Differences in health news from reliable and unreliable media

Abstract

The spread of 'fake' health news is a big problem with even bigger consequences. In this study, we examine a collection of health-related news articles published by reliable and unreliable media outlets. Our analysis shows that there are structural, topical, and semantic patterns which are different in contents from reliable and unreliable media outlets. Using machine learning, we leverage these patterns and build classification models to identify the source (reliable or unreliable) of a health-related news article. Our model can predict the source of an article with an F-measure of 96%. We argue that the findings from this study will be useful for combating the health disinformation problem.