Simulation Models for Diffusion over Multirelational Dynamic Networks

描述（由申请人提供）：传播是社会网络影响健康的一种主要方式，既可以是直接的，如性传播疾病在伴侣之间传播，也可以是间接的，如健康相关行为规范在社会网络中传播。虽然至关重要，但除了少数明确和相对简单的情况外，大多数网络扩散问题的复杂性使分析模型变得繁琐或难以处理。因此，社会模拟成为了解网络特征如何影响人口健康的有效工具，但我们缺乏一个通用的模拟框架，这使得很难比较调查团队的工作。因此，这个项目的主要目的是建立一个灵活的网络扩散模拟系统，准确地描绘动态的，多关系的社会网络。该项目利用新开发的多片网络结构作为一种灵活的工具，将网络从静态的单关系图扩展到由多种关系组成的复杂的、不断发展的系统，这些系统可以很容易地适应复杂的传输规则。这一提议的基本见解是双重的。从理论上讲，它基于这样一个概念，即复杂性自然地从更简单的子系统之间的连接中产生。在方法上，我们使用多层网络模型将简单的子系统连接到一个复杂的整体。因此，我们首先确定了一个广义的扩散框架，该框架基于网络结构在三个维度上的差异和二元传播过程的差异，也沿着三个维度变化。结合起来，这6个复杂性维度允许我们指定一个网络扩散模型，其范围从极其简单（固定病原体在静态网络上的特定传播）到复杂（在多个不断发展的网络上协商概念扩散）。由此产生的模拟系统将使跨多个领域的卫生研究人员能够模拟潜在的网络扩散过程，并回答有关网络结构特征如何影响健康和健康差异的明确问题。

DESCRIPTION (provided by applicant): Diffusion is a primary way in which social networks affect health, both directly, as when sexually transmitted diseases are passed between partners, and indirectly, as when norms about health relevant behavior travel through a social network. While crucial, with the exception of a few well specified and comparatively simple cases, the complexity of most network diffusion problems make analytic models cumbersome or intractable. As such, social simulations become an effective tool for understanding how network features affect population health, but we lack a generalized simulation framework, making it difficult to compare work across investigative teams. Thus, the primary aim of this project is to build a flexible network diffusion simulation system that accurately portrays dynamic, multi-relational social networks. The project makes use of newly- developed multi-slice network structures as a flexible tool for extending networks from static single-relation graphs to complex, evolving systems composed of multiple relations that can easily accommodate complex transmission rules. The fundamental insight of this proposal is twofold. Theoretically, it rests on the notion that complexity emerges naturally from the connections among simpler sub-systems. Methodologically, we operationalize linking of simpler sub-systems to a complex whole by using multi-slice network models. Thus, we first identify a generalized diffusion framework resting on differences in network structure along three dimensions and differences in dyadic transmission processes, also varying along three dimensions. Combined, these 6 complexity dimensions allow us to specify a network diffusion model that ranges from extremely simple (certain spread of a fixed pathogen over a static network) to complex (negotiated ideational diffusion across multiple evolving networks). The resulting simulation system will give health researchers across multiple domains the ability to model underlying network diffusion processes and answer clear questions about how features of the network structure affect health and health disparities. PUBLIC HEALTH RELEVANCE: At both the individual and population levels, health is affected by networks through a process of physical (as with infection spread) or social (as with peer influence) diffusion. Since analytic models for the vast majority of complex network diffusion situations are lacking, this project develops a generalized network simulation framework for health-relevant diffusion that can accommodate a wide variety of relevant situations. The result of this project is increased capacity to model the links between network structure, individual behavior, and the health status of individuals and communities.