The Sciences of Data – Moving Towards a Comprehensive Systems Perspective

Data science’s rapid development in a dynamically growing data environment endows it with unique characteristics among scientific disciplines, juxtaposing challenges typically encountered in theoretical as well as empirical sciences. This raises questions as to the identification of the most pressing problems for data science, as well as to what constitutes its theoretical foundations. In this contribution, we first describe data science from the perspective of philosophy of science. We argue that the current mode of development of data science is adequately described by what we term the differentiational-expansionist mode. This leads us to conclude that data science concerns the acquisition of scientific theories relating to the application of methods, workflows and algorithms that generate value for users – which we term the integrative view. This definition emphasizes the interdependent nature of human and algorithmic elements in complex data workflows. We then offer four challenges for the future of the field. We conclude that since full control of entire data workflows is unfeasible, attention should be redirected towards the creation of an infrastructure by which data workflows will self-organize in a useful manner.