Representation in Public Media: Six Years of Continuous Coverage of BBC News

• In Event: Automated Annotation of Text, Image, and Video Data

Abstract

In many modern democracies, public media have a large market share of news consumption. The way politicians are represented in media outlets can affect public opinion. Thus, public media may have incentives to differently represent political figures. To limit this bias, existing legislation regulates the time of exposure of politicians in public media. However, the legislation is limited to the time of exposure, and it does not consider that the type of exposure can be biased as well. In this paper, we investigate the qualitative aspect of public media bias. We analyze continuous daily coverage of two main news programs of the BBC News over a period of more than six years covering two election cycles, for a total of more than 2500 hours of video and 5 million image frames. We utilize artificial intelligence facial recognition software to identify individuals in public media news videos and measure the time and type of exposure and possible bias in multiple ways, including emotional state and personal smile, and location, size, and occlusion of the face. Our paper has important implications for the quality of political representation in public media and can be used as a framework for new legislation regulating public media.

Authors

• Cantay Caliskan, University of Rochester

• Alessio Albarello, Princeton University