Much attention has recently been given to discussing the effects, potential and actual, of artificial intelligence on clinical medicine. Many, like Sparrow & Hatherley, have begun anticipating and addressing the challenges arising from integrating AI into medicine, including concerns about privacy, bias, power, responsibility, trust, and empathy. Sometimes a dilemma is presented between, as Hatherley puts it, substitutionism and extensionism: either AI will surpass physicians in performing clinical tasks, thereby making physicians obsolete, or AI will extend and improve upon physicians’ capabilities. But to better appreciate the effects of AI on clinical medicine, one needs, I believe, a more comprehensive understanding of the nature and ambitions of the latter: If we understand what clinical medicine is or should be about, we will be in a better position to understand how AI might influence issues such as privacy, bias, power, responsibility, trust, and empathy, as well as what AI can and cannot substitute for or extend. In this post, I aim to give readers a sense of the complex nature of clinical medicine, with the hope of getting readers to think about what it is or should be before worrying about what it will become in the age of AI. My focus will be on the nature of the physician-patient relationship and the craft of medicine.
It is a perennial question in medicine and medical ethics: how best to conceive of the relationship between physicians and their patients. When paternalistic models of the physician-patient relationship, which place little importance on patient autonomy, fell out of favor several decades ago, fact-provider models, which champion patient autonomy, seemed like a promising alternative. According to the latter models, discussed, for example, by Veatch and Emanuel & Emanuel, physicians merely provide patients with non-value-laden medical information, and patients then choose their preferred intervention based on their own values. However, many, like Emanuel & Emanuel and Savulescu, have criticized fact-provider models, since the assumption underlying these models, according to which there is a clear distinction between facts and values, is untenable. Indeed, not only can physicians not avoid making value judgments, but patients do not necessarily have fixed, known values, and they can make choices that frustrate their own values. Given the problems associated with the fact-provider model, alternative models have been suggested. For example, Emanuel & Emanuel famously suggested both a.) an “interpretive model,” in which the physician elucidates the patient’s values and preferences and helps the patient select medical interventions that realize these values and preferences; and b.) a “deliberative model,” in which the physician helps the patient, through joint deliberation, determine and choose the best health-related values that can be realized in the clinical situation.
Three points are worth emphasizing about the latter two models. First, both models emphasize certain values. The Emanuels’ emphasis on interpreting patients’ values and preferences can be understood in the context of the central role that these values and preferences have attained in medical practice: as Pellegrino has argued, medicine should focus on the good of the patient, which includes not only the patient’s medical good, but, importantly, also the patient’s perception of the good, which concerns his values and preferences. The Emanuels’ emphasis on health-related values raises interesting questions about the nature and source of such values, an issue to which I return below. Second, both models suggest greater equality between physicians and patients, either because patients’ values are incorporated into medicine or because physicians and patients deliberate together. Third, both models suggest a meaningful relationship between physician and patient, one in which the physician cares about the patient and potentially shares a common goal with the patient.
One way to make sense of these elements is to appeal to a long-standing and well-known tradition in medicine that understands the relationship between physician and patient in terms of friendship. This idea is clearly stated, for example, by Seneca, who, in On Benefits, asks why it is “that I owe something more to my doctor and my teacher, but I do not quit my debt by payment,” and answers that “from being a doctor or a teacher they turn into a friend” (6.16.1). But even earlier, for the ancient Greeks, the relation between doctor and patient was one of philia, or “friendship.” Although philia is broader than our contemporary understanding of friendship, the image of the physician as friend to the patient has been prominent throughout the Western medical tradition, as Lain Entralgo has shown. The long history of this tradition may be explained in part by the “Hippocratic ethic,” which has dominated Western medicine. According to the Hippocratic Oath, the physician is required to benefit the sick according to her ability and judgment as well as to keep the sick from harm and injustice. Indeed, as Veatch once pointed out, in such a tradition, the physician’s task is to use her judgment to benefit her patient, which often requires intimate knowledge of the patient and is attained in the context of a relationship of friendship.
