Aligning Political and User Visions of Ethically Acceptable Healthcare AI

Artificial intelligence (AI) systems are fraught with ethical issues that inhibit long-term, sustainable use of AI. AI systems might use biased data to reproduce current inequalities or provide answers for which reasoning cannot be explained. Despite these possible ethical issues, AI systems are rapidly being developed and implemented worldwide.<br/><br/>A key focus in this context is healthcare. Healthcare professionals are expected to meet high standards, as their decisions and practices significantly impact patients. Thus, ethical concerns of healthcare AI are extremely visible – for example, biased AI systems may lead to patients receiving inappropriate treatment, potentially leading to life-long adverse health outcomes. However, the advancement of healthcare AI is also politically driven, with healthcare systems such as Denmark’s facing mounting pressure from demographic changes and recruitment challenges.<br/><br/>This implies that political calls for healthcare AI need to address ethical considerations and explore ways to mitigate or promote them. How do perspectives from policy stakeholders on ethically acceptable healthcare AI align, differ, and interact with visions of users?<br/><br/>To explore this, we are analyzing Danish healthcare AI policy documents. We deductively categorize documents’ text for their descriptions of desirable human impacts, i.e. ethics, and their place in the AI pipeline. We inductively analyze arguments for considering ethics. Preliminary results show an unequal distribution of ethical priorities in terms of content and pipeline placement. <br/>Thus, political visions of ethically acceptable healthcare AI might prioritize ethical concerns differently than users. We will investigate this to chart paths for ethically acceptable healthcare AI. <br/>