Rehumanizing Medicine: Medical Virtue

The differentiating human qualities go deep, and they encompass far more than the oft-repeated empathy. Here we resurrect two terms: ‘rehumanization and medical virtue to convey the broader set. The question is what the prefix in rehumanizing is doing. One does not rehumanize what was never dehumanized.

What Was Lost, and How

The erosion long predates machine intelligence and came from the industrialization of care. In an increasingly corporatized ecosystem, the encounter has been reorganized around throughput: visit lengths set by templates and not need; relative value unit targets that price the physician's hour and thereby price attention; documentation that routes the gaze from person to screen; protocols that convert deliberation into compliance. The RVU is emblematic. Designed to measure physician work fairly, it now rewards the short visit and penalizes the patient for whom the physician is most needed. A second erosion runs through the division of labor. Specialization parses a patient's care across ever more specialists, and when twelve clinicians hold a twelfth of a patient, no one is vested in the whole. Connection requires duration, and duration requires someone who stays.⁴⁶

What fragmentation costs is the physician-patient I-Thou: Buber's distinction between meeting another as a subject and handling them as an object of management. An I-It relation is not a moral failure on anyone's part. It is the default that systems produce when nobody is positioned to hold the whole. Restoring that human relation, and the system integration that makes it possible, is what rehumanization connotes. The two are inseparable, since human connection without system integration is sentiment, and integration without connection is mechanical administration.

The relation does not stay dyadic. What Buber described between two subjects now takes place with a third party present in nearly every encounter: the physician-patient-AI triad, in which a machine intelligence contributes whether or not it is acknowledged, having often spoken to the patient first. Reading that third party as a competitor produces the wrong response. A machine cannot stand in an I-Thou relation, since it is not a subject and has nothing at stake. What it can do is occupy the space in which one would otherwise form, or, arranged differently, clear that space of the administrative work that had been crowding it out. The triad is therefore the structure within which rehumanization now has to be achieved, and the design question is whether the third participant is positioned to protect the relation between the first two or to mediate it out of existence.

Precedent and Departure

The term is not new. David Kopacz's Re-humanizing Medicine (2014) set out a framework for transformation at three levels, the self of the practitioner, the practice, and the culture of medicine, arguing that systems trained to relieve suffering had come to generate it. This report inherits that diagnosis and departs from it in one respect, since Kopacz wrote in a pre-AI frame, where rehumanization meant recovering ground lost to bureaucracy. Here the technical layer is about to be absorbed by machines, so what remains of the physician's distinctiveness is the human dimension. Machine intelligence may clear the forest, making rehumanization possible by relieving physicians of the work that displaced it. Whether that space is filled or merely emptied is a system design decision, not a technological one.

Read against the Hippocrates continuum of Chapter I, this is in part a return. The pre-Flexnerian artisanal medicine of Hippocrates 2.0, the Oslerian priesthood of craftsmen, bedside observation as method, the physician who knew the patient and not merely the case, held exactly the integration that industrialization dissolved. The proposal is not nostalgia, since Hippocrates 2.0 also lacked most of what makes medicine effective. It is a recombination: Hippocrates 2.0 sensibility on Hippocrates 4.0 infrastructure.

Why Empathy Became the Proxy, and Why It Is Insufficient

Empathy somehow took center stage in conveying the human dimension, and the most influential instrument for measuring it is the Jefferson Scale of Empathy. Developed by Mohammadreza Hojat and colleagues, it is a twenty-item Likert instrument resolving into perspective taking, compassionate care, and standing in the patient's shoes, translated into some fifty-nine languages and used in more than eighty countries.⁴⁷ Its influence is deserved, since it established that a supposedly ineffable quality could be measured with psychometric respectability and supplied the evidence linking empathy to adherence, trust, and outcome.

It also illustrates a hazard. Empathy became the organizing term for the human dimension of medicine largely because it was the part that could be measured, and a measured proxy displaces the unmeasured whole. In medicine the distinction is sharper than the general one made in Chapter VI. Empathy is a perceptual capacity; rehumanization is a moral and relational condition, and the first does not entail the second. A clinician may perceive distress accurately and still be unable to act on it, unwilling to bear its cost, or absent when it matters. It is also the wrong ground on which to stake the human claim, since of all these qualities, it is the one machines simulate best.

The qualities that constitute rehumanization form a wider field, and they group naturally into six clusters.

