Explainable Fake News Detection with Source Credibility and Decision Transparency

The propagation of made-up news on digital platforms poses tremendous societal challenges, undermining public trust in the media and democratic processes. Traditional fake news detection techniques are black-box solutions, which do not expose much about decision-making processes in classification. In this paper, we introduce an explainable system for detecting fake news that includes source credibility evaluation and clear decision-making. We have employed various deep learning architectures, i.e., Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and Bidirectional LSTM networks, to analyze text content. We boost these networks with ensemble training and complement them with a source verification module, which employs generative AI in deciding content origin reliability. Our system achieved $\mathbf{9 5. 2 \%}$ accuracy on benchmarked datasets and provides human-interpretable explanations behind decisions. Through thorough various experiments, we have established that improved transparency significantly increases the level of trust and understanding for users concerning the detection outcome. Our multi-layered approach addresses the foundational requirement of explainability in detecting fake news and achieves high accuracy and computational efficiency.