Identification of Influential Users to Curb the Spread of Fake News on Social Networks

为了减少假新闻在社交网络中给人们带来的伤害，为社交平台运营商和政府提供有效的思路，本项目将对这一问题进行深入分析。本课题的研究问题是：(1)如何利用多源结构信息而不是有限信息来识别网络中有影响力的用户？(2)我们如何建立一个模型来模拟个人信仰和决策对他人的影响？(3)个人态度变化对SN中的信息传播过程有怎样的影响，如何通过现实世界的数据集对信息传播模型中的参数进行量化，使模型表现良好？为了回答这些问题，本项目将主要研究个人在信息传播中的作用和虚假新闻在SNs中的传播模式，从而为遏制虚假新闻在网络社交平台上的传播提供建议。本项目的目的是：(1)为了避免由于考虑有限信息而导致的结果偏差，本项目将采用典型的多准则决策（MCDM）方法（Yang & Xu 2002），在具有不确定性和无知的多层结构中聚集多条信息，并考虑SNs中的多源拓扑信息来识别影响节点；包括用户特征和SN的统计属性。具体步骤包括但不限于信息源的选择（Lu et al. 2016）、权重的确定（Zhou et al. 2020）以及信息融合模型的构建。这个项目将分析这些重要用户在SN中的分布，观察他们是否遵循一定的规律。定量大规模实验和定性分析将同时进行，以验证有效性。(2)在获得SN中的这些重要用户后，本项目将基于证据理论（Ni et al. 2021）探索他们如何影响网络中的其他用户，例如改变他人的决策和行为。证据理论可以通过质量函数同时表达和组合各种情况发生的概率，而不是简单的是或否，适合描述用户对假新闻的反应。通过这一部分，我们将研究一些有影响力的用户如何操纵或推动社交网络中的假新闻传播过程，从而制造信息泛滥，误导大量不知情的用户。这将在现实世界和合成网络中进行实验评估，以观察这些重要用户对假新闻传播造成的社会影响。(3)通过现有的动态模型，结合一些合理的因素（Zhang et al. 2016, Zhu & Zhao 2017），如用户的心理因素和行为延迟，以及社会推荐，来研究假新闻的传播模式。通过与真实传播数据集的比较和分析，该模型的性能将得到改进。本项目旨在提出一个机制驱动和数据驱动的假新闻传播模型。

In order to reduce the harm that fake news brings to people in SNs and provide an effective idea for social platform operators and the government, this project will conduct an in-depth analysis of this issue. The research questions of this project are:(1) How can we identify influential users in the SN by considering multi-source structural information rather than limited information?(2) How can we build a model that can simulate the impact of personal beliefs and decisions on others?(3) How is the information propagation process in the SN affected by personal attitude changes, and how can we quantify the parameters in the information propagation model through real-world datasets to enable the model to perform well?To answer these questions, this project will mainly study the role of individuals in information propagation and fake news propagation pattern in SNs, thereby providing recommendations for curbing the spread of fake news on online social platforms. The aims of this project are:(1) To avoid the biased results caused by considering limited information, the evidential reasoning algorithm (Yang & Xu 2002) that is a typical multiple-criteria decision-making (MCDM) method to aggregate several pieces of information in a multilevel structure with uncertainty and ignorance will be applied in thisproject to consider multi-source topology information in SNs to identify influential nodes, including the characteristics of users and statistical properties of the SN. The specific steps include but are not limited to the selection of information sources (Lu et al. 2016), the determination of weights (Zhou et al. 2020), andthe construction of the information fusion model. This project will analyze the distribution of these important users in the SN and observe whether they obey certain laws. Quantitative large-scale experiments and qualitative analyses will be undertaken simultaneously to verify the validity.(2) After obtaining these important users in the SN, this project will explore how they can influence other users in the network based on the evidence theory (Ni et al. 2021), such as changing the decisions and behaviours of others. Evidence theory can express and combine the probability of various cases simultaneouslyinstead of simple yes or no through the mass function, which is suitable for describing users' responses to fake news. Through this part, we will examine how some influential users manipulate or promote the fake news propagation process in SNs, thereby creating infodemic and misleading large numbers of uninformed users. This will be experimentally evaluated in real-world and synthetic networks to observe the impact of social influence on fake news propagation caused by these important users.(3) The fake news propagation pattern will be studied through the existing dynamical models by combining some reasonable factors (Zhang et al. 2016, Zhu & Zhao 2017), such as psychological factors and behavioural delay of users, as well as social recommendations. The performance of this model will beimproved by comparison and analysis with real-world propagation datasets. This project aims to propose a mechanism-driven and data-driven fake news propagation model.