• Relational qualities, meaning how the clinician meets another person: empathy, feeling with the patient instead of recognizing their state from outside; compassion, the will to relieve suffering; presence, full undistracted attention; deep listening, attending to the story and the silences; and warmth, which makes a frightened person feel safe.
• Moral qualities, the ethical spine of practice: integrity, courage in advocating or admitting error at cost, humility about the limits of one's knowledge, justice toward those most easily overlooked, and trustworthiness, which makes vulnerability possible.
• Reflective and wisdom qualities: practical wisdom, knowing the right thing for this person in this moment; curiosity about the patient as a mystery rather than a case; self-awareness of one's own biases; and discernment, holding ambiguity without premature closure.
• Existential and spiritual qualities: reverence, sensitivity to meaning and mortality; hope that is realistic; and equanimity amid suffering and death.
• Expressive and sustaining qualities: humor, which reasserts a person's standing as more than a patient; creativity, the flexibility to improvise; and self-compassion, tending one's own woundedness so that empathy is not extinguished by burnout and moral injury.

Two of these are the least discussed and among the least mechanizable.

Spirituality, understood broadly and not necessarily religiously, is the capacity to register that a life is at stake, not a set of parameters, and to accompany a person into questions of meaning and mortality that have no clinical answer. It is engaged at the moments medicine matters most, and the profession has less vocabulary for it than for anything else. The resulting de-spiritualization leaves physician and patient without language for the largest thing in the room, and a system that cannot register transcendence will optimize as though it were not there.

Humor seems trivial until one notices what it does: shared laughter restores reciprocity between two people whom illness has placed in radically unequal positions. Both spirituality and humor require judging a particular moment with a particular person, and both fail catastrophically when got wrong, which is why neither will be delegated. Alongside them belongs an allowance for imperfection, since a profession that permits no error produces concealment rather than excellence.

Underlying the catalogue of human qualities set forth is a broader conception of intelligence itself, as an assemblage of complementary intelligences that are cognitive, emotional, moral, spiritual, social, and aesthetic, of which the profession has cultivated and assessed only the first. Machine intelligence is formidable in the cognitive register and largely absent from the other five, and a physician's distinctiveness will increasingly reside there. This also exposes the inadequacy of professionalism, which names competencies assessed against checklists, where what is called for is the wider category of human-defining qualities, and it argues for the rehabilitation of wisdom as a word the profession may use without embarrassment.

The Arnold P. Gold Foundation, which has perhaps done more than any other body to advance this territory in American medical education, defines humanism in healthcare as clinically excellent care that is kind, safe, and trustworthy, and characterizes the humanistic physician through integrity, excellence, compassion, altruism, respect, empathy, and service.⁴⁸ The framing is instructive as it links compassion and scientific competence, which is this chapter's position too. The argument is that by 2050 the technoscientific and the transcendent physician will be the same person. The two are not in tension.

From Qualities to Virtue

A listing of qualities may be useful as an inventory. It is itself a weak foundation. A list suggests that these are traits some clinicians happen to possess, pleasant when present and forgivable when absent. An older and more demanding framing is virtue, whose application to medicine has a substantial scholarly tradition that the profession has largely allowed to lapse.

The classic statement is Edmund Pellegrino and David Thomasma's The Virtues in Medical Practice (1993), and their approach is the one that matters. Instead of importing a general ethics and applying it to cases, they derive the virtues from the nature and ends of medicine itself. Fidelity to trust follows from the patient's necessary vulnerability; prudence, or phronesis, from the particularity of every case; self-effacement from the asymmetry of the relationship. A physician who lacks these is less competent, not merely less pleasant, in the strict sense that they cannot reliably achieve what medicine is for.⁴⁹ Virtue ethics suits medicine because the relationship is intimate, asymmetric, and founded on vulnerability, which rules mediate poorly, and because practice demands judgment under uncertainty, which rules cannot specify in advance. Whether such virtue survives commercial pressure governs whether it can be taught, because virtues are formed by habituation, and habituation is a property of environments. A system that prices attention will not produce attentive physicians, however well it teaches attentiveness.

The Highest-Order Qualities

No curriculum forms attentive physicians inside a learning environment that rewards throughput. Students learn what the setting practices rather than what the syllabus asserts, and a whirlwind fifteen-minute visit conveys more about the value of attention than a course on attentiveness can undo. This report points to delivery-system reform in Part III, but framing parameters for redesign of the system is beyond this report’s scope.

David Brooks's The Road to Character supplies a distinction that can be mapped onto the physician: the résumé virtues, the skills one brings to a market, against the eulogy virtues, the qualities spoken of at a funeral, whether one was kind, brave, honest, capable of deep love. Recovering the second, Brooks argues, means reviving an older moral realism in which character is built through honest confrontation with one's own weakness.⁵⁰

Applied to medicine the distinction is disturbing, because physician formation of today is very nearly a pure résumé-virtue enterprise: examination scores, class rank, publication counts, procedural volumes, productivity, citation indices. Almost nothing in that apparatus measures the second column, and the familiar consequence is that the profession cultivates the wrong half of itself. The newer consequence in the age of machine intelligence is this: every résumé virtue of medicine is, in principle, machine-attainable; not one eulogy virtue is. A system can hold more knowledge, recall it losslessly, and integrate evidence better than any person. It cannot be answerable, bear the cost of being wrong, or be the same kind of creature as the patient. The profession has spent a century optimizing for the capacities it is about to lose, which makes virtue formation not an adornment of medical education but its core business. Character becomes the profession's differentiated asset.

