MisInfoWars: A linguistic analysis of deceptive and credible news

Misinformation, bias, and deceit, clandestine or not, are a pervasive and continual problem in media. Real-time mass communication through online media such as news outlets, Twitter, and Facebook, has extended the reach of deceptive information, and increased its impact. The concept of fake news has existed since before print, but has acquired renewed attention due to its perceived influence in the 2016 U.S. Presidential election. Previous studies of fake news have revealed much about why it is produced, how it spreads, and what measures can be taken to combat its rising influence. Despite the continued interest in fake news, current research on the language of deceptive media has been largely superficial. This thesis serves to provide a profound understanding of the stylistic and linguistic features of fake news by comparing it to its credible counterpart. In doing so, it will advocate for differentiation between disingenuous and respectable media based on linguistic variation. With a dataset of approximately 80,000 articles from known fake and legitimate news sources, specific stylistic differences will be examined for saliency and significance. Using multidimensional analysis for discourse variation established by Biber (1988), this thesis will confirm that there exist sufficient textual differences between the articles of fake news and credible news to consider them distinct varieties. Detecting misinformation has not proven to be simple, neither has minimizing its reach. As the ambition of fake news articles is to appear authentic, acquiring knowledge of the subtleties which serve to discriminate realism from fabrication is crucial. A better understanding of the linguistic composition of deception and fabrication in comparison to credibility and veracity will facilitate future attempts at both manual and automatic detection.