论文标题
使用快照人群行为没有网络数据的大规模社交网络的推论和影响
Inference and Influence of Large-Scale Social Networks Using Snapshot Population Behaviour without Network Data
论文作者
论文摘要
人口行为（例如投票和疫苗接种）取决于社交网络。社交网络可能会因行为类型而有所不同，并且通常被隐藏。但是，我们经常有大规模的行为数据，尽管仅在一次时间点拍摄的快照。我们提出了一种仅使用快照种群行为数据的人类行为的大规模网络结构和网络模型的方法。这利用了一些参数，几何社会人口统计网络模型和基于旋转的行为模型的简单性。我们为欧盟公投和两次伦敦市长选举说明了该模型如何提供对同质倾向的预测和解释。除了从大规模行为数据集中提取特定于行为的网络结构外，我们的方法还产生了一个粗略的微积分，将不平等和社会偏好与行为成果联系起来。我们提供了潜在网络敏感政策的例子：收入不平等，社会温度和同质偏好的变化如何在最近的选举中降低了极化。
Population behaviours, such as voting and vaccination, depend on social networks. Social networks can differ depending on behaviour type and are typically hidden. However, we do often have large-scale behavioural data, albeit only snapshots taken at one timepoint. We present a method that jointly infers large-scale network structure and a networked model of human behaviour using only snapshot population behavioural data. This exploits the simplicity of a few parameter, geometric socio-demographic network model and a spin based model of behaviour. We illustrate, for the EU Referendum and two London Mayoral elections, how the model offers both prediction and the interpretation of our homophilic inclinations. Beyond offering the extraction of behaviour specific network structure from large-scale behavioural datasets, our approach yields a crude calculus linking inequalities and social preferences to behavioural outcomes. We give examples of potential network sensitive policies: how changes to income inequality, a social temperature and homophilic preferences might have reduced polarisation in a recent election.
