Misinformation in social media - such as fake news, rumors, or other forms of deceptive content - poses a significant threat to society and, hence, scalable strategies for an early detection of online cascades with misinformation are in dire need. The prominent approach in detecting online cascades with misinformation builds upon neural networks based on sequences of simple structural features of the propagation dynamics (e.g., cascade size, average retweeting time). However, these structural features neglect large parts of the information in the cascade. As a remedy, we propose a novel tree-structured neural network named Cascade-LSTM.

Our Cascade-LSTM draws upon a tree-structured long short-term memory network that is carefully engineered to the structure of online information cascades. Specifically, we suggest a novel bi-directional encoding similar to the information flow, extend inner nodes with further covariates from retweets, and fuse the network with global information from the root. As a result, our Cascade-LSTM overcomes inherent limitations from feature engineering, since it learns propagation features along the complete cascade. The effectiveness of our Cascade-LSTM is demonstrated based on a classification task to predict the veracity of 2,156 Twitter cascades. We improve the detection if misinformation in terms of AUC over the status quo with cascade features by 2.8%.

Altogether, our Cascade-LSTM entails important implications: (1) it presents the first neural classifier that learns the complete cascade. (2) It demonstrates a promising approach to practitioners for detecting misinformation through mining retweet behavior. (3) The model is fairly general, which ensures widespread applicability for inferences from online cascades.