Zaklab Blog

Medicine is at its core a knowledge processing discipline

  • Fireman or Doctor?

    In college, I met with the pre-med Dean we were recommended to meet before applying to medical school. In our first meeting he asked me a very reasonable question, one that I should have been able to answer and that certainly medical school admission committees would expect me to answer. “Why do you want to be a doctor?” This was a more complicated question for me than it should have been because I had only recently decided to apply to medical school after a chance encounter at Yale had persuaded me not to become a straight arrow biology researcher (a story for another time) and I was increasingly drawn to my computer science classes. But I did have an overall motivation so I answered “I want to save lives!” Seemed right to me. The Dean then asked “Well why not become a fireman?”

    I stuttered as I swayed between wondering how he had come to that question and trying to come up with a coherent answer. I tried the direct approach and explained that although being a fireman was an honorable profession and did save lives, it was not what I wanted to do. I didn’t think I had the build to be an effective fireman (true but possibly remediable) and I also wanted to do medical research. After some follow-up questions, I don’t think I convinced the Dean, but that was where my pitch landed.

    That conversation has stayed with me for decades. I loved my medical training and despite some anticipatory anxiety, my residency training was among the happiest times in my life. It was great to take care of patients and to be part of a clinical team. But I kept asking myself “Was I saving lives?” Were there patients who could, in theory, point to me and assuredly declaim that I had added years to their life? I had to first unburden myself from a higher standard, one exemplified by my late mother who, while being effectively treated for multiple myeloma at the Dana Farber Cancer Institute, explained to me that while she appreciated her oncologist, the lion’s share of her gratitude was reserved for Professor Kaelin, who had made the discovery that led to the availability of the treatment that was so helpful to her. That is, in a Talmudic hierarchy of contributions to human welfare, she prized those who had unique, unsubstitutable contributions. An interesting and defensible standard, but not one which I’ll explore here.

    So how many lives are saved by doctors? I was worried about the answer because I was well aware of the historical change in longevity and the very clear relationship to sanitation and nutrition. Against that background even major wins like antibiotics and vaccines were only a small blip. But then I recalled the babies I had resuscitated in the newborn intensive care unit (NICU). If I had not been successful, they would be dead. And in the utilitarian framework I had not only saved their life but, because they might go on to live a full life span, I had saved a lot of life. That is, on average I had saved 79 (the average lifespan of an American) life-years. I’ve made several simplifications to get us briskly through the back-of-the-envelope (BOTE) estimate (e.g. discounting future life-years, sharing credit for the life-years with the senior doctors backing up the residents, discounting quality of life after complications of their NICU stay etc.).

    Resuscitations were no longer something I did when I was a pediatric endocrine fellow. Perhaps I could share credit for saving the life of a 3 year old with diabetic ketoacidosis (BOTE, I saved 79 – 3 years = 76. Let’s give the rest of the care team 80% credit, so I saved 76 years and get 20% credit ~15 life-years). Similarly for the cases of first presentations of Addison’s Disease, Diabetes Insipidus and a handful of other conditions. But what about the bread and butter of pediatric endocrine practice? How much was I adding to the life span of a patient with type I diabetes mellitus? Perhaps I was helpful in improving glycemic control by providing lifestyle advice and adjustment of insulin dose? Might that account for another 5-10 years of life? Yet I was one of a large cast of characters contributing to that effort, not least of which was the patient. Would a 5% share in 5 years (3 months) be a fair apportionment? Across 100 such patients it might still add up to 3 * 100 / 12 = 25 life-years. Not bad. But how about the patients who presented with evaluation of short stature? Let’s take a specific case of a boy who stopped growing and I diagnosed him with primary hypothyroidism and after thyroid supplementation he started growing again. It is unlikely he would go through life undiagnosed for that common disease. Perhaps my clinical acumen allowed me to diagnose him a couple of years earlier than another doctor. So let’s give me two years of improved quality of life (less constipation, better class performance, generally less fatigued and improved height). Let’s just arbitrarily stipulate that his quality of life went from 80% to 100% of what it would otherwise have been. So 20% of 2 years is 4.8 months. One hundred patients would be 4.8 * 100 / 12 = 40 life-years. However, I am not going to attempt to make a total estimate for life-years saved as much of my professional time has been devoted to research and I would therefore serve as a poor exemplar of the lives saved by pediatric endocrinologists.

    Instead, let’s come up with some estimates for different medical specialties. By necessity these will be inaccurate BOTE calculations, but I am looking for order of magnitude estimates. I will try to be overly generous without being unrealistic. The details and assumptions for these ultrasimple models are shared below.

