Social Context-Aware Trust Prediction: Methods for Identifying Fake News

Fake news, a type of yellow journalism or propaganda, consist of false or incorrect information and have the potential to spread very fast on online social networks. This false information is mainly distributed by social actors who has influence in a specific context (e.g., politics or industry) with the intent to mislead in order to damage an entity (e.g., a politician or a product). Identifying fake news is a challenging task and requires analyzing the reliability, truth, or ability (i.e., trust) of social actors in a certain context and to a certain extent. To address this challenge, in this paper, we present a context-aware trust prediction approach which considers the notion of a context (which conceptually refers to any knowledge to specify the condition of an entity) as well as the social actor’s behavior (supported by theories from social psychology) as first class citizens. We present novel algorithms that employ social context factors inspired by social phycology theories and mathematically model our approach based on Tensor Decomposition. We perform an extensive empirical study and present evaluation results on the effectiveness and the quality of the results using real-world datasets.