Offline Harm Potential

Offline harm potential research focuses on identifying and mitigating the risk of real-world violence stemming from online interactions, particularly on social media. Current research employs multimodal approaches, leveraging natural language processing and computer vision techniques (like CLIP-ViT) within machine learning models (including ensemble methods and simpler classifiers like SVMs) to analyze text and images for indicators of potential harm. This work is crucial for developing effective tools to prevent offline violence and improving online safety, with applications ranging from content moderation to personalized interventions designed to reduce individual risk.