The Upend of Theory: Post-theory, Correlationism, and the Epistemic Injustice of Big Data Science

In 2008, Wired Magazine’s editor-in-chief Chris Anderson famously prophesized that big data and the data science methods used to exploit them would cause the end of theory. One and a half decades on, give and take, our scientific methods and theories of the world have not been rendered altogether obsolete. Not even close. However, it is safe to say that big data analytics techniques have made inroads in both academic research, public administration, and business, leaving a strong imprint on the way theories and concepts are perceived and mobilized in problem-solving. In the era of big data science, theories need not be prerequisites of inquiry, but merely tools for framing an excavation for correlations in massive datasets. This change in attitude towards theory use is shifting attention from categories and concepts to model-generated correlations and classifications. The theory-bulwark of knowledge domains is weakened by the domain-agnostic machine and deep learning analytics tools afforded by data science. Correlationism, post-theory thinking, and domain-agnosticism risk contributing to data science-inflicted epistemic injustice with implications well beyond the confines of academia. These changing attitudes toward theory-use are examined through two distinct cases: the UK government’s Covid-19 response and the investment bank J. P. Morgan’s machine learning-driven hedging program, ‘deep hedging’.