Collisions of Methodology and Epistemology in Data Science

Abstract

Data Science, which emerged relatively recently, has taken its rightful place in the structure of sciences. The application of data science has shown outstanding possibilities for solving many complex problems in various fields of activity. The basis of its success was a new methodology of cognition, including the concepts and methods of Big Data, Artificial Intelligence, an interdisciplinary approach (computer science, statistics, mathematics, social and humanitarian sciences). The new scientific paradigm of Data Science radically transforms scientific methodology and therefore needs to be substantiated. To solve the problem, the scientometric method, case-study methods, comparative analysis, methodological and epistemological analysis are used. The article considers cases of methodological and epistemological collisions that hinder the effectiveness of data science, their causes and consequences. Specifically, examples of improving search engines on the Internet, optimizing the management of scientific research, and the operation of car navigators in megacities are analyzed. As a result of the conducted research, two groups of contradictions between the methodology and epistemology of data science are distinguished. The first group is associated with subjective causes of dilemmas, the second — with objective ones. In the first group, methodological reasons for the emerging conflicts prevail, while in the second group — epistemological reasons for the emerging contradictions. In the author’s opinion, objective paradoxes are more complex. They touch upon deep questions of the philosophy of science. In any case, the identified contradictions lead to a decrease in the potential of data science, lead to erroneous decisions and erroneous forecasts, and they must be eliminated.