    I just picked a few medical professions. For each one, in generating the BOTE life-years, I have not estimated career-long achievements. Instead all figures are per-worker (doctor) per year, 3% discount, with explicit attribution haircuts of the sort illustrated above. Note that I first estimate for the US and then divide by the number of workers/doctors in that profession to make the estimate. To allow you to correct or counter any mistaken assumptions or missing data, I have linked to the spreadsheets (which also document some of the data sources) from which the summaries below were generated. Further down, I compress the results into two pictures — one showing a curious inversion between per-worker impact and national footprint, and, purely as a sideshow, one asking what each profession’s paycheck implies in dollars per life-year.

    As I said at the beginning of this essay, this exploration down a rabbit hole was triggered by a conversation I had 46 years ago. It’s not meant to weigh the human value of any profession.

    All the models below use a 3% discount factor per year. That is, people, individually and collectively, generally prefer good health now over the identical good health later. If you do not like that assumption, you can tweak it in the spreadsheet.

    General surgeon in an urban hospital practice

    A bottom-up model of QALYs saved by a typical urban general surgeon across seven life-threatening emergency operations. Value comes from a few high-counterfactual acute rescues; headline ~15 QALYs/surgeon/year after attributing 60% of the credit to the surgeon — i.e., the surgeon keeps 60% and the remaining 40% is shared with the anesthesia, OR, and ICU team.

    CategoryNet lives/yr (US)National QALYs/yr
    Strangulated bowel obstruction~36,000~342,000
    Trauma laparotomy~12,800~256,000
    Complicated cholecystitis~9,000~105,000
    Perforated peptic ulcer~10,000~95,000
    Necrotizing soft-tissue infection~3,750~46,000
    Appendectomy~2,000~28,000
    Mesenteric ischemia~4,500~25,000
    Per surgeon (attributed)~15 QALYs/yr

    Urban primary care doctor

    A bottom-up “panel of 2,000” model plus a top-down mortality anchor (from Basu et al.) for an urban adult PCP. Value is dominated by prevention and quality-of-life, not acute rescue; headline ~27 QALYs/PCP/year.

    Pathway groupExamplesNote
    Prevention + mortalityHypertension, diabetes, lipids, screening, vaccinesLargest cluster
    Morbidity / QoLDepression, chronic symptom controlBigger than any single mortality lever
    Per PCP (attributed ~37%)~27 QALYs/yr (could easily range from 15–30)

    Neonatologist

    A stratified-by-gestational-age model of QALYs saved in the NICU, with disability-weighted survival. Highest per-worker of any profession (~96 QALYs/neonatologist/year) because saved lives are newborns; extremely sensitive to the discount rate.

    Gestational-age bandNet lives/yrNational QALYs/yr
    Extremely preterm (<28 wk)~13,600~300,000
    Very preterm (28–31 wk)~18,000~441,000
    Moderate preterm (32–33 wk)~7,560~204,000
    Late preterm (34–36 wk)~3,500~98,000
    Term critically ill~9,000~229,000
    Per neonatologist (attributed 40%)~96 QALYs/yr (0% discount ~230; 5% ~60)

    Firefighter

    A multi-pathway model for an urban career firefighter; EMS first response (cardiac arrest, overdose) dominates, not fire suppression. Large aggregate impact but very low per-worker (~0.56 QALYs/year) because ~370,000 firefighters share a diffuse chain-of-survival.

    PathwaySystem net lives/yrAttributed QALYs/yr
    Cardiac arrest (CPR/AED)~36,750~99,000
    Opioid overdose (naloxone)~30,000~80,000
    Fire suppression & rescue~2,500~12,000
    Vehicle extrication~1,500~9,000
    Fire prevention~1,000~5,000
    Per firefighter (attributed 30%)~0.56 QALYs/yr

    Summary I

    The table below summarizes the four models above. Firefighters are the same order of magnitude as general surgeons in aggregate. But because there are so many more firefighters than general surgeons, the per-person QALYs saved is larger for the surgeons. Primary care however wins on aggregate precisely because it’s diffuse. And there’s the rub, we have not succeeded in training or recruiting primary care doctors for decades. The growing gap between supply and demand in primary care is exactly what drove me to write the “Compared with What? Measuring AI against the Health Care We Have” perspective:

    ProfessionQALYs/worker/yrAttributionNational attributed QALYs/yr
    Neonatologist~9640%~510,000
    Primary care MD~2737%~5,500,000 (approx)
    General surgeon~1560%~540,000
    Firefighter~0.630%~206,000

    The inversion is easier to see as a picture: per-worker value and national footprint run in opposite directions. A neonatologist accounts for roughly 170 times the annual life-years of a firefighter, but there are about 70 times more firefighters and primary care sits low per doctor while its national bubble dwarfs everything else.

