Inference in the Data Science Era –Do we Need a New Epistemological Debate in the Social Sciences?

Abstract

Where does trustworthy knowledge come from? This article suggests that the answer to this epistemological question has fundamentally changed due to the recent bloom of machine learning (ML) techniques, artificial intelligence (AI), and data science (DS). Where previously the quantitative social sciences were committed to a hypothetico-deductive research logic, machine learning, artificial intelligence, and data science frequently use abductive and inductive inferences and integrate them with deductive and abductive approaches.This article therefore proposes that Social Science needs a new epistemological debate to clarify what role abduction, induction, and deduction can and shall play in scientific inference.To give starting points to this debate, we first present an example of paradigmatic epistemological changes that machine learning and artificial intelligence bring to Social Sciences: We show how the traditional separation between quantitative and qualitative approaches becomes increasingly blurry and integrative due to innovations enabled by ML, AI, and data science.Following that, we present a new framework for integrating data-driven and theory-driven approaches in research with ML and AI.We conclude by discussing whether and how machines can make trustworthy inferences. Following that, we discuss under which circumstances machines can make trustworthy inferences, to provide a starting point for the new epistemological debate in the Social Sciences that we call for.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 1 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.