Towards experimental standardization for AI governance in the EU

Abstract

The EU has adopted a hybrid governance approach to address the challenges posed by Artificial Intelligence (AI), emphasizing the role of harmonized European standards (HES). Despite advantages in expertise and flexibility, HES processes face legitimacy problems and struggle with epistemic gaps in the context of AI. This article addresses the problems that characterize HES processes by outlining the conceptual need, theoretical basis, and practical application of experimental standardization, which is defined as an ex-ante evaluation method that can be used to test standards for their effects and effectiveness. Experimental standardization is based on theoretical and practical developments in experimental governance, legislation, and innovation. Aligned with ideas and frameworks like Science for Policy and evidence-based policymaking, it enables co-creation between science and policymaking. We apply the proposed concept in the context of HES processes, where we submit that experimental standardization contributes to increasing throughput and output legitimacy, addressing epistemic gaps, and generating new regulatory knowledge.