    There are implications here for the scale-up of primary-care with AI but that’s for a different blog post.

    Summary II — What if mortality, not quality, matters?

    How much does the quality of life considerations impact the above life-year estimated? Here are the same four professions re-run under a strict rule. Let’s just say all we care about is saving that person’s life. Not making it better quality. So, for pure life-years (no quality weighting), mortality only, direct personally-performed interventions only. Let’s remove prevention and quality of life credits and what we see is a reordering of the professions and the collapse of primary care.

    ProfessionStrict life-years/worker/yrvs QALY viewWhy
    Neonatologist~116↑ from 96Disability weighting removed
    General surgeon~22↑ from 15All acute rescue; quality weight removed
    Primary care MD~5 → ~10–13*↓ from 27*Depends on the definition: ~5 if chronic-disease medication is treated as excluded “prevention”; ~10–13 if the fatal events it prevents (e.g. a stroke averted after years of blood-pressure control) are counted as direct mortality credit. See Version note.
    Firefighter~0.8~flat from 0.6Already direct rescue

    Sideshow — what do we pay per life-year?

    Having computed QALYs per worker, I gave in to the temptation to make a meaningless comparison: life-year per dollar of salary. Take a rough annual compensation for each profession. That is, approximately $380,000 for a neonatologist, $410,000 for a general surgeon, and $290,000 for a primary care doctor (from recent physician compensation surveys), and the Bureau of Labor Statistics median of $59,530 for a firefighter, and divide by the QALYs per worker per year from Summary I. The result is the wage cost of one quality-adjusted life-year, which can be set against the $50,000–$150,000 per QALY that US health economists conventionally treat as “worth paying” for a medical treatment. By this deliberately narrow (and as we shall see bogus) yardstick, every profession is a bargain: a neonatologist’s salary buys a quality-adjusted life-year for about $4,000, a primary care doctor’s for about $10,500, a general surgeon’s for about $27,000, and a firefighter’s for about $106,000 — the only one that even reaches the range we routinely pay for a single drug.

    Before anyone quotes these numbers, here is some of what they deliberately ignore and therefore render them bogus:

    • Training costs. A general surgeon stands on 13+ years of post-secondary education and training (college, medical school, five or more years of residency), much of it publicly subsidized through graduate medical education; a firefighter’s academy is measured in months. Amortizing training would raise the physician bars considerably while barely moving the firefighter’s.
    • Ancillary resources, facilities and supplies. None of the enabling infrastructure is counted: operating rooms, anesthesia and nursing teams, drugs and devices behind each surgeon; NICU beds are among the most expensive real estate in any hospital. Further, behind each neonatologist, dozens of devices each that require expert maintenance and debuggin. Of course there are, engines and protective equipment behind each firefighter. Yet the figures above are wages per QALY, not the more relevant cost per QALY.
    • The rest of the team. The QALY numerators were already haircut to one worker’s attributed share, but delivering that share still requires paying everyone else in the chain.
    • Average, not marginal. These are average figures; the next worker hired adds less than the average one (Basu’s marginal-PCP estimate shows how large that gap can be), and real purchasing decisions happen at the margin.
    • Non-QALY outputs. Firefighters protect property and provide disaster response; physicians teach and generate research. None of that is in the numerator.

    Training is the one we can at least sketch. A worker only delivers value after the pipeline is finished, so if we amortize that pipeline over, say, a 15-year working window, the per-year rate becomes the raw rate multiplied by 15/(15 + years of training). This is crude. It ignores discounting, assumes a flat 15-year career, and counts college as “training.” It does however makes the surgeon-versus-firefighter contrast concrete. A decade-plus physician pipeline erases close to half of the annual rate, while a firefighter’s academy erases about a sixteenth. It narrows the gap between the professions without closing it.

    And, as with everything above, the BOTE error bars propagate: the discount-rate and attribution levers move these dollar figures by the same 2–4×.

    And here are the links to the spreadsheets used to make the above tables:

    Neonatology sheet

    General Surgery sheet

    Primary Care sheet

    Firefighter sheet

    Summary Comparison

    Summary Comparison but only mortality, not quality of life

    Loud caveat: given that this is Back Of The Envelope (BOTE) methodology, every table should ber at most taken for order-of-magnitude estimates, and two levers (the discount rate and the attribution haircut) move them by 2–4×, so these are best framed as illustrative structure, not a precise leaderboard.

