ZICO: A Credit Scoring Approach to Detecting Zombie Papers

This paper proposes a new framework for assessing scientific credibility using probabilistic retraction risk. We introduce the Zombie Integrity and Credibility Oversight (ZICO) framework, a multidimensional credibility score inspired by credit-scoring models in finance. Rather than relying on isolated indicators of misconduct, ZICO integrates heterogeneous signals extracted from the semantic characteristics of scientific manuscripts and the broader organization of scientific production to estimate the ex ante probability that a publication will eventually become problematic or be retracted. We apply the framework to a corpus of publications in economics and finance. Our findings show that retraction risk is driven primarily by linguistic and semantic characteristics rather than by conventional metadata, such as authorship or institutional affiliations. In particular, measures of readability, lexical diversity, textual complexity, and citation practices emerge as the strongest predictors of scientific credibility. These signals capture deeper structural and stylistic irregularities rather than differences in English language proficiency, thereby highlighting semantic organization as a robust marker of research integrity. We further develop a dynamic version of ZICO that continuously updates credibility assessments, allowing the framework to operate as an adaptive early-warning system for editorial screening. More fundamentally, our framework shifts the assessment of scientific integrity from the ex post detection of misconduct to the ex ante probabilistic evaluation of manuscript credibility. Beyond its predictive performance, ZICO provides a transparent and interpretable framework for editorial decision-making, research evaluation, and the governance of research integrity.