Rand Hirmiz

ABSTRACT This paper contributes to the ongoing debate over the authority of advance directives in cases where patients with dementia express desires that conflict with their earlier wishes. Drawing on Harry Frankfurt's concept of second‐order volitions, I argue that the preferences of the pre‐dementia self (the “then‐self”) should, in most cases, take precedence over those of the post‐dementia self (the “now‐self”)—particularly in instances where the now‐self has lost the capacity to form second‐order volitions and is no longer able to meaningfully repudiate prior values and commitments.

This dissertation focuses on the ethical implications of implementing artificial intelligence in healthcare. Three approaches are considered regarding the role that artificial intelligence ought to play in healthcare: 1. the neo-luddite approach, which urges against the implementation of artificial intelligence in healthcare altogether; 2. The substitutive approach, which favours the goal of ultimately substituting artificial intelligence systems for human clinicians; and 3. The assistive approach, which favours the implementation of artificial intelligence systems in healthcare, but only as a tool to assist, rather than replace, human clinicians. The dissertation begins by looking at what excellence in the practice of medicine entails and deriving some formal duties of clinician, before proceeding to assess the risks or threats that different approaches to the use of AI may pose to the practice. The first position evaluated is the neo-luddite approach. In evaluating this approach, some arguments in its favour are considered, including worries regarding biased algorithms, decreased communication with patients, confidentiality, the resurgence of paternalism, the potential for unethical design, overdependence on AI, and inscrutability. After responding to these concerns and rejecting the neo-luddite approach, the dissertation shifts its focus on to the assistive and substitutive approaches. Before evaluating these two approaches, the concepts of “assistance” and “substitution” are analyzed and definitions of assistive and substitutive technology are provided. With these definitions at hand, the remaining two approaches are assessed, beginning with the substitutive approach, which is then rejected for reasons related to both AI’s limited capacity to foster a strong patient-clinician relationship (due to its limited capacity for compassion, recognition, communication, creating trust, and respecting patient autonomy) and concerns on a more systemic level (including its likelihood to cause responsibility gaps and produce erroneous and biased decisions). The final part of the dissertation focuses on the assistive approach. Issues related to clinician-AI disagreement and the use of opaque decision support systems are addressed and recommendations for handling these issues (derived after a discussion of the epistemic role of assistive AI) are provided. The dissertation concludes with the recommendation that the assistive approach offers a superior outlook for the future of medicine.

Despite the importance of discussions over the epistemic role that artificially intelligent decision support systems ought to play, there is currently a lack of these discussions in both the AI literature and the epistemology literature. My goal in this paper is to rectify this by proposing an account of the epistemic role of AI decision support systems in medicine and discussing what this epistemic role means with regard to how these systems ought to be utilized. In particular, I argue that AI decision support systems are not epistemic superiors, inferiors, or peers. Instead, I recommend that they be classified in an epistemic category of their own, as “epistemic nudges.” With my proposed account of the epistemic role of AI decision support systems, I aim to provide answers the following two questions: 1. How ought disagreements between a clinician and an AI decision support system be handled? 2. How ought responsibility and accountability be allocated when an AI decision support system used by a human clinician in decision-making results in harm?

Paul Bloom has famously argued against the need for empathy in clinicians, while Sally Dalton-Brown has argued that AI need not be capable of empathy in order to be a good carer. In this paper, the capacity for AI to substitute for human clinicians is assessed from a bioethical perspective, primarily through the evaluation of the arguments put forth by Dalton-Brown and Bloom concerning empathy in healthcare. In opposition to both Bloom and Dalton-Brown, this paper argues that (1) empathy is essential to providing good care or deep care (that is, care that goes beyond the mere fulfillment of medical tasks), (2) deep care is a significant aspect of healthcare and yields important benefits, and (3) AI’s lack of capacity for empathy and mutual recognition prevents it from being able to provide deep care for patients, and for this reason, we ought to be extremely cautious about considering allowing AI to ultimately replace human health care providers.