Collaborative Research: Dynamics for Social Networks Processes: Comparing Statistical Models with Intelligent Agents

这个项目的目标是调和两种随时间变化的社交网络建模方法：一类统计模型和智能代理模型。该研究对比了这两种方法的特性，探索了社会网络中哪些定性的动态行为可以被每一种方法有效地捕获和解释。主要的工具是潜在变量表示和动力系统分析。结果是一个框架，该框架描述了两类模型在多种设置下的优势和局限性，并提供反馈，刺激现有理论的改进和新理论的发展。社会网络理论已经产生了大量的范式来描述相互作用的人群（可能是在子群体中）的性质和进化。社会网络理论被应用的环境包括从朋友到公司，从政治集团到宗教团体。反恐是当前特别重要的一个应用，其中网络可能具有潜在的（不可观察的）特征，这些特征对其随时间的演变非常重要，这一概念尤为重要。然而，目前，人们对不同的社交网络动态模型的优势和局限性还没有很好的理解。该项目的研究人员来自社会科学、统计学和应用数学，他们正在构建一个概念和操作框架，允许在多种情况下对不同模型进行原则性比较。除了理论和方法上的进步外，该框架还使社会科学研究人员和其他分析人员能够选择最适合每种应用的动态特性的模型。

The goal of this project is to reconcile two methods for modeling change insocial networks over time: a class of statistical models and intelligent agentmodels. The research contrasts the properties of these two approaches,exploring what qualitative dynamic behaviors in social networks are capturedusefully and interpretably by each. The primary tools are latent variablerepresentations and dynamical systems analysis. The result is a framework thatcharacterizes the strengths and limitations of the two classes of models inmultiple settings, and provides feedback that stimulates refinement ofexisting theories for social networks and development of new theory.Social network theory has produced a wealth of paradigms to describe thenature and evolution of groups of people who interact, possibly in subgroups,with one another. Settings in which social network theory has been appliedrange from friend to corporations to political blocs to religious groups.Counterterrorism is one application of particular current importance, in whichthe notion that the network might have latent (unobservable) characteristicsthat are important to its evolution over time is especially germane.Currently, however, there is not good understanding of strengths andlimitations of different models for the dynamics of social networks.The researchers on this project, drawn from the social sciences,statistics and applied mathematics, are constructing a conceptual andoperational framework that allows the principled comparison of differentmodels in multiple contexts. In addition to advances in theory andmethodology, the framework enables social sciences researchers and otheranalysts to choose a model whose dynamic properties are most appropriate toeach application.