It could be argued that the healing relationship model is outdated, as “the notion of patients placing themselves under the care of a doctor and seeking their expert advice has moved to the concept of patients as producing health knowledges and as acquiring expert knowledge so as to manage their illness themselves.” This characterisation of medicine suggests that the doctor-patient relationship has evolved and can seamlessly incorporate AI without altering the character of medical care.

As the practice of medicine changes in the face of emerging technologies, “something of the past is inevitably lost, not always for the worse.” Medicine has long been affected by advances in technology that disrupt the traditional one-to-one, face-to-face model of clinical care between doctor and patient. The Internet, for example, has empowered patients with greater access to medical information, but introduced risks owing to misleading or inaccurate information. Introducing new stakeholders into care relationships is not self-evidently problematic, but must be measured in terms of impact on the healing relationship and the ends of medicine; in other words, in the impact on patient care.

The healing relationship must be understood as an idealistic framework of the relationship between ‘expert’ doctors and ‘vulnerable’ patients. As an ideal, the model is not reflective of the ‘empowered patient’ model of care that has emerged in parallel over the past several decades. Assuming modern medicine is characterised by ‘empowered’ patients eroding the privileged position of doctors as ‘experts’, trust cannot be assumed to exist whenever healing occurs.

Read more

However, the healing relationship describes the motivations of patients to seek professional care, or knowledge and technologies for self-care. Whether addressed through professional or self-directed care, the vulnerability of the patient is not eliminated. Similarly, the fiduciary duties created by this vulnerability do not change when diffused to different sources of expertise, be they medical professionals, databases of medical knowledge and advice, or other technologies and systems supporting self-care such as telemedicine or readily available medical information on the Internet.

Finding new ways to live up to the fiduciary duties of medicine in practice takes on renewed importance in this context and in the future deployment of AI in medicine. Pertinent questions have been asked, for example, about the validity and efficacy of medical knowledge available through internet portals. Furthermore, although medical information is increasingly available through other mediums, the role of expertise as an indication of fidelity to trust does not change. Providers of low-quality medical advice, information or care can be criticised, regardless of format.

On this basis, the healing relationship model can be understood as a description of the moral character and obligations of medical practice, traditionally embodied by health practitioners but increasingly diffused across various platforms and persons, including web portals, consumer device developers, providers of wellness services, and others. Even if modern medicine has moved beyond the single doctor-patient model described in the healing relationship, the obligations of this relationship have not disappeared. Rather, the diffusion and displacement of these obligations by new technological actors in medicine is a cause for concern in considering how best to govern the introduction of AI in medicine. Our notion of the healing relationship could, of course, be revised to give primacy to patient autonomy above all else. However, doing so risks reducing the doctor to a mere service-provider, incapable of exercising the full range of medical virtues and practice-internal norms.

When evaluating the impact of AI and algorithmic technologies on the doctor-patient relationship, choice of metric is key. If measured solely in terms of cost-benefits, or utility, the justification for AI mediation and augmentation of care is straightforward. However, while algorithmic technologies may allow for a greater number of patients to be treated more efficiently or at lower cost, their usage can simultaneously undermine non-mechanical dimensions of care. A distinction can be drawn between those effects of algorithmic systems (and components of utility) which contribute to the good of the patient or medicine as a practice governed by well-established internal norms and codes of conduct, and those which contribute to the good of medical institutions and healthcare services.

The moral complicity that characterises the doctor-patient relationship, wherein treatment is ideally guided by the professional’s contextually and historically aware assessment of a patient’s condition, cannot be easily replicated in interactions with AI systems. The role of the patient, the factors that lead people to seek medical attention, and the patient’s vulnerability are not changed by the introduction of AI as a mediator or augmenter of medical care. Rather, what changes is the means of care delivery, how it can be provided, and by whom. The shift of expertise and care responsibilities to AI systems can be disruptive in many ways, which are explored in the section entitled “Potential impact of AI on the doctor-patient relationship”.