Rehumanization as Foundation, Compassion as Science

Chapter V framed holistic, fully personalized medicine, the N-of-1 goal, as a problem of data, mechanism, and computation. It is at least equally a relational problem, and the relational half is also very much a limiting factor, contributing to the irreducible N-of-1 gap. Genuine personalization requires knowing what this person's health is for: what they will trade, what they fear, what a good outcome means in a life only they can describe. None of that is recoverable from a genome or proteome, and none of it is disclosed to a clinician who has not earned the standing to hear it. Rehumanization is the obligate foundation for personalized medicine, and the technical layer rests on it. A system that automates the technical layer while the relational layer erodes will have built an extraordinary apparatus for personalizing care to a person it no longer knows. Caring is the core differentiator, and systems must protect it deliberately, through protected time, relational continuity, and roles defined around connection. That protection is a condition of the performance.

The deeper reason for treating the relationship as infrastructure is that the relationship does clinical work. Compassion carries physiological consequences, operating on the patient's state in ways that are measurable and that affect outcome: physician warmth and expectation-setting affect pain report, physiological stress markers, and adherence, and the long literature on placebo and expectation supplies a route by which a relationship changes a physiological state. Its erosion is therefore the withdrawal of an active ingredient, and a system that has priced attention out of the encounter has removed something that was doing work. Bringing the human dimension of medicine back to life is therapeutic in its own right, and the physician who restores it is administering a modality the profession has never costed, licensed, or measured with any seriousness.

Compassion is the portion of that relational work which can be specified, and it should be held apart from the empathy with which it is routinely conflated. Empathy is affective resonance, the registering of another's state, and a machine can now produce a convincing performance of it. Compassion adds motivation and action directed at relieving what has been registered, which is why patients, asked to describe it, describe behaviour: the clinician who sits instead of standing, who leaves the opening question uninterrupted, who names the emotion in the room and then does something about it. Empathy without action is associated with burnout; compassion, which carries the behavioural component, appears protective. Defined this way, compassion becomes a set of trainable behaviours for which a clinician can be prepared and against which performance can be assessed, which is what makes it a legitimate object of a curriculum rather than a graduation-day sentiment.

An evidence base has begun to form around that proposition. Trzeciak and colleagues named the project compassionomics in 2017 and framed it as a hypothesis with a testable design: that compassionate care improves patient outcomes, lowers cost, and protects the clinician. Measurement was the bottleneck for a decade and has now largely been cleared. The Sinclair Compassion Questionnaire, built from a patient-derived empirical model and validated across four care settings, was the first patient-reported instrument to meet formal psychometric criteria, and shorter tools now exist for routine administration. That step is what makes everything downstream possible, since compassion cannot serve as an outcome, a quality indicator, or the dependent variable in a trial without it. Where the relationship itself has been manipulated under randomized conditions, the effect on healthcare outcomes is real and modest. Few trials qualified, and most were never designed to answer the question. The fair statement of the position is a causal effect of small size resting on a thin evidentiary base, with no plausible harm on the other side of the ledger. The response that follows is to run the trials.⁵¹

Training moves the compassion dial. A neuroscience-informed curriculum for resident physicians improved empathy as patients experienced it, and the protocol-based programs developed at Stanford and Emory have since been adapted by medical schools on several continents. The T. Denny Sanford Institute for Empathy and Compassion at UC San Diego was built research-first, establishing the neurobiology before redesigning the curriculum on it, and the Compassion Research Lab at Calgary supplies the conceptual and measurement backbone. The most interesting structural experiment is the newest. The Miriam and Aaron Gutwirth Medical School at the Weizmann Institute of Science, which admitted its first cohort in October 2025, is pioneering programming that elevates compassion into a science and gives its therapeutic potential an evidence base, alongside an education it describes as evidence-based and technology-driven. An MD-PhD program of forty students a year, embedded in a basic-science institute with several hundred laboratories, is a unique setting capable of running adequately powered mechanistic and interventional studies.⁵²

Compassion warrants what medicine gives to therapeutics. That means formal research centers devoted to compassion science and relational healing, investigating mechanism, dose, and durability with the seriousness applied to any other intervention; compassion hardwired into medical education and into the research enterprise, not merely commended in mission statements; self-awareness and reflective practice required as components of formation instead of offered as electives to the willing; and healthcare systems that preserve authentic human caring as a protected dimension of medicine by design. Ethics and compassion then stand as foundational pillars of physician identity in the age of machine intelligence.

