Predictive Intelligence Against Fake News Through Intent-Based Language Analysis

The proliferation of sophisticated, digitally disseminated misinformation poses a critical threat to public discourse and democratic processes. Existing fake news detection systems, primarily reliant on content veracity or superficial stylistic features, struggle to adapt to the evolving, multi-faceted nature of deceptive communication. Problem: Current models fail to explicitly account for the author’s underlying, often complex, manipulative intent, leading to limited generalizability and interpretability. Solution: This paper presents the Intent-Aware Fake News Detector (IAFND), a novel predictive system that employs a multi-label classification framework to identify five distinct authorial intents (Deceive, Sensationalize, Propagandize, Manipulate, and Incite) using fine-grained linguistic features. Key Findings: Through rigorous experimental validation on a large, publicly identified dataset (25,000 articles from LIAR, FakeNewsNet, and CoAID), the IAFND demonstrates statistically significant performance improvements over state-of-the-art baselines (p<0.001). Furthermore, the system’s intent-based interpretability module is quantitatively shown to be more robust and actionable than established XAI methods (LIME/SHAP), providing a transparent and scalable solution for combating real-world disinformation.