Research on ethical dilemmas and governance pathways of artificial intelligence in the medical field

Artificial intelligence (AI) is increasingly used in clinical practice, raising ethical concerns about algorithmic opacity, accountability, informed consent, and professional autonomy. Existing ethical and governance frameworks provide important normative direction, but questions remain regarding how their principles can be translated into operational guidance for specific medical AI technologies and frontline clinical contexts. This study examines ethical dilemmas perceived by healthcare professionals in AI-assisted practice and considers practical approaches to governance. This study employed a qualitative research design using semi-structured interviews to examine hospital staff’s perceptions of ethical issues surrounding medical artificial intelligence. Interview transcripts were analysed using a deductive–inductive thematic analysis supported by charting in a framework matrix. Latent Dirichlet allocation (LDA) topic modelling was used as an exploratory, unsupervised complement to the qualitative analysis. The deductive–inductive thematic analysis identified six themes, while LDA topic modelling identified five topics. Comparison of the two outputs showed semantic overlap as well as differences in thematic boundaries and granularity. Clinical practice risks emerged as a separate LDA topic but not as a separate qualitative theme, whereas informed consent dilemmas and ethical governance expectations were identified as distinct qualitative themes but not as separate LDA topics. The participating professionals described perceived gaps in current institutional arrangements for accountability, informed consent, and oversight. Based on the interview findings and existing ethical frameworks, we propose two complementary approaches: integrating ethical requirements across the medical AI lifecycle and establishing a multi-actor, risk-based review process. These findings may inform policymakers, healthcare institutions, and technology developers seeking to translate ethical principles into clinical governance. The proposed framework has not yet been implemented or evaluated and requires empirical testing in future research.