科研再现性危机:行为表现、内生逻辑及治理路径

Scientific research reproducibility crisis: Behavioral rerformance, endogenous logic, governance path

再现性危机已成为制约科研高质量发展的“瓶颈”。科研再现性危机主要有三大行为表现:科研统计的技术性“失真”、科研数据的过程性错误、科研决策的程序性“陷阱”。科研再现性危机主要是由科学域内在逻辑因素生发而成:学术无限自由引发科研行为失范,科研过程开放性缺失导致科研再现受限,不良专业文化滋生假阳性结果泛滥。对此,科研利益相关者需要在实验统计技术治理上修正P值域和设立置信区间,在科研过程治理上维护自由限度和推进开放科学,在科研专业文化治理上推进评价制度创新和改革公众科学媒体,在学科专业治理上改革科学课程体系和培养科学核心素养,在政府外部科学治理上推进政府科学政策改革和维护再现性研究合法性。

Reproducibility is the core nature and lifeblood of scientific research. However, significant changes have taken place in scientific research ecology, such as the lack of scientific research ethics, the politicization of science, the scientific culture that cannot tolerate failure, the unlimited expansion of academic freedom, and the repeated misconduct, especially in the increasingly complicated social environment. These changes, consequently, bring serious challenges to Merton’s “scientific norms” in the sense that a “trap” emerges in scientific domain in which the results of scientific research cannot be reproduced. The reproducibility crisis has become a “bottleneck” restricting the high-quality development of scientific research from three perspectives. Firstly, the technical "distortion" of scientific research statistics, among which the most prominent one is P-hacking which leads to the false-positive results and the irreproducibility of scientific experiments. Secondly, the procedural error of scientific research data, that is generating statistical results from “recalcitrant data” or creating data set on the basis of excluding the obvious “bizarre cases” and only employing data consistent with the preset ones. Thirdly, the procedural “trap” of scientific research decision-making, specifically the inappropriate practice in data collection and analysis thus weakening the statistical significance of these data. The reproducibility crisis of scientific research mainly arises from the logical factors inherent in the scientific domain. Firstly, the lack of necessary academic norms and supervision due to excessive worship or misunderstanding of academic freedom leads to anomie of scientific research behavior and irreproducibility of research results. Secondly, the lack of openness in the scientific research process or the failure of researchers to make their research data and procedures available to the public results in limited scientific reproduction. Thirdly, the proliferation of false-positive results bred from negative professional cultures such as the premium on positive results and the scientific politicization of "group thinking" makes scientific research results irreproducible. In this regard, the scientific community needs to strengthen governance of the reproducibility crisis in scientific research through the following. Firstly, in terms of experimental statistical technology, the stakeholders of scientific research should modify and narrow the range of P value and establish confidence intervals to enhance the reproducibility of scientific research results. Secondly, in terms of scientific research process, the scientific community should manage to maintain the limits of academic freedom and share its scientific research methods, data, computational models and other scientific research results with the public so as to provide conditions to reproduce discoveries and test original research. Thirdly, in terms of scientific research professional culture, the government should promote the innovation of evaluation system, encourage researchers to actively carry out more reproducible research, fully respect those who produce negative research results, and guide the public scientific media to pay rational attention to the reproducibility crisis in scientific research. Fourthly, in terms of discipline construction, reform of scientific curriculum system should be made to cultivate students’ ethics in scientific research. Fifthly, in terms of external governance, an overhaul of the science policy should be made to vindicate the validity of the reproducible research.