In-Silico Sandbox to Model Multi-Platform Computational Disinformation

这项拟议的研究计划旨在推进对社交网络和舆论动态建模的量化方法，以弥补人们对通过社交媒体平台开展的虚假信息活动日益关注的研究空白。对舆论动态的研究，特别是在多个社交媒体平台上传播“假新闻”的研究正在增长，但往往被社交媒体平台的发展所超越。现有文献的最新进展集中在检测和表征在社交媒体上传播的虚假新闻，留下了挑战和悬而未决的问题，包括算法干预在防止虚假信息传播方面的有效性。这项研究计划试图使用基于代理的建模技术来模拟多平台网络机制、社交互动和舆论动态，以创建加拿大社交媒体环境的统计代表性合成环境。重点是开发必要的新方法来创建和告知这个现实的模拟器，它捕捉潜在的复杂性和异构性，但仍然适合计算实验。该计划的短期目标是：1)建模并构建社交媒体环境的合成群体，该群体以统计方式表示用户特征、观点和交互行为以及网络模式；2)开发计算算法以从经验数据中捕获和归结代理、网络和观点特征的分布；3)构建基于计算代理的模型模拟器，从统计上重建社交媒体网络和观点的内生动态；以及4)创建计算机沙盒以测试算法干预机制的效果。拟议计划的结果将是加拿大社交媒体环境的真实计算模拟器，该环境捕捉到人类和社交行为、舆论动态以及跨多个社交媒体平台的复杂网络动态的复杂性。这一提议将导致在一个直接重要的领域--通过社交媒体话语的舆论动态--为应用程序提供信息的新方法。我们开发的方法将有可能导致社会动力学的新发现。如果成功，拟议的计算模拟器还将有助于向社交媒体组织和公共机构提供有关开发干预措施以防范虚假信息幽灵的信息。

This proposed research program aims to advance quantitative methods in modeling social networks and opinion dynamics, addressing a research gap about the growing concern around disinformation campaigns via social media platforms. Research on opinion dynamics, particularly the propagation of "false news" across multiple social media platforms, is growing but often outpaced by developments on social media platforms. Recent advances in extant literature focus on detecting and characterizing false news that propagates on social media, leaving challenges and open questions, including the efficacy of algorithmic interventions in preventing the spread of disinformation. This research program seeks to create a statistically representative synthetic environment of the Canadian social media environment using agent-based modeling techniques to simulate multi-platform network mechanisms, social interactions, and opinion dynamics. The focus is on developing new methods necessary to create and inform this realistic simulator that captures underlying complexity and heterogeneity but remains tractable for computational experimentation. The short-term objectives of the program are: 1) Model and construct a synthetic population of the social media environment that statistically represents user characteristics, opinions and interaction behavior, and network patterns; 2) Develop computational algorithms to capture and impute distributions of the agent, network, and opinion characteristics from empirical data; 3) Build a computational agent-based model simulator that statistically recreates the dynamics of social media networks and opinions endogenously; and 4) Create an in silico sandbox to test the effects of algorithmic intervention mechanisms. The outcome of the proposed program will be a realistic computational simulator of the Canadian social media environment that captures the complexity of human and social behavior, opinion dynamics, and complex network dynamics across multiple social media platforms. This proposal will lead to new methods that inform applications in an area of immediate importance - opinion dynamics via social media discourse. The methods we develop will have the potential to lead to novel discoveries in social dynamics. If successful, the proposed computational simulator will also help inform social media organizations and public institutions on developing interventions to guard against the specter of disinformation.