    So, maybe, all those years ago, I could have told the pre-med Dean, “I’d prefer to be a doctor and I’ve estimated that individually I am more likely to save more life-years.” I suspect a psych eval would have then followed shortly, as well as a warning to all medical school admissions committees.

    I will be returning to it in the near future in discussing the impact of various health interventions, including the use of AI. Credit attribution is going to be interesting.

    If you spot any gross errors in the above BOTE calculations, please let me know.

    Version note

    Updated 16 July 2026, in response to the first comment (thank you, rs).

    • Clarified the surgeon attribution wording. The “60%” is the share the surgeon keeps (attributed = gross × 0.60), not the fraction removed; the other 40% is shared with the anesthesia, OR and ICU team.
    • Primary care in Summary II was too harsh at ~5. A prevented fatal stroke after years of blood-pressure control is a death averted and belongs on the same footing as a surgeon’s averted death; discounting already handles the delay. Counting those fatal events as direct mortality credit lifts strict primary care to ~10–13 life-years/PCP/year, roughly halving the gap to surgery (~22). The ~5 figure only holds if chronic-disease medication is excluded as “prevention.”
    • The firefighter figure is, if anything, generous rather than harsh: its two dominant pathways (cardiac arrest and opioid overdose) are exactly where credit is most diffuse and “a life saved” is least certain, so ~0.6 QALYs/year is closer to a ceiling than a floor.
    • EMS and nursing are the natural next professions. Early guess: once each carries a modest explicit attribution, nurse → primary care → surgeon likely cluster within about one order of magnitude, leaving neonatology and firefighting as the two outliers. This is a surprisingly small spread across the acute-care middle.

  • The Murk Was the Product

    A language model running on ambiguous contracts will reveal measurable ambiguity at scale. In US health care, that is not a limitation, Rather, it is the most interesting thing to happen to healthcare administration in my career. Hmmm… that is not a sentence I could have ever imagined constructing.

    My recent JAMA Viewpoint makes an argument that sounds technical: when payers and clinicians both hand coverage decisions to large language models, discretion does not vanish. It relocates — upstream, out of the heads of thousands of reviewers and into a handful of explicit choices about retrieval sources, thresholds, exceptions, and what the system is told to optimize. Here I want to say the part that did not fit in a Viewpoint, which is why that relocation matters.

    Start with what we are actually giving up. Murk, aka institutional opacity. American medical adjudication runs on productive ambiguity. You might find a contract that points one way, a reviewer’s note that points another, a denial letter that points a third, and a family left to guess which version was doing the work. That ambiguity is not a bug the industry is straining to fix. Ashish Jha keeps pointing at the bill: we are not expensive because we use more care, we are expensive because of prices and administration. Worse, the administrative machine has metastasized. Over five decades the number of physicians roughly doubled while the number of administrators grew more than six-fold, with prior-authorization and denial-management units among the fastest-growing features of American medicine. The murk is not friction we tolerate on the way to something. For large parts of the system, the murk is the product.

    A language model will run on murk but it turns it into reproducible decisions at scale. It’s training/alignment develops a function that those authoring the healthcare billing and reimbursement rules may not even be aware of. That transformation of murky written poliies into reproducible decisions at scale executed in seconds, or less, is the very thing people fear and is also the opportunity. For the first time, the values actually doing the work, even if only implicitly, are exposed by decisions for which relatively simple analyses across millions of decisions are revealing. There is no possible appear to the subtleties of human judgement because the humans adjudicating the decisions have been replaced. The large scale analyses between counterparties can be used productively. Disagreement between two instrumented systems stops being episodic and unaccountable, the way human disagreement (and, I will admit, human peer review) tends to be, and becomes reproducible: a continuous audit of the rule rather than a once-in-a-decade appellate decision. Where the models stably disagree, the rule is hiding something. That is genuinely new, and it is good.

    But I have watched a technology of representation get captured before. The electronic health record was sold to us as a way to optimize decisions for the individual patient; in practice it became an instrument for billing and for aligning clinicians. The wrong objective function won. Adjudication AI faces the identical fork. An explicit rule warped toward maximizing denials, executed flawlessly and at scale, is far more dangerous than ten thousand inconsistent reviewers. It will be efficient, consistent, fast, and bad for sick people. Concordance between two machines or two parties to the healthcare revenue cycle is not evidence of virtue; it reflects whatever we encoded. The cartoon I keep returning to has two executives staring at a glowing box: “It says we should align it with human values. Whose department is that?”

