Multi3Hate: Multimodal, Multilingual, and Multicultural Hate Speech Detection with Vision-Language Models

TL;DR

This paper introduces Multi3Hate, a multilingual, multicultural dataset of memes for hate speech detection, and evaluates vision-language models' ability to understand cultural nuances in hate speech across five languages.

Contribution

It creates the first multimodal, multilingual hate speech dataset with multicultural annotations and analyzes how cultural backgrounds influence model performance and annotation agreement.

Findings

Cultural background significantly impacts hate speech annotation.

Models align more with US annotations than others, regardless of meme language.

Annotator agreement varies greatly across cultures, with only 67-74% agreement.

Abstract

Warning: this paper contains content that may be offensive or upsetting Hate speech moderation on global platforms poses unique challenges due to the multimodal and multilingual nature of content, along with the varying cultural perceptions. How well do current vision-language models (VLMs) navigate these nuances? To investigate this, we create the first multimodal and multilingual parallel hate speech dataset, annotated by a multicultural set of annotators, called Multi3Hate. It contains 300 parallel meme samples across 5 languages: English, German, Spanish, Hindi, and Mandarin. We demonstrate that cultural background significantly affects multimodal hate speech annotation in our dataset. The average pairwise agreement among countries is just 74%, significantly lower than that of randomly selected annotator groups. Our qualitative analysis indicates that the lowest pairwise label agreement-only 67% between the USA and India-can be attributed to cultural factors. We then conduct experiments with 5 large VLMs in a zero-shot setting, finding that these models align more closely with annotations from the US than with those from other cultures, even when the memes and prompts are presented in the dominant language of the other culture. Code and dataset are available at https://github.com/MinhDucBui/Multi3Hate.