Using Machine Learning to Investigate Potential Image Bias in News Articles

• Published in: 2024 IEEE 22nd Mediterranean Electrotechnical Conference (MELECON) pp. 174 - 179
• Main Authors: Hili, Gabriel, Seychell, Dylan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 25.06.2024
• Subjects: Computer Vision; keyword extraction; Manuals; Media Monitoring; NER tagging; News Analysis; Philosophical considerations; Pipelines; Sentiment analysis; Technological innovation; Transforms; Visualization; VLP
• DOI: 10.1109/MELECON56669.2024.10608508

Media bias refers to the deviation from objective reporting in media, where journalists introduce external beliefs or agendas into the journalistic process, altering the perception of an event or issue. A newspaper article may introduce bias in numerous ways, including image use. Analysing media bias manually is challenging and time-consuming, as determining bias requires nuanced human judgement. This paper proposes a simple yet effective technique for investigating picture-related bias in online newspaper articles by utilising BLIP (Bootstrapping Language-Image Pretraining), a Vision-Language Pretraining model. We scraped six online newspapers to demonstrate this technique, achieving promising results for adopting this methodology to automate media bias research. The technique enables large-scale analysis of historical articles to uncover previously undetected biases, complementing modern studies reliant on manual work. It also facilitates monitoring to maintain news integrity. Furthermore, this research highlights innovative applications of AI for journalism to augment human analysis. This study illustrates how AI can transform the media landscape by automating bias analysis to improve news quality and reader trust.