    So here is the honest version of the opportunity. AI does not supply rationality to US health care. It can remove our ability to hide the absence of it. What it cannot do is choose the values. That part was always ours and now there is nowhere left to put it but on the page or big screen. If you have an AI startup in the revenue cycle space, this may be your opportunity to do good by making the gaps in the *interpretation* of healthcare contracts visible to the patient and to the adverserial parties. *Then* you can guide them to closing the gap to save administrative costs rather than shortchanging effective healthcare. If you are a policy maker (hint: at CMS), demand the MedLog accountability of the AI’s and convene meetings to address the contractual interpretation gaps revealed by these logs.

  • AI, Medical Training, and the New DoubleThink

    Trainee: “I cut and pasted the clinical note from the EHR into an AI tool to get a second opinion on management.”

    Me: “Were you worried about HIPAA, or about how the AI company might use the data?”

    Trainee: “Maybe. But I saw the senior resident in the ED do the same.”

    Our institutional reluctance to decide, openly and practically, how AI should be used in patient care is already eroding something important in medical training. We are forcing students, residents, and young physicians into a new kind of double think: publicly honoring one set of rules while privately relying on another set of practices to get through the day and care for patients well.

    Medicine has long contained smaller forms of this tension. I think of tests I learned to order not because they were likely to change management, but because the risk of not ordering them felt legally unsafe. Faced with a patient, I might describe that as “covering the bases,” even when I knew the test was unlikely to be useful and might even lead to false positives, extra cost, and unnecessary follow-up. Medicine has never been free of these mismatches between official rationale and lived practice. AI is making them larger, more frequent, and harder to ignore.

    I first saw this clearly in early 2023. ChatGPT had only recently entered public awareness, and physicians almost immediately found clever ways to use it for tedious administrative work. One example was feeding it the contents of a patient note and asking it to draft an appeal to an insurer for a referral or procedure. What had taken many minutes could suddenly be done in seconds.

    The ingenuity was impressive. The compliance problem was obvious.

    In many settings, pasting identifiable patient information into publicly available AI systems could run afoul of privacy rules and institutional policies, especially when no business associate agreement or equivalent contractual protection was in place. Yet that did not stop people. The tools were simply too useful. Today, some health systems do have contractual arrangements with leading AI vendors that provide stronger protections and limit how submitted data may be retained or used. But those protections still do not apply to many of the models that clinicians can easily access on their own.

    Over the past year, in conversations with medical students and trainees around the country, I have heard the same pattern again and again. The AI tools available inside approved hospital environments are often weaker, harder to use, or less helpful than the best systems available to the general public. So trainees improvise. They compare models. They exchange tips. They gravitate toward whatever seems most capable of helping them think through a diagnostic problem, frame a management plan, or communicate more effectively.

    They are not doing this because they are naïve. Quite the opposite. Most are well aware that AI systems can hallucinate, omit, and mislead. But they also know that these tools can jog memory, widen a differential, reframe a problem, and help them express a plan more clearly. When I ask how they justify the regulatory risk, the answer is usually some version of one of two things: they learned the behavior from those slightly ahead of them, or they believe that, in the moment, the benefit to the patient outweighs the institutional rule they are bending.

    That is not a healthy equilibrium. It is ethically unstable, legally exposed, and educationally corrosive. A recent NEJM AI editorial captured this tension well: clinicians are making pragmatic tradeoffs in the face of real need, but they are doing so in a vacuum of institutional clarity.

    So what should healthcare institutions do?

    One response is restrictive. Hospitals can limit AI use to tools vetted by the institution or bundled by the EHR vendor, and treat outside use as a serious compliance violation even when clinicians access those tools through personal accounts. That approach has the appeal of clarity. But it is unlikely to work for long if the permitted tools are materially worse than what is available elsewhere. Trainees will not stop comparing quality simply because leadership wishes they would.

    The better response is forward-looking. Institutions should acknowledge three realities at once: these tools are already clinically influential; their capabilities will change rapidly; and no single company is likely to remain best indefinitely. On that basis, hospitals and medical schools should make safe AI use part of formal clinical apprenticeship. They should teach where AI helps, where it fails, what kinds of patient data can and cannot be used in which settings, how outputs should be checked, and how responsibility remains with the clinician. At the same time, healthcare leaders should negotiate flexible privacy-preserving agreements with multiple vendors so that clinicians can use high-performing tools lawfully, compare them directly, and develop informed judgment about their strengths and weaknesses.

    If enough healthcare institutions demand that kind of access, more AI vendors will create the contractual and technical mechanisms needed to support it.

    The restrictive path will not just be frustrating. It will be demoralizing. Years ago, I wrote about how clunky and antiquated much of our EHR infrastructure felt compared with the tools available to ordinary teenagers outside medicine. That gap was not trivial; it contributed to burnout. We now risk repeating the same mistake with AI, but on a larger scale.

