech in YouTube videos with an F1-score of 0.71. We also build multilabel models that can detect different types of counterspeech in a comment with an F1-score of 0.60.

Insights into Countering Online Hate Speech

The paper "Thou Shalt Not Hate: Countering Online Hate Speech" investigates the growing issue of hate speech on social media, focusing on counterspeech as a potential solution. Despite efforts by major platforms like Facebook, Twitter, and Google to mitigate hate speech, the outcomes have been largely ineffective. This research provides a new perspective by promoting counterspeech—defined as direct responses that oppose hateful or harmful content—as an alternative to content removal strategies, which may infringe on free speech rights.

Dataset Development and Analysis

The authors present the first dataset specifically curated for counterspeech, sourced from YouTube comments responding to videos containing hate speech aimed at Jews, African-Americans, and the LGBT community. The dataset comprises 13,924 comments, annotated to identify counterspeech, with further categorization into types. Notably, counterspeech comments garnered more likes compared to non-counterspeech, indicating community support or endorsement.

Through psycholinguistic analysis, the research identifies distinct linguistic features in counterspeech. The paper finds that counterspeech comments are more emotionally charged, exhibiting higher frequencies of words related to anxiety, anger, and sadness. Furthermore, language indicating biological processes, such as references to 'body' and 'health,' appeared more frequently in counterspeech, illustrating how commenters may invoke personal or humanistic sentiments to counteract hate.

Machine Learning Models and Classification Tasks

The authors developed machine learning models to automatically detect counterspeech, achieving an F1-score of 0.71 for identifying counterspeech versus non-counterspeech. They extended their models to classify different types of counterspeech, obtaining an F1-score of 0.60. The models demonstrated a notable ability to generalize across communities, evident from cross-community classification tasks with F1-scores ranging from 0.62 to 0.65.

Implications and Future Prospects

This work lays a significant foundation for understanding and leveraging counterspeech in social media platforms. The findings offer practical implications in managing hate speech, suggesting that platforms could benefit from promoting counterspeech through algorithmic support or user training. The dataset and models provide a framework for future exploration into the effectiveness of different counterspeech strategies across diverse communities.

The research highlights the complexity of addressing online hate speech, emphasizing the potential for counterspeech to maintain the balance between community safety and freedom of expression. Future work could explore generating auto-counterspeech, evaluating the change in attitudes resulting from counterspeech interactions, and adapting these strategies to other social media contexts. The broader impact of counterspeech and its role in fostering inclusive online environments remains a promising area for continued investigation in the AI field.