Contemporary data-intensive science is often contrasted with traditional hypothesis-driven science and the former is thought to be the successor to the latter. Indeed, advocates of the former champion it as the future for scientific knowledge and practice simply because of the shear complexity of many natural phenomena. Data-intensive science, as its proponents insist, is not subject to this limitation. Rather, through use of computational algorithms to analyze big data obtained from quantifying natural phenomena, accounts of these phenomena are possible without resorting to theories. However, advocates of hypothesis-driven science counter that data-intensive science is incapable of providing meaningful interpretation of its big data without an explanatory theoretical framework, which only hypothesis-driven science can provide (Kitchin 2014).

The question arises as to whether data-intensive and hypothesis-driven science can be reconciled with one another. My thesis is that they can, in a complementary manner. Towards that end, I first discuss the debate between the two approaches to scientific knowledge and practice, initially in terms of empiricism and rationalism. Next, I propose the notions of empirically-directed rationalism and rationally-directed empiricism to forge a complementarity model between empiricism and rationalism. Empirically-directed rationalism involves expected observation or experience that direct formulation of rational inferences, which are then subjected to observation, while rationally-directed empiricism pertains to reason directing the collection of observations, which can then be interpreted intelligibly.

The complementarity model between empiricism and rationalism is then used to propose a complementarity model between data-intensive and hypothesis-driven science, in terms of rationally-directed data-intensive science, which provides a rational framework to incorporate data into data-intensive science and thereby to interpret them intelligibly, and empirically-directed hypothesis-driven science, which provides an empirical framework to generate data that can be used to justify a hypothesis derived from hypothesis-driven science. Rather than contrasting rationally-directed data-intensive science with empirically-directed hypothesis-driven science in binary oppositional terms, a complementary approach for integrating them into scientific practice is proposed. And the complementary relationship between them is cyclical. Moreover, the process is not simply cyclical but also iterative in that both hypotheses and data reinforce and supplement one another as science progresses to a more accurate and practical understanding and knowledge of natural phenomena. In conclusion, both the empiricism of rationally-directed data-intensive science and the rationalism of empirically-directed hypothesis-driven science are two sides of the same epistemic coin.

Reference

Kitchin, R. 2014. Big Data, New Epistemologies and Paradigm Shifts. Big Data & Society 1 (1): 1-12.