Quality assurance for data science: Making data science more scientific through engaging scientific method

摘要

Credibility of science is fundamentally due to the strenuous efforts made to verify the general consistency among relevant facts, theories, applications, research methodologies, etc. and scientific method which emphasizes the significance of continuously building and testing hypotheses has withstood the test of time as a successful methodology of acquiring a body of knowledge, we can rely on, at least within a certain context. A paradigm based on composition of data rich services to gather data to replicate real world scenarios through complexity science based simulators, where quality of data as well as theories explaining them is primarily assured via building and testing of hypotheses, can improve our understanding of what we try to comprehend by engaging data science. While simulators would at least partially automate the implementation of scientific method, a credibility ranking mechanism, would not only help determining and disseminating rankings pertaining to the quality of data as well as theories explaining them but also receive & publish feedback regarding rankings. Including methods used in complex system analysis as part of simulators would enhance the scientific rigor of establishing the credibility of knowledge we have. Providing simulation as a service and making graphical hypothesis builders & testers available for external parties (sometimes even members of general public) would democratice the process of ascertaining the believability of data & associated theoretical models thereby further enhancing the Quality of Knowledge.