全数据模式的幻象与网络大数据的代表性

The Whole Data Model Illusion and the Representativeness of Online Big Data

大数据时代为计算社会科学的发展提供了契机。有一种观点认为,由于大数据是"样本=总体",因此它不存在采样偏差和数据代表性问题。虽然大数据驱动下的社会科学研究取得诸多成果,但也有不少失败的案例,对这些案例进行分析可见,"总体数据"是相对于具体的研究对象和研究问题而言的,大数据时代并不能保证社会科学开展全数据模式研究。数字鸿沟、用户偏好等客观存在的问题,使网络大数据往往是用户自我选择样本。在很多情况下,"全数据模式"只是缺乏深思明辨而勾勒出的一幅幻象,社会科学研究者应对此具备清醒的认识,方能作出高质量的研究。

The era of big data provides opportunity to the development of computational social science. There is a view that given "everything can be digitized",social science can acquire research-required "whole data",as "big data is whole data",sampling bias and data representativeness issue no longer exist. Although big-data-driven social scientific research has made a series of achievements,there are also certain unsuccessful cases,through which it can be found that "whole data" is relative to the specific research object and issue,the era of big data cannot guarantee whole data research model. Digital divide,users’ preferences and other objective problems make online big data mostly user self-selected sample. In many cases,"whole data model" is an illusion created by lacking of care discernment,social science researchers should have a clear understanding of this,so that they can conduct high quality research.