An Antidote for Moral Injury

This agenda has a second beneficiary—the physician. The discourse about what modern practice does to physicians has keyed in on burnout, a term that locates the deficiency in the physician and therefore locates the remedy there too, in resilience modules and wellness officers. The deeper wound is moral injury, which has a different origin. Moral injury entered clinical language through Jonathan Shay, a Veterans Affairs psychiatrist working with Vietnam veterans, who described injury of character and defined the injury as the betrayal of what is right, by someone holding legitimate authority, in a high-stakes situation. Shay's veterans generally recovered from horror and grief; what they did not recover from was the violation of themis, the customary order of what is right and proper. This relates to Jameton's moral distress: knowing the ethically correct action and being institutionally prevented from taking it, which accumulates as moral residue across a career.⁵³

Talbot and Dean carried the term into medicine in 2018, with Caplan joining to separate moral dilemma, moral distress, and moral injury as sequential phenomena. Cynda Rushton draws the same sequence at the bedside: burnout is a mismatch between the demands placed on a clinician and the resources available to meet them; moral distress is the state of knowing the ethically correct action and being unable to enact it; and moral injury is what the repetition of the second produces. It is a more corrosive form of moral suffering, linked to guilt, shame, depression, and suicidality. The physician's version of the betrayal is structural. Competing allegiances to the patient, the employer, the insurer, the record, and the productivity target place the clinician in a double bind many times a day: bound by oath to the patient's interest and bound by the RVU to act against it. The clinician does so knowingly while remaining the person the patient holds responsible.

The statistics make the case. Symptoms of burnout were reported by thirty-eight percent of United States physicians before the pandemic and sixty-three percent by 2022, and a physician now spends something close to two hours on documentation for every hour spent with a patient. What corrodes under those conditions is standing as a moral agent, and with it the professional identity built on the promise to put the patient first. The construct is young, its boundary with moral distress unfixed, and its empirical base thin and drawn heavily from one health system, though instruments now exist to measure it. The pandemic established that the phenomenon runs across the whole workforce.⁵⁴

Reframing the problem this way shifts the locus of the remedy. The interventions that hold up at institutional scale are the ones that restore agency rather than simply fortifying tolerance. Mayo's COMPASS groups, in which physicians meet over months to work on what makes the job meaningful and what obstructs it, reduced burnout and the stated intention to leave practice in a randomized trial. The Mindful Ethical Practice and Resilience Academy at Johns Hopkins was built so that nurses acquire the ethical discernment to act as agents of change in their workplaces, its designers explicit that the object is never to fortify clinicians to tolerate unacceptable conditions. The dyad model of management pairs clinical with administrative leadership so that decisions are taken where their consequences are visible. And the United States Surgeon General's advisory on health worker burnout addressed itself to governments, insurers, technology vendors, training programs, and health systems together, which is the correct level of address for a wound the individual did not inflict.

Every element of this report's agenda points the same way. Attention protected rather than priced, continuity preserved, compassion treated as a clinical modality with a research program behind it, formation aimed at the qualities the market does not score. A physician able to deliver the care they know the patient needs is not being indulged; the profession is being repaired. The case for medical virtue therefore cannot rest on patient benefit alone. It is equally the case for the physician's own survival in the work, and it is the strongest reason to treat the clinical environment, and not only the curriculum, as an object of reform.⁵⁵

More Than a Physician Simulacrum

More than a physician simulacrum is needed. In Baudrillard's sense, a simulacrum is a copy detached from any original rather than merely a copy of something real. It generates its own self-referential reality and ending in hyperreality, where the representation is experienced as more authoritative than whatever it once referred to. A system that reproduces the outward form of the encounter while lacking anyone home behind it may satisfy a satisfaction survey and still fail the patient in the way that matters. The risk is that simulated care becomes the standard against which real care is judged. A patient deceived once is the lesser harm.

Three practical steps follow. Measure simulated care against outcomes, not satisfaction, since satisfaction is a metric that a good simulacrum optimizes best. Require disclosure, so that a patient always knows whether a person is at the other end. The patient should know it at the beginning rather than discover it later. And reserve certain encounters for human presence by policy rather than by preference: the delivery of a diagnosis that changes a life, the conversation about stopping treatment, the hours around a death. Policy matters here, because preference under time pressure can skew to the cheaper option, and the encounters that most need a person are surrendered at times to schedule constraints.