How Media Necessarily Bias the News via the Selective Reporting of Events

• In Event: Global Media Flows and Impacts

Abstract

Most existing studies of media bias measure overt ideological language or framing effects. However, many mainstream media eschew overt ideological language, adopting a more objective and neutral tone. But even the most neutral news outlet must still decide which which events to include in a news story, and these myriad decisions are necessarily moral and ideological, with significant effects on readers. We develop a suite of new, state-of-the-art large language model NLP methods for measuring the ingredients of event-level selection: the partisanship of individuals and other entities; the pool of events in a general news story, and the subsets selected for individual articles; and the sentiments and moral values expressed alongside entities and events. We apply our model first to a paired set of corpora -- articles on the wars in Ukraine and in Gaza -- and show at scale how different media selectively report different events in ways that reflect and promote different moral and ideological values, even when their overt language remains scrupulously neutral. We also apply the method to a less controversial topic, economic news, to show that even here there are similar selection effects. Our tools can be deployed at scale to show how the selective reporting of events shapes the media and its consumers.

Authors

• Nicholas Beauchamp, Northeastern University

• Xinliang Frederick Zhang

• Winston Wu, University of Hawaii at Hilo

• Kaijian Zou, University of Michigan

• Lu Wang, University of Michigan