规律与因果:大数据对社会科学研究冲击之反思——以社会学为例  

Regularity and Causality: Rethinking the Challenge of Big Data for Social Sciences Research——A Case of Sociology

作　　者：刘林平[1] 蒋和超[1] 李潇晓[1] 

机构地区：[1]南京大学社会学院

出　　处：《社会科学》2016年第9期67-80,共14页Journal of Social Sciences

摘　　要：在社会科学中,大数据研究还刚刚起步,但也取得了一定成果。大数据为社会学和社会科学重新发现社会历史发展规律提供了可能性:它提供了认知宏观社会、检验社会现象的"异质性假设"和"结果稳定假设"的数据基础;它以实时记录的特点较大程度上排除了获取数据时的人为干扰;它将抽样数据中被排斥的极端值重新纳入统计分析。在因果关系上,大数据有助于从根本上克服由于抽样偏颇所引起的样本选择性偏误;匹配数据可以克服或缓解变量遗漏问题;作为面板数据和分层数据,大数据对确定因果效应、检验因果关系比抽样数据更为有利、稳健和可靠。大数据也许可以重构社会学和社会科学的研究目标。The big data studies in social science fields have just started and some preliminary achievements are obtained. Big data provides a probability of discovering the development rules of social history once again by sociology or social sciences. First, it offers the data basis of ＂ heterogeneity hypotheses＂ and ＂stable results hypotheses＂ to understand the macro society and examine the social phenomena. Second, factitious disturbances in data collection process might be mostly overcame as it is characterized by instantaneity. Third, extreme values which are usually ignored in the sampled data are adopted in the statistical analysis. While from the perspective of casual relationship, big data is helpful to radically overcome the samples＇ selective bias which caused by sampling bias. The matched data can solve the problem of missing variables. And, as the panel and stratified data, it is more beneficial, stable and reliable to examine the causality using the big data instead of sampled data. Therefore, this study has different opinions on the statement indicating that big data research only needs to study the correlation relationship rather than the causal relationship. Big data can reconstruct the research objectives of sociology or social sciences. At least, the regularity of human behaviors could be identified by sociology, economics and other social sciences studies according to the big data. Furthermore, the law of development of social history might be discovered. However, this type of reconstruction is neither the revisit for the classical historical determinism, nor the inevitable understanding of the regularity. It advocates a probabilistic pattern. The research logics, methods, techniques and relevant social conditions of sociology and social sciences are also discussed in this paper.

关 键 词：大数据 规律 因果关系 冲击 反思 

分 类 号：C91-03[经济管理]