“See the Image in Different Contexts”: Using Reverse Image Search to Support the Identification of Fake News in Instagram-Like Social Media

Abstract

Social media are an integral part of the daily lives of today’s young generation. In addition to the positive impact on learning through these channels, there are also risks related to toxic content like “fake news” on various social media. Fake news aims to change opinions based on disinformation or misinformation supporting conspiracy theories, e.g., related to the pandemic. Fake news creators use various multimedia artifacts, including images taken from serious and valid news sources, to attract the audience’s attention. Tracking images in different contexts can give social media users important clues to distinguish fake news from credible information. We report on the development of a web-based learning environment that includes a “virtual learning companion” to help learners improve their understanding, awareness, and critical thinking concerning such social media threats. The learning environment mimics Instagram and includes toxic and non-toxic content in a controlled way. The companion is implemented as a browser plugin that communicates with students via chat. The companion poses knowledge activation questions and answers according to an underlying script. The companion offers other sources with the same image identified through Reverse Image Search (RIS). The goal is to help learners find the same image in different contexts with different textual descriptions and keywords. For this purpose, we added basic NLP mechanisms to extract keywords from these contexts, including keywords that signal persuasiveness. Currently, we evaluate the impact of this tool and the provided support in distinguishing fake or credible news.