Detecting Misinformation: Identifying False News Spread by Political Leaders in the Global South

Abstract

We provide and examine an approach for detecting false stories that circulate as text and without hyperlinks, which are commonly found in the Global South. Our text-based approach relies on a combination of false stories identified by fact-checkers, supervised learning methods, natural language processing, and human review. We contrast our approach with the established domain-based and with Facebook’s URL approaches by applying them in the case of Brazilian political leaders. The results show that sharing false news by politicians is a rare event: less than 1% of political leaders’ social media posts contain misinformation. However, we find little overlap across the approaches. The text-based approach leads to different conclusions about which politicians share misinformation and the type of false content shared, while demographic and political predictors of misinformation-sharing behavior are typically similar across approaches. Our approach produces fewer false positives than other approaches and only a small number of false negatives. Our results show that the text-based approach is an important complement to the dominant approaches as it is more effective at detecting false news.