Analyzing the Potential of Feature Groups for Misinformation Detection in WhatsApp

Abstract

The new mass media (e.g., social networks, and instant messaging applications) have drastically changed how users generate, propagate, and consume information. Given this scenario, an emerging problem is the abuse of these platforms for the propagation of disinformation that, consequently, affects the credibility of the news ecosystem in these environments. The dissemination of misinformation on these platforms has become a worldwide phenomenon, and scalable strategies to contain or mitigate the problem are scarce. Thus, using automated approaches to detect disinformation on digital media can help journalists and fact-check teams identify content that needs to be verified. In this context, in this work, we investigate the potential of feature groups for automatic misinformation detection shared on digital platforms. Our results reveal that using a set of features from some groups, can be useful to build models with satisfactory performance, which can make the application of the model viable in a practical scenario.