On Big Data: How should we make sense of them?

The topic of Big Data is today extensively discussed, not only on the technical ground. This also depends on the fact that Big Data are frequently presented as allowing an epistemological paradigm shift in scientific research, which would be able to supersede the traditional hypothesis-driven method. In this piece, I critically scrutinize two key claims that are usually associated with this approach, namely, the fact that data speak for themselves, deflating the role of theories and models, and the primacy of correlation over causation. In so doing, I will also refer to a recent case history of data mining projects in the field of biomedicine, i.e. EXPOsOMICS. My intention is both to acknowledge the value of Big Data analytics as innovative heuristics, and to provide a balanced account of what could be expected and what not from it. Besides, I also focus on one aspect that today is subject to growing attention, i.e. the opacity that surrounds the algorithms underlying Big Data.