Mistakes vs. Malice: Automatic Classification of Unintentional Fake News on Celebrity Deaths

Abstract

This study addresses the growing problem of information clutter, particularly in the context of celebrity death announcements. As the rapid spread of misinformation becomes a critical social issue, especially through mass and social media, it is essential to understand the mechanisms behind this phenomenon. The research employs an interdisciplinary approach, integrating computational, linguistic, sociological, and journalistic perspectives to analyze the characteristics of unintentional fake news. Using machine learning classification techniques, it seeks to differentiate between unintentional fake news, real news that debunks these false claims, and real news. The findings reveal significant linguistic features that contribute to the classification process. However, while there are models capable of classifying certain specific news types, none of them are able to correctly classify all news types considered, underscoring the complexities involved in distinguishing between correct, misinformation, and disinformation. This work not only sheds light on the nature of unintentional fake news but also emphasizes the need to improve fact-checking processes in journalism to combat the viral spread of misinformation. Ultimately, the study calls for further research into the implications of these findings for media practices and the role of technology in addressing the challenges of information disorder in contemporary society.