Obviously, the relationship is not strictly speaking one of friendship, given, as Davis notes, the lack of mutuality and reciprocity, among other differences. Hence, I prefer to understand the model’s ambitions in terms of the physician being a friend to the patient, rather than as the physician and patient as friends. This more one-sided idea, according to which the physician could and should be a friend to the patient, can incorporate the desiderata noted above. As I have argued elsewhere, the physician–patient relationship can and should have the following characteristics: a.) a common goal to this relationship, that is, one that physician and patient share; b.) certain forms of equality between physician and patient; and c.) an ideal of a caring physician. The idea is that there is a broad goal shared by both physicians and patients, such as “benefiting patients in need of prima facie medical treatment and care.” With this type of goal in mind, physicians and patients can deliberate as equals. This is so because physicians focus on the values over which they have epistemic authority, health-related values, and patients can focus on the values over which they have epistemic authority, their own values. The third component, the ideal of a caring physician, distinguishes this relationship from others that also share a common goal and equality (e.g., relationships between colleagues). To tie together all three characteristics, the physician ought to care about the patient like a friend (rather than like a parent), because this attitude occurs in the context of a social relationship that is constituted in part by a commitment to a common goal, in light of which physician and patient jointly pursue certain goods as equals.
Consider, as a case in point, Quill’s well-known description of a leukemia patient named Diane. Despite having a 25 percent chance of survival, Diane declined treatment. Quill, who had been Diane’s physician for several years, knew her well. When she was diagnosed with cancer, they “together […] lamented her tragedy and the unfairness of life” (692). Quill and Diane met several times to discuss her decision to forego treatment, especially since Quill had previously seen Diane fight to overcome alcoholism and depression and so expected her to change her mind. However, he “gradually understood the decision from her perspective and became convinced that it was the right decision for her” (692). Diane then raised the option of physician-assisted dying:
Diane [wanted] to maintain control of herself and her own dignity in the time remaining to her. […] When the time came, she wanted to take her life […]. I acknowledged and explored this wish […]. In our discussion, it became clear that preoccupation with her fear of a lingering death would interfere with Diane’s getting the most out of the time she had left […]. I told Diane that information was available from the Hemlock Society that might be helpful to her […]. [I]t was important to me […] to be sure that she was not in despair or overwhelmed in a way that might color her judgment. […] We agreed to meet regularly, and she promised to meet with me before taking her life […]. [When] it was clear that the end was approaching […], she let me know […]. [I]t was clear that she knew what she was doing, that she was sad and frightened to be leaving, but that she would be even more terrified to stay and suffer. […] Two days later her husband called to say that Diane had died. […] [W]e talked about what a remarkable person she had been. [693]
If we are approving of this case, it is because Quill was a friend to Diane: he knew her well and cared about her, and he treated her as an equal in sustained deliberation toward a common goal, paying attention to her values and preferences. The case of Diane embodies an ideal relationship between physician and patient. This last point is important: the physician-qua-friend model is not intended to explain current medical practices; rather, it is more plausible to view it, as James notes in connection with the friendship model, as a normative ideal toward which actual physician–patient interactions may aspire.
Now, I have repeatedly talked of “health-related” values, but I have said little about their nature and source. Some, like Schwartz, have argued that medical values encompass not only trivial imperatives such as not treating patients by nonmedical or scientifically futile means, but also more substantive norms determined by the ends of medicine. Consider a patient who comes to a surgeon requesting that her leg be amputated above the knee. She says that her leg has always felt “alien” to her and that her life would be much better without it. Indeed, she compares herself to a person who feels that they have been born in a body of the wrong gender, and wants an operation to put things right, to be “whole.” Should the surgeon operate? One could argue that amputating a limb constitutes a medical means (physicians amputate limbs for various reasons) and that the procedure is not scientifically futile (the limb would be removed, and the patient might feel better). Nevertheless, as some, like Gibson, have noted, the requested procedure might conflict with the ends of medicine and its values. But what are these ends and values and where do they come from? I want to conclude by delineating some possibilities.
