奥迪斯·邓肯的学术成就：社会科学中用于定量推理的人口学方法

Otis Dudley Duncan’s Legacy:the Demographic Approach to Quantitative Reasoning in Social Science

奥迪斯·戴德里·邓肯,于2004年11月去世。他对当今社会学和人口学的量化推理实践起到了极为重要的作用。本文将在科学史的背景下追溯邓肯作为一位量化社会学家所产生的影响。将O.D.邓肯的社会科学哲学置于"总体逻辑思维"的传统中进行分析,这一传统起源于查尔斯·达尔文,其后由弗朗西斯·加尔顿引入社会科学。作为分析的一部分,我将对统计分析的两种方法加以区分(起源于两种主要的有关科学的哲学视角):高斯或类型逻辑思维和加尔顿或总体逻辑思维,并进一步对邓肯在社会科学中的量化推理视角进行细致讨论,特别是他对于社会测量、路径分析、结构模型、计量经济学和Rasch模型的看法。本文的一个重要主题就在于O.D.邓肯很快意识到社会科学中量化方法所遇到的困难及其局限性,他尤其被内生的总体异质性所困扰。正是这种异质性使得在社会科学中通过统计分析得到具有定律性质的推论变得徒劳。因此,O.D.邓肯十分鄙视那些试图效仿物理科学来探寻假想的社会普适法则的做法,因为他坚信这种法则并不存在且毫无意义。对于O.D.邓肯来说,统计工具的主要作用在于对总体差异的系统性模式进行归纳总结。本文将大量引用O.D.邓肯过去未经发表的私人通信。

Otis Dudley Duncan, who died in November 2004, had enormous impact on the practice of quantitative reasoning in sociology and demography today. This paper traces the influence of Duncan as a quantitative sociologist within the context of the history of science. I locate Duncan’s philosophy of social science within the tradition of 'population thinking' that was begun by Charles Darwin and introduced to social science by Francis Galton. As part of this exploration, I distinguish two approaches to statistical analysis (emanating from the two main philosophical views of science): Gaussian or typological thinking, and Galtonian or population thinking. I examine in detail Duncan’s views of quantitative reasoning in the social sciences, particularly his opinions on social measurement, path analysis, structural modeling, econometrics, and the Rasch model. An important theme of the paper is that Duncan quickly realized the difficulties and limitations of quantitative methodology in social science. In particular, he was bothered by inherent population heterogeneity that makes it futile to draw 'law-like' inferences from statistical analyses in social science. Thus, Duncan was disdainful of the search for supposedly universal laws of society that would mimic those of the physical sciences, because he believed that such laws did not exist and would be meaningless. Instead, to Duncan, the main use of statistical tools is to summarize systematic patterns of population variability. The paper draws heavily on Duncan’s previously unpublished personal communications.