Echoes of Unrest framework detects fake news and mob violence with 98% accuracy

A new research paper titled 'Echoes of Unrest: A Multimodal NLP Framework for Early Warning of Fake News and Violence-Driven Mob Activity' presents a sophisticated AI system designed to tackle the growing threat of misinformation-fueled social unrest. Developed by Md. Maruf Bangabashi and five other authors (accepted as a Taylor & Francis book chapter, 2026), the framework addresses how false narratives on platforms like Facebook and WhatsApp—especially in South Asia—can trigger real-world mob violence faster than fact-checkers can intervene.

The system combines multiple state-of-the-art AI components: XLM-RoBERTa for multilingual text understanding (Bangla and English), CLIP for visual embedding of images and memes, and a multi-head attention mechanism for multimodal fusion. It also incorporates auxiliary signals such as sarcasm detection and geospatial metadata to identify escalation patterns. Trained on a fused dataset of 138,256 samples, the model achieved 98% test accuracy on a stratified 30% subset, with strong precision and recall. The authors highlight that adding geospatial features significantly improved the system's ability to anticipate real-world violence, making it a potential early-warning tool for authorities and platform moderators.