Comparative Analysis of Artificial Intelligence Techniques for the Detection of School Violence

Abstract

School violence is a growing phenomenon that affects the overall well-being of students, manifesting in forms such as verbal, physical, social, digital, socioeconomic, and sexual violence. In response to this issue, artificial intelligence has emerged as a promising tool to strengthen prevention and detection strategies. This study aims to conduct a comparative analysis of three artificial intelligence-based approaches applied to the detection of school violence, evaluating their effectiveness, advantages, and limitations. A systematic literature review was carried out using the PRISMA protocol, focusing on articles indexed in the Scopus database. Three studies were selected, each addressing a different input modality: text, audio, and video. The results show that computer vision techniques offer higher accuracy in the direct detection of violent acts; natural language processing enables early identification of risk behaviors through linguistic analysis; and voice recognition proves useful for real-time immediate responses. The conclusions emphasize that integrating these techniques into hybrid systems, supported by clear ethical policies and active participation from the educational community, can improve prevention and intervention processes. Finally, remaining challenges and future work are identified, particularly in areas related to privacy, system adaptability, and contextual precision.