AIE-FND: Enhancing UAV-Based Fake News Detection Technique via AI-Generated Insights From Media Experts

Fake news proliferation poses a growing threat to information credibility. Current researches mainly focuses on analyzing news content and external resources for fake news detection. External resources are known to boost detection accuracy. However, the abundance of external sources implies a wider spread of fake news, which creates a conflict with the high demands for real-time detection. This paper introduces a novel approach, AI-Generated Expert Comments for Enhanced Fake News Detection (AIE-FND), which utilizes large language models (LLMs) to simulate expert-like comments, replacing external resources.. We construct expert profiles through in-depth interviews with 30 journalists and scholars, capturing key characteristics such as demographics, the Big Five personality traits, and writing style. The LLMs generate expert-like comments based on these profiles, using the Role-Background-Objectives-Condition (RBOC) prompt method. Then, by splitting comments based on different perspectives, the model can capture the characteristics of news comments from various aspects, thereby reducing the risk of bias and incorrect judgments caused by a single perspective. Experiments on three datasets show that AIE-FND significantly improves detection, particularly in the early detection phase when the number of actual comments is limited.