The disinformation pandemic : understanding, identification, and mitigation in COVID-19 era

In 2020 during COVID-19, in addition to the spread of coronavirus disease, we also observed a pandemic of disinformation about the disease. This pandemic of disinformation became known as Infodemic in the medical world. Just as coronavirus was infecting our bodies, Infodemic was infecting our information ecosystem and exasperating the fight against the COVID-19 pandemic. Disinformation can be produced by various sources including scientists, media personalities, and others and it can be disseminated by news media, webpages, and social media from one source to another. Additionally, disinformation can spread easily from web media to social media where it can spread even faster to a wider audience. Therefore, it is important that disinformation be detected before it has a chance to spread. However, the identification of disinformation is fraught with several challenges. This fact highlights the importance of studying and identifying disinformation both in the content of web pages and social media posts before it is allowed to spread. In this dissertation, I pursued a three-essay approach to understand, identify, and mitigate the disinformation pandemic. While manual fact-checking is difficult, time-consuming, and expensive, various automated detection solutions could speed up this process. Therefore, in my first essay, I explored whether Machine Learning (ML) techniques can be used to develop predictive models for automatic identification of disinformation. Computational linguistics methods are used to extract content-based, and sentiment-based features of selected webpage’ articles to construct our study dataset. This dataset is used to train various ML algorithms to develop predictive models to identify disinformation. The results showed that there are significant differences among features of true and false information that can be used to identify disinformation. Since the spread of disinformation happens both on media pages and on social media platforms, it is important to analyze disinformation at both levels. Moreover, the literature shows that disinformation spreads six times faster than true information on social media, demonstrating that users get more engaged with disinformation. Therefore, I extended my research to enhance the understanding of disinformation detection based on content-based features and its impact on users’ engagement in social media posts. The findings of the second essay highlighted the critical role of linguistic structure, emotional tone, and the psychological load of social media posts on users’ engagement that can be used to differentiate information from disinformation. The results of the first two essays confirmed that negative emotional tone was one of the most important factors in disinformation posts and was associated with a high engagement score. So, in the third essay, I explore the impact of negative emotional tones in developing users’ perceptions regarding the accuracy of the content. Three separate experiments were developed to explore this. The results of experiments in the third essay highlighted the significant role of negative emotional tones on the believability of the content and their potential influence on behavioral change. My research findings allow for a better understanding and identification of disinformation by highlighting and identifying content-based features that are meant to mislead users to falsely perceive disinformation as information