Unpacking Hateful Memes: Presupposed Context and False Claims

2025-10-10

AI Summary

Existing hate meme detection methods primarily optimize classification accuracy while neglecting the underlying causal mechanisms. This work, grounded in philosophy and cognitive psychology, identifies two essential characteristics of hateful memes for the first time: *presupposed context* and *false claims*. To address these, we propose SHIELD—a novel cross-modal framework comprising two core modules: (1) the PCM module, which jointly models image-text presupposed context via multimodal alignment; and (2) the FACT module, which performs fact-checking by integrating external knowledge bases and a cross-modal coreference graph. Evaluated on multiple benchmark datasets, SHIELD achieves significant improvements over state-of-the-art methods. Further ablation and transfer experiments demonstrate that its learned semantic representations generalize effectively to downstream tasks such as fake news detection. This work establishes a new paradigm for interpretable, traceable, and causally grounded multimodal harmful content identification.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

While memes are often humorous, they are frequently used to disseminate hate, causing serious harm to individuals and society. Current approaches to hateful meme detection mainly rely on pre-trained language models. However, less focus has been dedicated to extit{what make a meme hateful}. Drawing on insights from philosophy and psychology, we argue that hateful memes are characterized by two essential features: a extbf{presupposed context} and the expression of extbf{false claims}. To capture presupposed context, we develop extbf{PCM} for modeling contextual information across modalities. To detect false claims, we introduce the extbf{FACT} module, which integrates external knowledge and harnesses cross-modal reference graphs. By combining PCM and FACT, we introduce extbf{ extsf{SHIELD}}, a hateful meme detection framework designed to capture the fundamental nature of hate. Extensive experiments show that SHIELD outperforms state-of-the-art methods across datasets and metrics, while demonstrating versatility on other tasks, such as fake news detection.

Problem

• Detecting hateful memes using presupposed context analysis
• Identifying false claims in memes with external knowledge
• Developing multimodal framework to capture hateful meme characteristics

Innovation

• PCM models contextual information across modalities
• FACT integrates external knowledge for verification
• SHIELD combines PCM and FACT for detection