Ethical Responsibility in Medical AI: A Semi-Systematic Thematic Review and Multilevel Governance Model

Abstract

<b>Background</b>: Artificial intelligence (AI) is transforming medical practice, enhancing diagnostic accuracy, personalisation, and clinical efficiency. However, this transition raises complex ethical challenges related to transparency, accountability, fairness, and human oversight. This study examines how the literature conceptualises and distributes ethical responsibility in AI-assisted healthcare. <b>Methods</b>: This semi-systematic, theory-informed thematic review was conducted in accordance with the PRISMA 2020 guidelines. Publications from 2020 to 2025 were retrieved from PubMed, ScienceDirect, IEEE Xplore databases, and MDPI journals. A semi-quantitative keyword-based scoring model was applied to titles and abstracts to determine their relevance. High-relevance studies (n = 187) were analysed using an eight-category ethical framework: transparency and explainability, regulatory challenges, accountability, justice and equity, patient autonomy, beneficence-non-maleficence, data privacy, and the impact on the medical profession. <b>Results</b>: The analysis revealed a fragmented ethical landscape in which technological innovation frequently outperforms regulatory harmonisation and shared accountability structures. Transparency and explainability were the dominant concerns (34.8%). Significant gaps in organisational responsibility, equitable data practices, patient autonomy, and professional redefinition were reported. A multilevel ethical responsibility model was developed, integrating micro (clinical), meso (institutional), and macro (regulatory) dimensions, articulated through both ex ante and ex post perspectives. <b>Conclusions</b>: AI requires governance frameworks that integrate ethical principles, regulatory alignment, and epistemic justice in medicine. This review proposes a multidimensional model that bridges normative ethics and operational governance. Future research should explore empirical, longitudinal, and interdisciplinary approaches to assess the real impact of AI on clinical practice, equity, and trust.

Ähnliche Arbeiten