Perhaps medicine’s values come from outside of medicine, for example, from a given society. But then medicine can be easily corrupted, for, as Pellegrino argued, medical knowledge and skills might be used to further whatever ends a particular society deems appropriate. Now, it may be the case that universal morality has normative authority over the values of medicine. The ends and norms of medicine would then come from outside of medicine, not in the descriptive sense that a particular society has decided that these are medicine’s ends and norms, but, as Beauchamp notes, in the normative sense that these ends and norms are, in fact, justified because they are derived from justified principles. But even if this is the case, there is a worry that if medicine’s values derive from sources external to medicine, then, as Miller & Brody argue, the physician’s identity as a medical craftsperson will be compromised, which, in turn, will lead to a compromise in their integrity. For even if a physician can appeal to justified external norms, rather than merely doing whatever patients or society tell her to do, the physician qua physician is still only providing medical knowledge and skills, while the values she advocates are external to medicine. Physicians would not have the resources to justify what they can and cannot do qua physicians. Thus, even if a moral system were to show that amputating a healthy limb is morally permissible, many physicians might argue that this is not what physicians ought to be doing qua physicians.
An alternative to this externalist approach is an internalist one. One of the key ideas in discussions of a morality internal to medicine is that clinical medicine is a practice governed by an end definitive of it and by standards of excellence internal to the type of activity it is. This end and these standards determine the goods that cannot be had independently of the practice, as well as the obligations of physicians who partake in the practice. The thought builds on MacIntyre’s conception of a practice, according to which a practice is a “coherent and complex form of socially established cooperative human activity through which goods internal to that form of activity are realized in the course of trying to achieve those standards of excellence which are appropriate to, and partially definitive of, that form of activity” (187). Importantly, since practices are not to be identified with mere technical skills, the successful employment of physicians’ knowledge and skills cannot be dissociated from physicians’ having medicine’s end and standards of excellence in view: that end and those standards unify physicians’ knowledge and skills, as well as medicine’s values, into one coherent and intelligible activity. Accordingly, a good physician has not merely mastered knowledge and skills, but also has the virtues that allow her to attain the goods and further the end and the values that are internal to medicine. A good physician is, in other words, a competent medical craftsperson.
To be sure, various criticisms have been leveled against the idea that there is a morality internal to medicine. These criticisms, raised primarily by Veatch (here, too), pertain to the observations that it is not clear that there is in fact an end that is definitive of the practice of medicine, that external, often societal, factors influence medicine’s values, and that medicine’s values are not necessarily best identified by medical professionals. Many of these criticisms were aimed at Pellegrino’s (here, too) “realist” account of the internal morality of medicine, according to which it is something out there in the world, existing independently of any human construction. But there are more plausible accounts of the internal morality of medicine. For example, according to the “evolutionary” approach, championed by Miller & Brody, the ends of medicine and its morality are not timeless, but rather evolve together with human history. Nevertheless, there is a core ethic internal to medicine that develops historically through interactions between the medical profession and society. In my own work, I have developed a “constructivist” approach to the internal morality of medicine, according to which the norms of medicine are constructed by medical professionals, other professionals, and patients, given the broad end of “benefitting patients in need of prima facie medical treatment and care.” What is important for my current purposes is to highlight the complexities involved in the idea of “health-related” values, especially if these values come from within the practice of medicine—albeit potentially in ways that are influenced by external factors—and the associated idea that the physician is a medical craftsperson, who makes sense of, and, indeed, is a master of, this practice.
In conclusion, clinical medicine is more than the sum of discrete issues such as privacy, bias, power, responsibility, trust, and empathy. In thinking about the possible effects of AI on clinical medicine, one ought to take a more holistic approach and consider what type of relationship characterizes, or ought to characterize, the interactions between physicians and patients. Such considerations will likely involve answering complex questions about medicine’s own values and the role the physician may or should have, qua craftsperson, in determining and acting in accordance with them. Therefore, an answer to the question of whether AI will surpass physicians in the performance of clinical tasks, thereby making physicians obsolete, or extend and improve on their capabilities, ultimately depends on how we understand the value-laden framework of clinical medicine. Indeed, the dilemma between substitutionism and extensionism may prove false if AI transforms the very nature of the craft of medicine and the physician-patient relationship. It could do so by enabling the implantation of the type of ideals discussed in this post, albeit in novel ways.
Nir Ben-Moshe
Nir Ben-Moshe is Associate Professor in the Department of Philosophy and Health Innovation Professor in the Carle Illinois College of Medicine at the University of Illinois Urbana-Champaign. Ben-Moshe’s research falls into two areas. The first area lies at the intersection of contemporary moral philosophy and 18th-century moral philosophy. The second area is biomedical ethics, especially the values at play in the physician-patient relationship.
