Abstract:In the age of the infodemic, it is crucial to have tools for effectively monitoring the spread of rampant rumors that can quickly go viral, as well as identifying vulnerable users who may be more susceptible to spreading such misinformation. This proactive approach allows for timely preventive measures to be taken, mitigating the negative impact of false information on society. We propose a novel approach to predict viral rumors and vulnerable users using a unified graph neural network model. We pre-train network-based user embeddings and leverage a cross-attention mechanism between users and posts, together with a community-enhanced vulnerability propagation (CVP) method to improve user and propagation graph representations. Furthermore, we employ two multi-task training strategies to mitigate negative transfer effects among tasks in different settings, enhancing the overall performance of our approach. We also construct two datasets with ground-truth annotations on information virality and user vulnerability in rumor and non-rumor events, which are automatically derived from existing rumor detection datasets. Extensive evaluation results of our joint learning model confirm its superiority over strong baselines in all three tasks: rumor detection, virality prediction, and user vulnerability scoring. For instance, compared to the best baselines based on the Weibo dataset, our model makes 3.8\% and 3.0\% improvements on Accuracy and MacF1 for rumor detection, and reduces mean squared error (MSE) by 23.9\% and 16.5\% for virality prediction and user vulnerability scoring, respectively. Our findings suggest that our approach effectively captures the correlation between rumor virality and user vulnerability, leveraging this information to improve prediction performance and provide a valuable tool for infodemic surveillance.

What problem does this paper attempt to address?

This paper attempts to address the problem of effectively monitoring rumor propagation and identifying susceptible users in the era of infodemic. Specifically, the paper proposes a multi-task learning framework based on Graph Neural Networks (GNN) aimed at simultaneously predicting viral rumors, the scale of information dissemination, and user susceptibility. Through this approach, researchers hope to take preventive measures in advance to reduce the negative impact of misinformation on society. ### Core Issues of the Paper: 1. **Rumor Detection**: How to distinguish between rumors and non-rumors when observing information dissemination on social media platforms at an early stage. 2. **Virality Prediction**: How to predict the final dissemination scale of specific information at the early stage of its spread. 3. **User Susceptibility Scoring**: How to identify users who are more likely to spread rumors during the observed dissemination process and assign them a susceptibility score. 4. **Joint Prediction Performance Enhancement**: How to enhance the joint prediction performance of rumors, virality, and user susceptibility by improving task-related feature representations. ### Solution: The paper proposes a unified multi-task learning framework that includes the following four main components: 1. **User Interaction Graph Construction**: Constructing a user interaction graph based on the dissemination network to reflect the actual interaction relationships between users. 2. **Input Embedding**: Generating time-aware post embeddings and user embeddings based on contrastive learning to capture users' original features. 3. **Refined Embedding**: Further optimizing user embeddings through a user-post cross-attention mechanism and a community-enhanced vulnerability propagation method. 4. **Output Layer**: Feeding the final graph representations and updated user representations into the corresponding classifiers to complete the predictions for the three tasks. ### Method Overview: - **User Interaction Graph Construction**: Extracting interaction relationships between users from the dissemination network to construct a user interaction graph. - **Input Embedding**: Using a pre-trained BERT model to generate embeddings for post content and combining time information to generate time-aware post embeddings. Simultaneously, generating initial user embeddings through contrastive learning methods. - **Refined Embedding**: Utilizing a user-post cross-attention mechanism to further optimize user embeddings by combining post content and user features. Additionally, discovering user community information through a community-enhanced vulnerability propagation method to further improve user embeddings. - **Output Layer**: Feeding the optimized graph representations and user representations into the rumor detector, virality predictor, and susceptibility scorer respectively to complete the final prediction tasks. Through these methods, the paper aims to provide an effective tool to help infodemic monitoring systems timely identify and handle high-risk rumors and susceptible users.

Predicting Viral Rumors and Vulnerable Users for Infodemic Surveillance

Interpretable Graph Neural Network for Social Media Rumor Detection

Detecting the Rumor Patterns Integrating Features of User, Content, and the Spreading Structure.

Rumor Detection with a novel graph neural network approach

Epidemiology-informed Network for Robust Rumor Detection

Rumor detection on social media using hierarchically aggregated feature via graph neural networks

Rumor Localization, Detection and Prediction in Social Network

Modeling the Uncertainty of Information Propagation for Rumor Detection: A Neuro-Fuzzy Approach

Early detection of rumors based on source tweet-word graph attention networks.

A Deep Neural Networks-Based Fusion Model for COVID-19 Rumor Detection from Online Social Media

Graph Convolution-Based Joint Learning of Rumor with Content, User Credibility, Propagation Context, and Cognitive as Well as Emotion Signals

Adaptive Spatial–Temporal and Knowledge Fusing for Social Media Rumor Detection

Identifying Possible Rumor Spreaders on Twitter: A Weak Supervised Learning Approach

Study on the Rumor Detection of Social Media in Disaster Based on Multi-Feature Fusion Method

A Bi-GRU-DSA-based social network rumor detection approach

Continuous-Time Dynamic Graph Networks Integrated with Knowledge Propagation for Social Media Rumor Detection

Predicting social media rumours in the context of public health emergencies

A Novel and High-Accuracy Rumor Detection Approach Using Kernel Subtree and Deep Learning Networks

Rumor Detection on Social Media with Bi-Directional Graph Convolutional Networks

On Early-stage Debunking Rumors on Twitter: Leveraging the Wisdom of Weak Learners

Early Detection of Rumors in Heterogeneous Mobile Social Network.