A Web-Based Multi-Modal AI Framework for Intelligent Content Moderation with Constructive Feedback

Corresponding Author ORCID Identifer

Naveen Palanichamy https://orcid.org/0000-0003-4601-9770
Subhashini Gopal https://orcid.org/0000-0001-9486-9732

Section

Regular Issue

Abstract

Traditional content moderation systems primarily rely on strict measures that focus on removing offending or inappropriate content without offering users any form of educational feedback or guidance. This lack of constructive interaction often leads to a poor user experience, leaving users frustrated and uninformed regarding the reasons behind content removal. Moreover, such a system fails to encourage long-term positive behavioural changes within online communities, as users do not gain insights on how to improve their conduct. This paper proposes a new constructive Artificial Intelligence (AI) intervention system that integrates a number of AI models in order to conceptualize content moderation as a limiting process rather than an educational process. The framework suggested in the paper combines HuggingFace's BERT-based toxicity analysis, SightEngine's image processing, and Google's Gemini large language model to provide end-to-end content moderation with construction of feedback. Based on comprehensive testing through user interaction analysis and comparative analysis, the study demonstrates the effectiveness of educational intervention in content moderation. The study advances the field by developing a novel approach to content moderation that puts user education and community development first while upholding safety regulations.

Cite This Article

Mohammed Al-Helali, M. Z., Palanichamy, N., Haw, S.-C., & Gopal, S. A Web-Based Multi-Modal AI Framework for Intelligent Content Moderation with Constructive Feedback. Journal of Informatics and Web Engineering, 5(3), 165–177. https://doi.org/10.33093/jiwe.2026.5.3.10