Improving Implicit Crowd Signals Based Fake News Detection on Social Media: A Time-Aware Method for Early Detection
Resumo
Context: The proliferation of Fake News on social media has raised significant concerns around the world. In this context, conceiving information systems capable of early detecting such a type of News remains a challenge. Early detection is relevant once it can enable mitigation actions against the negative effects of Fake News. Problem: A promising approach for Fake News detection relies on Crowd Signals. It considers the reputation of social media users (i.e., crowds) based on their opinions (i.e., signals) when interacting with News in the past. However, despite its high detection rates, this approach lacks exploring early detection. Solution: To overcome this limitation, this work proposes an extension to a State-of-the-Art Crowd Signals-based method, aiming to detect Fake News as soon as possible after its disclosure. IS Theory: Social network theory1. Method: Drawing upon Social Network Theory, the proposed method was validated through experimental evaluation including a comparison with two Crowd Signals-based methods and three early detection methods. Summary of Results: Experiments conducted on four real-world datasets demonstrate the effectiveness of the proposed method and its ability to successfully early detect Fake News by achieving promising classification results. Contributions and Impact in the IS area: The contribution of this paper lies in introducing a Crowd Signals-based method that detects Fake News at earlier stages than the existing State-of-the-Art methods while maintaining comparable detection rates. In terms of impact, this work enables the development of information systems that can filter social media interactions using Fake News early detection. Moreover, it can inspire future research focusing on temporal aspects of Fake News detection.
