A Hybrid Multimodal Graph Neural Network for Explainable Fake News Detection on Social Media

The rapid spread of misinformation across social media platforms has created significant problems for information integrity and public trust. This research presents a new hybrid multimodal graph neural network architecture that combines text, images, and social network patterns to identify fake news articles while explaining its reasoning. Our proposed system uses heterogeneous graph construction techniques to map out the complex relationships between users, the content they share, and how it spreads. By using attention mechanisms and graph convolution, the system achieves excellent classification results and provides human-readable explanations by highlighting key subgraphs and features. Experiments on three benchmark datasets show that our approach achieves 92.4% accuracy, outperforming existing stateof-the-art methods while offering meaningful visual and textual explanations. These explanations help identify the specific propagation patterns and content features that separate real news from fake, giving content moderators and platform administrators the tools they need to make better decisions.