Key points are not available for this paper at this time.
The proliferation of fake news and its detrimental impact has spurred significant research in fake news detection. Existing studies have primarily focused on classifying news content in a broader definition, for example, as real or fake or likely or unlikely, utilizing machine learning algorithms and natural language processing techniques. However, this classification approach must account for the diverse nature of fake news, which can vary in degrees of falseness. Moreover, relying on textual features alone overlooks the multimedia elements often involved in fake news dissemination, such as images and videos. The emphasis on individual instances of fake news also neglects the broader dynamics of information diffusion and contextual factors. To address these limitations, we propose adopting the problem understanding phase of data mining processes for formulating the fake news detection problem. This phase involves a comprehensive understanding of the characteristics of fake news, the credibility indicators across various modalities, and the social dynamics that shape its spread. This paper provides a literature background, explores the problem of understanding fake news detection, discusses the challenges in achieving optimal detection solutions, and highlights potential research opportunities for future work in this critical area.
Chua et al. (2023) studied this question.