This paper attempts to understand the mechanisms of fake news beyond the current focus on theoretical modelling of propagation or identification methods based on machine learning. The authors believe that it is of importance to understand the realistic mechanisms between theoretical models and black-box methods. The study tracks large databases of fake and real news from Weibo in China and Twitter in Japan and finds both social networks spread fake news distinctively from real news even at the early stages of propagation.
Highlights:
- The study provides a topological approach to the detection of fake news. It finds that fake news spreads with a very different network topology, even at early stages like five hours after posting, from real news.
- The fraction of re-posting fake news in the first layer is found to be significantly smaller than real news. Real news is shared more widely earlier, while fake news tends to be re-shared widely much later.
- Fake news networks typically have lesser heterogeneity as compared to real news. This is because they are propagated by fewer dominant broadcasters unlike real news.