Exploring Some Roles of Deep Neural Networks in Fake News Detection: A Mixed-Methods Investigation

Abstract

Given the rapid proliferation of fake news on social media, the need for intelligent methods to detect them has become increasingly urgent. This research aims to investigate the functions of deep neural networks in identifying fake news, developing more effective strategies to combat the spread of misinformation. Employing a mixed-methods approach, this study combined qualitative meta-synthesis and thematic analysis with quantitative t-tests to assess model validity. A systematic literature review of 15 articles identified 14 initial concepts, which were then categorized into three main concepts. Subsequently, semi-structured interviews with 12 experts, analyzed using thematic coding, yielded 21 components, grouped into three factors related to the functions of deep neural networks in fake news detection. The sample T-test results confirmed the validity of the proposed model, with findings (p<0.0001) including that the indicators generated were systematically linked, comprehensible, generalizable, and important for designing the model. By integrating the results from both stages, 27 initial concepts were consolidated into four main tasks: text feature extraction, news classification, learning patterns, and system tools. In conclusion, deep neural networks have the potential to play a significant role in combating the spread of fake news. However, achieving a fully accurate and reliable fake news detection system requires a multi-faceted approach that combines various methods and validates them through rigorous statistical analysis