AUTOMATING VISUAL ETHICS: A CONVOLUTIONAL NEURAL NETWORK FRAMEWORK FOR CLASSIFYING JOURNALISTIC PHOTO SUITABILITY

Abstract

The rapid digitization of journalism has exponentially increased the volume of visual content processed by newsrooms, making manual ethical review of photographs a significant bottleneck prone to subjectivity and inconsistency. This research proposes and evaluates a Convolutional Neural Network (CNN)-based framework designed to automate the classification of journalistic photographs as either "suitable" or "unsuitable" for publication based on visual ethical criteria. Leveraging an annotated image dataset, the proposed CNN model was trained and tested to identify visual elements that potentially violate journalistic codes of ethics. The experimental results demonstrate the model's robust performance, achieving a weighted average accuracy of 86% and an F1-score of 0.86. The model exhibited particularly strong performance in identifying suitable photos, with precision, recall, and F1-scores ranging from 0.88 to 0.89, while maintaining a solid F1-score of 0.81 for the unsuitable class. These findings confirm the substantial potential of CNN-based systems as efficient, objective decision-support tools for editorial workflows. The implementation of such a framework not only accelerates the content filtering process but also enhances adherence to professional ethical standards in digital media by minimizing the risk of publishing ethically questionable visual content.