Higher Ground? How Groundtruth Labeling Impacts Our Understanding of Fake News About the 2016 U.S. Presidential Nominees

Abstract

The spread of fake news in social media platforms has garnered much public attention and apprehension. Consequently, both the tech industry and academia alike are investing increased effort to understand, detect, and curb fake news. Yet, researchers differ in what they consider to be fake news sites. In this paper, we first aggregate 5 and 3 distinct lists of fake and mainstream news sites published by experts and reputable organizations. Then, focusing on tweets about the democratic (Hillary Clinton) and republican (Donald Trump) presidential nominees, we use each pair of fake and traditional news lists as an independent to examine i) the prevalence, ii) temporal characteristics and iii) the agenda-setting differences between fake and traditional news sites. We observe that depending on the groundtruth, the prevalence of fake news varies significantly. However, the temporal trends and agenda-setting differences between fake and mainstream news sites remain moderately consistent across different groundtruth lists.