Supervised Learning for Misinformation Detection in WhatsApp

Abstract

WhatsApp created anew channel for smartphone users to consume and share news. The easiness to create groups of people that partake similar interests and share content has made WhatsApp prone to abuse by misinformation campaigns. Although fact-checking is very effective for detecting misinformation, it cannot keep up with the sheer volume of information that is now generated online. In this context, we investigate the potential of automatic approaches based on supervised machine learning as a support tool to help fact-checkers identify misinformation shared through images on WhatsApp. Our results show that the predictive performance of the investigated approaches has a useful degree of discriminative power to detect misinformation. Finally, we discussed how WhatsApp misinformation detection approaches can be used in practice, highlighting challenges and opportunities.