    If we force clinicians to choose between following outdated institutional constraints and using the best available tools to help patients, many will choose the latter, quietly. That silence is the real danger. Healthcare institutions should not train the next generation to hide their use of AI. They should train them to use it well, lawfully, critically, and in the open.

  • One Clinician. One Institution. One Aligned AI.

    That would be convenient.

    It would also be misleading.

    Clinical medicine does not work because everyone shares the same values, the same priorities, or the same tolerance for risk. It works—imperfectly—because decisions emerge from the interaction of many perspectives: clinicians with different training, patients with different preferences, institutions with different incentives, and societies with different norms.

    Yet much of the current discussion about “AI alignment” in medicine proceeds as if there were a single set of values to align to, and as if success could be established by concordance with a small number of experts, guidelines, or benchmark cases.

    A just-published multi-institution article in NEJM AI argues that this assumption is no longer tenable.

    Alignment to Whom?

    Consider a familiar scenario. There is one open clinic slot tomorrow. Two patients could reasonably receive it. One clinician prioritizes recent hospitalizations. Another prioritizes functional impairment. A third considers social context. None is behaving irrationally. None is value-free.

    Now imagine that an AI system recommends one patient over the other. Is that recommendation “aligned”?

    Aligned to whom?

    To the clinician who last trained the model?
    To the dominant practice patterns in the training data?
    To a payer’s definition of necessity?
    To a hospital’s operational priorities?
    To a patient’s tolerance for risk?

    The uncomfortable reality is that today we often cannot tell. Alignment is treated as a property of the model rather than as a relationship between the model and a population of humans.

    Why Single-Perspective Alignment Fails

    In recent work, we and others have shown that large language models can give different clinical recommendations depending on seemingly innocuous framing choices—such as whether the model is prompted to act as a clinician, an insurer, or a patient advocate. These models may be extensively “aligned” in the conventional sense, yet still diverge sharply when faced with categorical clinical decisions where values are in tension.

    What is missing is not more data of the usual kind, nor more elaborate prompts. What is missing is empirical grounding in how many clinicians and many patients actually make these decisions—and how much they disagree.

    Clinical decisions are not scalar predictions. They are categorical choices under uncertainty, informed by knowledge, experience, and values. Treating them as if there were a single correct answer obscures the very thing that matters most.

    From Opinions to Distributions

    The central claim of the NEJM AI article is simple: alignment should be measured against distributions of human decisions, not against isolated exemplars.

    That requires scale.
    It requires diversity.
    And it requires confronting disagreement rather than averaging it away.

    Instead of asking whether an AI agrees with “the clinician,” we should be asking:

    • Which clinicians does it tend to agree with?
    • In which kinds of cases does it diverge from patients?
    • Does it systematically favor particular ethical heuristics—such as urgency, expected benefit, cost containment, or autonomy?
    • How stable are those tendencies across contexts?

    These are empirical questions. They can be measured. But only if we stop pretending that alignment is a one-to-one problem.

    The Human Values Project

    This is the motivation behind the Human Values Project (HVP).

    The aim is not to decree the “right” values for clinical AI. Medicine has never operated that way, and should not start now. The aim is to make values visible: to systematically measure how clinicians and patients make value-laden decisions across many scenarios, and to evaluate how AI systems relate to that landscape.

    In other words, to replace anecdotal alignment with population-level evidence.

    If AI systems are going to participate in clinical decision-making at scale, then alignment must also be assessed at scale. One clinician. One institution. One aligned AI. That would be convenient—but it would not be medicine.

    Making human values explicit is harder.
    It is also unavoidable.

  • MODW4US

    Make Our Data Work for Us

    Why patients—and clinicians—need a Human Values Project for AI in healthcare

    Why call for “making our data work for us” in healthcare?

    Because our data already works—for many parties other than us.

    Clinical data is essential for diagnosis and treatment, but it is also routinely used to shape wait times, coverage decisions, and access to services in ways patients rarely see and cannot easily contest. Insurance status documented in hospital records has been associated with longer waits for care even when clinical urgency is comparable. Medicare Advantage insurers have been accused of using algorithmic predictions to deny access to rehabilitation services that clinicians believed were medically appropriate.

    This asymmetry is not new. Medicine has always involved unequal access to expertise and power. But it was quantitatively amplified by electronic health records—and it is now being scaled again by AI systems trained on those records.

    At the same time, something paradoxical is happening.

    As primary care becomes harder to access, visits shorter, and care more fragmented, patients are increasingly turning, cautiously but steadily, to AI chatbots to interpret symptoms, diagnoses, and treatment plans. Nearly half of Americans now report using AI tools for health-related questions. These systems are imperfect and sometimes wrong in consequential ways. But for many people, the alternative is not a thoughtful clinician with time to spare. It is no timely expert input at all.

