AI Techniques and Applications for Online Social Networks and Media: Insights from BERTopic Modeling

Abstract

This study examines the role of Artificial Intelligence (AI) in enhancing personalization, analyzing information dynamics, and developing scalable methodologies within Online Social Networks and Media (OSNEM), with a focus on user protection. Through a systematic review using the PRISMA framework and BERTopic modeling, key AI applications in OSNEM were identified, including fake news detection, sentiment analysis, hate speech detection, big data analysis, bot detection, and insights into public health, disaster relief, and mental health. Although AI techniques and multimodal frameworks have significantly improved content personalization, challenges like algorithmic bias and echo chambers remain. To address these, the implementation of fairness-aware learning models is recommended to ensure personalization stays ethical. Advanced AI techniques, such as Dynamic Memory Networks and Temporal Convolutional Networks, have shown strong capabilities in tracking opinion dynamics and combating misinformation. Additionally, Generative AI offers opportunities for content creation but also raises concerns about misinformation, requiring robust moderation frameworks. Emerging technologies like Artificial Real Intelligence (ARI), which simulate human reasoning and decision-making, could further improve the management of complex online interactions. The study highlights the need for scalable AI methodologies, such as multitask learning frameworks, to efficiently handle the vast amounts of real-time data generated by social media while addressing cross-platform adaptability and computational efficiency.