Using machine learning to investigate potential image bias in news articles

Abstract

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 through a number of ways, one of which is image use. Media bias research is hard to perform without manual human intervention as what constitutes bias in a news article is not always clear, especially to a machine. Thus, this work aims to develop an automatic technique at investigating picture‐related bias in online newspaper articles, by integrating various machine learning models into the methodology, such as BLIP a Vision Language Pre‐training model. We scraped six online Maltese newspapers in order to demonstrate this technique. The demonstration achieved promising results for the adoption of this methodology by media bias researchers. This work can be used to fact‐check older studies which did not have access to such technology, as well as a way to alleviate some of the manual work performed by researchers in picture‐related media bias studies.