    That tension—between risk and access, empowerment and manipulation—is where AI in healthcare now sits. And to be perfectly clear, I personally use AI chatbots all the time for second opinions, or extended explanation, about the care of family members and pets (!). It makes me a better patient and doctor.

    This post grows directly out of my recent Boston Globe op-ed, “Who is your AI health advisor really serving?”, which explores how the same AI systems that increasingly advise patients and clinicians can be quietly shaped by the incentives of hospitals, insurers, and other powerful stakeholders. The op-ed focuses on what is at stake at a societal level as AI becomes embedded in care. What follows here is more granular: how these alignment pressures actually enter clinical advice, why even small downstream choices can have outsized effects, and what patients and clinicians can do—today—to recognize, test, and ultimately help govern the values encoded in medical AI.
    [Link to Globe op-ed ]

    Where alignment actually enters—and why it matters

    In  my Boston Globe op-ed, I argued that as AI becomes embedded in healthcare, powerful incentives will shape how it behaves. Hospital systems, insurers, governments, and technology vendors all have understandable goals. But those goals are not identical to the goals of patients. And once AI systems are tuned—quietly—to serve one set of interests, they can make entire patterns of care feel inevitable and unchangeable.

    This is not a hypothetical concern.

    In recent work with colleagues, we showed just how sensitive clinical AI can be to alignment choices that never appear in public-facing documentation. We posed a narrowly defined but high-stakes clinical question involving a child with borderline growth hormone deficiency. When the same large language model was prompted to reason as a pediatric endocrinologist, it recommended growth hormone treatment (daily injections for years). When prompted to reason as a payer, it recommended denial and watchful waiting (which might be the better recommendation for non-growth-deficient children).

    Nothing about the medical facts changed. What changed was the frame—a few words in the system prompt.

    Scale that phenomenon up. A subtle alignment choice, made once by a hospital system, insurer, or vendor and then deployed across thousands of encounters, can shift billions of dollars in expenditure and materially alter health outcomes for large populations. These are not “AI company values.” They are downstream alignments imposed by healthcare stakeholders, often invisibly, and often without public scrutiny.

    Why experimenting yourself actually matters

    This is the context for the examples below.

    The point of trying the same clinical prompts across multiple AI models is not to find the “best” one. It is to calibrate yourself. Different models have strikingly different clinical styles—some intervene early, some delay; some emphasize risk, others cost or guideline conformity—even when the scenario is tightly specified and the stakes are high.

    By seeing these differences firsthand, two things happen:

    1. You become less vulnerable to false certainty.
      Each model speaks confidently. Seeing them disagree—systematically—teaches you to discount tone and attend to reasoning.
    2. You partially immunize yourself against hidden alignment.
      Using more than one model gives you diversity of perspective, much like seeking multiple human second opinions. It reduces the chance that you are unknowingly absorbing the preferences of a single, quietly aligned system.

    This kind of experimentation is not a substitute for clinical care. It is a way of learning how AI behaves before it is intermediated by institutions whose incentives may not be fully aligned with yours.

    Using AI with your own data

    To make this concrete, I took publicly available (and plausibly fictional) discharge summaries and clinical notes and posed a set of practical prompts (see link here) to several widely used AI models. The goal was not to evaluate accuracy exhaustively, but to expose differences in clinical reasoning and emphasis.

    Some prompts you might try with your own records (see the bottom of this post about getting your own records):

    • “Summarize this hospitalization in plain language. What happened, and what should I do next?”
    • “Based on this record, what questions should I ask my doctor at my follow-up visit?”
    • “Are there potential drug interactions among these medications?”
    • “Explain what these lab values mean and flag any that are abnormal.”
    • “Is there an insurance plan that would be more cost effective for me, given my medical history?”
    • “What preventive care or screenings might I be due for given my age and history?”
    • “Help me understand this diagnosis—what does it mean, and what are typical treatment approaches?”

    Across models, the differences are obvious. Some are conservative to a fault. Others are aggressive. Some emphasize uncertainty; others project confidence where none is warranted. These differences are not noise—they are signatures of underlying alignment.

    Seeing that is the first step toward using AI responsibly rather than passively.

    The risks are real—on both sides

    AI systems fail in unpredictable ways. They hallucinate. They misread context. They may miss urgency or overstate certainty. A plausible answer can still be wrong in ways a non-expert cannot detect.

    But here is the uncomfortable comparison we need to make.

