This paper motivates institutional epistemic trust as an important ethical consideration informing the responsible development and implementation of artificial intelligence (AI) technologies (or AI-inclusivity) in healthcare. Drawing on recent literature on epistemic trust and public trust in science, we start by examining the conditions under which we can have institutional epistemic trust in AI-inclusive healthcare systems and their members as providers of medical information and advice. In particular, we discuss that institutional epistemic trust in AI-inclusive healthcare depends, in part, on the reliability of AI-inclusive medical practices and programs, its knowledge and understanding among different stakeholders involved, its effect on epistemic and communicative duties and burdens on medical professionals and, finally, its interaction and alignment with the public's ethical values and interests as well as background sociopolitical conditions against which AI-inclusive healthcare systems are embedded. To assess the applicability of these conditions, we explore a recent proposal for AI-inclusivity within the Dutch Newborn Screening Program. In doing so, we illustrate the importance, scope, and potential challenges of fostering and maintaining institutional epistemic trust in a context where generating, assessing, and providing reliable and timely screening results for genetic risk is of high priority. Finally, to motivate the general relevance of our discussion and case study, we end with suggestions for strategies, interventions, and measures for AI-inclusivity in healthcare more widely.

What makes cases of pure risking sometimes wrong? There is a strong intuition that the wrongness of pure risking stands in an explanatory relationship with the wrongness of the non-risky act. Yet, we cannot simply take this for granted insofar as in cases of wrongful pure risking, the risked outcome fails to materialize. To this end, I motivate and develop an under explored approach in the literature that I call Unificationism. According to the Unificationist account that I defend, the fact that pure risking φ is pro tanto wrong is grounded by a general moral fact that φ-ing is pro tanto or all-things-considered wrong, other things being equal. This relationship holds even if and when an agent's risky conduct fails to transpire or culminate into φ-ing ex post. I argue that this Unificationist account captures our explanatory intuition, avoids problems of extensional and explanatory inadequacy that existing alternative faces, and most importantly, renders Unificationism as a plausible view within the ethics of pure risking.

What is wrong with imposing pure risks, that is, risks that don’t materialize into harm? According to a popular response, imposing pure risks is pro tanto wrong, when and because risk itself is harmful. Call this the Harm View. Defenders of this view make one of the following two claims. On the Constitutive Claim, pure risk imposition is pro tanto wrong when and because risk constitutes diminishing one’s well-being viz. preference-frustration or setting-back their legitimate interest in autonomy. On the Contingent Claim, pure risk imposition is pro tanto wrong when and because risk has harmful consequences for the risk-bearers, such as psychological distress. This paper argues that the Harm View is plausible only on the Contingent Claim, but fails on the Constitutive Claim. In discussing the latter, I argue that both the preference and autonomy account fail to show that risk itself is constitutively harmful and thereby wrong. In discussing the former, I argue that risk itself is contingently harmful and thereby wrong but only in a narrow range of cases. I conclude that while the Harm View can sometimes explain the wrong of imposing risk when (and because) risk itself is contingently harmful, it is unsuccessful as a general, exhaustive account of what makes pure imposition wrong.