Ethics in Danish healthcare AI policy: A document analysis

Abstract

Introduction: Nations are increasingly turning towards artificial intelligence (AI) systems to support healthcare settings. While nations must then contend with ethical considerations surrounding healthcare AI, they do so in a variety of ways, emphasizing different ethical considerations in different ways. However, there is still limited knowledge on how Scandinavian healthcare AI policy emphasizes ethics. In this paper, we investigate ethics in Danish healthcare AI policy to highlight underlying policy preferences. Methods: We present a document analysis of Danish policy documents relating to AI. We view policy documents' contents as expectations that signal and frame what is perceived as a desirable future with healthcare AI. From 210 policy documents, we extracted data of text snippets related to categories of ethical principles and pipeline stages, as well as articulated reasons for considering ethics. We analyzed the proportions of ethical principles and pipeline stages quantitatively and reasons for considering ethics inductively. Results: The most frequently cited ethical principle was prevention of harm (n = 177), while the most commonly referenced pipeline stage was implementation, evaluation, and oversight (n = 189). Both ethical principles and pipeline stages significantly deviated from equal proportions (p<0.001). Additionally, five primary reasons for addressing ethics emerged in the documents: fit of AI with existing healthcare structures, the potential consequences of AI, its marketability, associated uncertainties, and the perceived inevitability of its adoption. These findings indicate that Danish healthcare AI policy predominantly frames ethical considerations based on the potential consequences of AI deployment. Conclusions: Our study suggests the need for steering Danish, and more broadly Scandinavian, healthcare AI policy toward other views of ethics that integrate non-potentiality.