Anticipating the unpredictable: A qualitative study on ethics experts' perspectives on review challenges related to AI.

Abstract

Background: Artificial intelligence (AI) is increasingly integrated into research across disciplines, raising various ethical concerns related to bias, privacy, transparency, and accountability. However, less is known about how these challenges are addressed within research ethics review practices. This paper examines how ethics experts perceive and manage AI-related challenges in ethics review.

Methods: Drawing on a qualitative study of 13 focus groups with 67 European ethics experts, we applied thematic analysis to identify key tensions shaping current review processes.

Results: Our findings show that AI challenges basic assumptions of ethics review. Participants described AI as difficult to define, rapidly evolving, and not readily accommodated by ex-ante regulatory frameworks. Ethical issues were seen as emerging across the research lifecycle, often after initial review. This creates tensions between rule-based approaches, focused on compliance and risk minimization, and more trust-based models emphasizing researcher reflexivity and ongoing ethical judgment.

Conclusions: We argue that AI highlights epistemic, temporal, and procedural limitations of existing ethics review systems. These findings suggest that ethics review of AI research may benefit from more anticipatory and flexible ethical governance, including iterative review mechanisms, interdisciplinary expertise, and continuous ethical engagement throughout the research process.