It is an online platform and not the real world, I don't care much: Investigating Twitter Profile Credibility With an Online Machine Learning-Based Tool

Abstract

Social media is now an important source of everyday information. Given the plethora of scandals concerning the rapid spread of misinformation and disinformation on social media, the credibility of the content on these platforms is now a pivotal research area. Much of the existing work on social media credibility focuses on content credibility. In this study, however, we focus on the credibility of the profile as the virtual representation of the content author. We developed a real-time machine-learning-based online tool that assesses the credibility of profiles on Twitter, one of the most common and versatile social media platforms. To investigate user perceptions on credibility-related issues, we used our tool as a stimulus for people to reflect on their profile's credibility and collected 100 responses. The combination of our quantitative and qualitative analysis reveals that the latest tweets and retweet behavior are two of the most critical factors for profile credibility. It is also observed that people demonstrate a limited interest in their profile credibility but agree that the author's credibility is of paramount importance. With an open-source tool to assess user credibility on Twitter and a user study to establish its utility, we contribute a timely piece of research on the topic of online credibility.