https://analista.in/10.71182/aijmr.2512.0302.5005

---

Abstract

The growth of digital communication technologies has made it easier to pose threats to one’s psychological well-being and even one’s safety due to cyberbullying and online toxicity. AIs have become one of the few technologies capable of mitigating the problem of harmful interactions and toxicity digitally via automated content moderation. The purpose of this review is to assess the development of AI-enabled systems in the automated moderation of cyberbullying and toxicity to digitally detect comments and to evaluate how the systems have adapted from using rules and algorithms to more advanced systems using deep learning and other new digital technologies.
This paper also reviews the criticisms and challenges surrounding benchmark datasets, evaluation research, text and discourse structures and even inconsistency and disregard for key AI ethics in cyber moderation, responsibility, algorithm bias, explanatory disaggregation, and even digital-privacy. We also examine new concepts of research in context integrating multimodal compositions for discourse, improvement of emotion, and computational efficiency in new forms of deep learning architectures. This advances digital moderation using low computational resources. This comprehensive review is intended to detail, critique, and analyze the breadth of research in cyberbullying and online toxicity.

Keywords: Cyberbullying Detection, Toxic Content Classification, Natural Language Processing, Machine Learning, Deep Learning, Transformers, Multilingual AI

Download PDF