Social Psychology of Interaction and Group Dynamics: The Problem of Interdependence

社会互动研究是社会科学和行为科学研究的核心部分。尽管具有中心性，但统计方法尚未很好地发展用于处理此类数据。现有统计技术的问题在于，它们要求观察结果是独立的。然而，在涉及社会互动的研究中，由于它们的相互作用，观察结果是非独立的。例如：两个兄弟姐妹会影响彼此的行为；工作小组的成员相互影响对方的表现；一群猴子在彼此面前表现出特殊的梳理行为；陪审员在审议过程中会影响彼此的裁决。本研究旨在开发一类统计模型，用于分析来自社会互动研究的数据。它提出了一个新的框架，两两方法，在这个框架中，涉及社会互动的广泛设计可以被测试和评估。其概念是将数据结构的非独立性直接嵌入到数据集中，然后对数据进行相应的建模。这些统计技术将允许研究社会科学和行为科学中的许多问题，这些问题没有得到太多的研究关注，主要是因为没有适当的方法来分析数据。这些技术具有广泛的应用前景，并将为美国国家科学基金会知识与分布式智能（KDI）计划的知识网络重点做出贡献。

The study of social interaction is a central part of research in the social and behavioral sciences. Despite its centrality, statistical methods have not been well developed for handling such data. The problem with existing statistical techniques is that they require observations to be independent. However, in research involving social interaction, the observations are nonindependent by virtue of their interaction. For instance: two siblings influence each other's behavior; the members of a work group influence each other's performance; a group of monkeys exhibit particular grooming behaviors in each other's presence; and jurors in deliberation influence each other's verdicts. This research seeks to develop a class of statistical models that will permit analyses of data from social interaction research. It presents a new framework, the pairwise approach, in which a broad class of designs involving social interaction can be tested and evaluated. The notion is to embed the nonindependence of the data structure directly into the data set, and then model the data accordingly. These statistical techniques will permit the study of many problems in the social and behavioral sciences that have not received much research attention primarily because there have been no appropriate methods for analyzing the data. The techniques can be broadly applied, and will contribute to the Knowledge Networking focus of NSF's Knowledge and Distributed Intelligence (KDI) initiative.