Mind the Gap: “Traditional” vs. Computational Research Logics in Fake News Detection

Abstract

Fake news detection has become an acutely important goal in both academic studies and editorial practice, creating a research area that comprises journalistic debunking of fakes, cross-disciplinary fact-checking projects, and automated efforts of fake news detection. However, with the growth of these industries, an epistemological gap between “traditional” (conceptual, qualitative, quantitative, or mixed-methods) and computational studies of detecting fakes has been deepening. We describe the two divergent logics of fake definition and detection. In particular, international regulation, industrial fake detection, and most media studies legitimize the “blurred border” between fact and interpretation, warning against too strict elimination of fakes and preserving freedom of expression. Computational methods, in their turn, are better in automated fake detection but are “yes/no”-oriented, often ignoring the variety of interpretive forms in public communication. Rooted deeper than just in individual research designs, the divergence of logics sharpens when the public need of clear-cut fake detection runs into freedom of interpretation that results from centuries of struggle for standards in public speech and journalism. Employing critical reviewing of 45 conceptual academic and industrial writings, we outline the major shortcomings of the lack of clear textual markers for fake news in “traditional” media studies, on one hand, and of the “yes/no” logic in computational fake detection, on the other. We propose a five-pillar epistemological framework for fake detection, including true/false, fact/interpretation, discrepancy/solidity, media/user-generated evidence, and human/AI authorship dimensions.