Calibration of predictive agent-based models of opinion dynamics

基于预测代理的舆论动态模型的校准

• 批准号: 2588247
• 金额: --
• 依托单位: University College London
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2021
• 资助国家: 英国
• 起止时间: 2021 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2588247
• 关键词: Calibration; predictive; agent; based; models

Opinion dynamics is a fast-growing interdisciplinary research area where problems of fundamental importance to our societies are studied with a mix of data-driven approaches and mathematical modelling. This PhD aims to develop quantitative approaches aimed at modelling and contrasting the diffusion of "fake news" and misinformation in online social networks (OSNs), which have been the subject of major research efforts in recent years. As the literature on the subject has grown, an apparent gap has emerged between its two main branches. On the one hand, increasingly sophisticated mathematical models have been put forward to understand how different rules of communication between individuals may lead to very different states (e.g., consensus, polarisation) in networked populations. On the other hand, data-driven studies have revealed several stylised facts that characterise the diffusion of information in the most popular OSNs, such as Facebook and Twitter. Yet, there is very little overlap between these two strands of research, as the currently available mathematical models struggle to come up with testable predictions, which in turn prevents from any meaningful model validation/rejection against empirical data. The focus of this PhD will be that of filling this gap by leveraging agent-based modelling and the wide array of existing techniques to calibrate agent-based models (ABMs) on empirical data.

舆论动力学是一个快速发展的跨学科研究领域，对我们的社会至关重要的问题是通过数据驱动的方法和数学建模的混合研究。该博士旨在开发定量方法，旨在模拟和对比在线社交网络（OSNs）中“假新闻”和错误信息的传播，这是近年来主要研究工作的主题。随着关于这一主题的文献越来越多，它的两个主要分支之间出现了明显的差距。一方面，人们提出了越来越复杂的数学模型来理解个体之间不同的交流规则如何导致网络群体中非常不同的状态（例如，共识，极化）。另一方面，数据驱动的研究揭示了几个程式化的事实，这些事实反映了Facebook和Twitter等最受欢迎的社交网络中信息传播的特征。然而，这两种研究之间几乎没有重叠，因为目前可用的数学模型很难提出可测试的预测，这反过来又阻止了对经验数据进行任何有意义的模型验证/拒绝。本博士的重点将是通过利用基于代理的建模和广泛的现有技术来根据经验数据校准基于代理的模型（ABMs）来填补这一空白。