Separating Evaluation from Attribution: A Case for Result-Oriented Reform of Academic Assessment in the Age of AI

This is a preprint of the manuscript: "Separating Evaluation from Attribution: A Case for Result-Oriented Reform of Academic Assessment in the Age of AI" Abstract:This paper advances a normative institutional argument: the evaluation function of academic assessment (judging output quality) should be structurally separated from the attribution function (verifying identity, affiliation, and compliance). Currently, these functions are conflated, allowing author identity to influence quality judgments before content has been fully examined—a distortion documented by decades of bias research. The argument rests on three claims. First, scientific quality is a judgment about results; results do not depend on the identity of their producer. Second, using group-level correlations between affiliation and quality to filter individual submissions constitutes a procedural error: academic evaluation judges individual texts, not group averages. Third, identity filtering’s signal basis is eroding in the AI era: as institutional researchers increasingly use generative AI, the quality floor that affiliation once guaranteed is no longer reliable. This paper advocates for the piloting of a sequential two-stage editorial process—blind content review followed by identified compliance review—and argues that competitive dynamics in academic publishing will naturally favor journals adopting more effective evaluation methods. The manuscript has been submitted to a peer-reviewed journal and is currently under review. Keywords: peer review reform, identity filtering, evaluation-attribution separation, generative AI, research integrity