因果推断五十年:成就、挑战与应对

Fifty Years of Causal Inference:Achievements,Challenges,and Responses

鲁宾因果模型问世五十年以来,学术界对因果关系的理论认识和分析方法有了长足发展,应用因果推断方法开展的实证研究也不断增多。但鲁宾因果模型在取得巨大成就的同时,也面临很多挑战和局限。这既包括将该模型应用于观察研究时无法保证随机分配假定的局限,也包括为了满足近似随机分配的合理性,而无视研究问题的实质重要性的局限。除此之外,鲁宾因果模型在分析结果的原因和识别因果机制方面也存在不足。为了应对这些挑战,研究者既要改进定量研究设计,也要有意识地借鉴和吸纳定性分析工具,只有将定性和定量两种方法有机结合起来,才能对因果关系进行全面和透彻的分析,进而有助于对因果问题的全面认识,促进定量与定性两个研究范式的融通和促进。

In the 50 years since Rubin causal model was introduced,significant progress has been made in the theoretical understanding and analytical methods of causality,leading to a marked increase in empirical research using causal inference methods.Despite these achievements,the model faces various challenges and limitations.One such limitation is itsinability toeensure the random assignment assumption when applied to observational studies.Furthermore,the model overlooks the substantive importance of the research problem in its pursuit of satisfying the criterion of approximate random assignment.In addition,the model is deficient in analyzing the causes of effects and identifying causal mechanisms.To address these challenges,researchers must not only enhance the quantitative research design but also conscientiously integrate and learn from qualitative analysis tools.This organic combination of qualitative and quantitative methods is imperative for conducting thorough and comprehensive analyses of causality.