An explainable machine learning framework for fake financial news detection

In recent years, we have witnessed a continuing onslaught of fake news, or other forms of biased information on social media platforms. These types of information can influence people's beliefs, attitudes, and behaviors by its ubiquity with significant social and economic implications. In this study, we examine fake news on crowd-sourced platforms for financial markets. Assembling a unique dataset of unambiguous fake news articles that were prosecuted by the Securities and Exchange Commission, along with propagation data of such news on other digital platforms and the financial performance data of the focal firm, we develop a well-justified and explainable machine-learning framework to predict fake financial news on social media platforms. Our framework design is rooted in the Truth Default Theory, which emphasizes contextualized information for deception detection. Extensive analyses are conducted to evaluate the performance and efficacy of the proposed framework.