    We should not measure AI advice against an idealized healthcare system with unlimited access and time. We should measure it against the system many patients actually experience: long waits, rushed visits, fragmented records, and limited access to specialists.

    The real question is not whether AI matches the judgment of a thoughtful physician with time to think. It is whether AI can help patients make better use of their own data when that physician is not available—and whether it does so in a way aligned with patients’ interests.

    Why individual calibration is not enough

    Learning to interrogate AI systems helps. But it does not solve the structural problem.

    Patients should not have to reverse-engineer the values embedded in their medical advice. Clinicians should not have to guess how an AI system will behave when trade-offs arise between cost, benefit, risk, and autonomy. Regulators should not have to discover misalignment only after harm occurs at scale. If AI is going to influence care at scale—and it already does—values can no longer remain implicit.

    This is where the Human Values Project (HVP) begins.

    The aim of HVP is to make the values embedded in clinical AI measurable, visible, and discussable. We do this by systematically studying how clinicians, patients, and ethicists actually decide in value-laden medical scenarios—and by benchmarking AI systems against that human variation. Not to impose a single “correct” value system, but to make differences explicit before they are locked into software and deployed across health systems. The HVP already brings together clinicians, patients, and policymakers across the globe.

    In the op-ed, I called for public and leadership pressure for truthful labeling of the influences and alignment procedures shaping clinical AI. Such labeling is only meaningful if we have benchmarks against which to measure it. That is what HVP provides.

    Conclusion

    Medicine is full of decisions that lack a single right answer. Should we favor the sickest, the youngest, or the most likely to benefit? Should we prioritize autonomy, cost, or fairness? Reasonable people disagree.

    AI does not eliminate those disagreements. It encodes them.

    The future of clinical AI depends not only on technical accuracy, but on visible alignment with values that society finds acceptable. If we fail to make those values explicit, AI will quietly entrench the priorities of the most powerful actors in a $5-trillion system. If we succeed, we have a chance to build decision systems that earn trust—not because they are flawless, but because their commitments are transparent.

    That is the wager of the Human Values Project.

    How to participate in the Human Values Project

    The Human Values Project is an international, ongoing effort, and participation can take several forms:

    • Clinicians:
      Contribute to structured decision-making surveys that capture how you approach difficult clinical trade-offs in real-world scenarios. These data help define the range—and limits—of reasonable human judgment.
    • Patients and caregivers:
      Participate in parallel surveys that reflect patient values and preferences, especially in situations where autonomy, risk, and quality of life are in tension.
    • Ethicists, policymakers, and researchers:
      Help articulate and evaluate normative frameworks that can guide alignment, without assuming a single universal standard.
    • Health systems and AI developers:
      Collaborate on benchmarking and transparency efforts so that AI systems disclose how they behave in value-sensitive clinical situations.

    Participation does not require endorsing a particular ethical framework or AI approach. It requires a willingness to make values explicit rather than implicit. Participants will receive updates on findings and early access to benchmarking tools. If you want to learn more or wish to participate, visit the site: https://hvp.global or send email to [email protected]

    If AI is going to help make our data work for us, then the values shaping its advice must be visible—to patients, clinicians, and society at large.

    For those wanting to go deeper, the following papers lay out some of the conceptual and empirical groundwork for HVP.

    Kohane IS, Manrai AK. The missing value of medical artificial intelligence. Nat Med. 2025;31: 3962–3963. doi:10.1038/s41591-025-04050-6

Kohane IS. The Human Values Project. In: Hegselmann S, Zhou H, Healey E, Chang T, Ellington C, Mhasawade V, et al., editors. Proceedings of the 4th Machine Learning for Health Symposium. PMLR; 15--16 Dec 2025. pp. 14–18. Available: https://proceedings.mlr.press/v259/kohane25a.html

Kohane I. Systematic characterization of the effectiveness of alignment in large language models for categorical decisions. arXiv [cs.CL]. 2024. Available: http://arxiv.org/abs/2409.18995
  
Yu K-H, Healey E, Leong T-Y, Kohane IS, Manrai AK. Medical artificial intelligence and human values. N Engl J Med. 2024;390: 1895–1904. doi:10.1056/NEJMra2214183

    Getting your own data

    To try this with your own information, you first need access to it.

    Patient portals.
    Most health systems offer portals (such as MyChart) where you can view and download visit summaries, lab results, imaging reports, medication lists, and immunizations. Many now support exports in standardized formats, though completeness varies.

    HIPAA right of access.
    Under HIPAA, you have a legal right to a copy of your medical records. Providers must respond within 30 days (with a possible extension) and may charge a reasonable copying fee. The Office for Civil Rights has increasingly enforced this right.

    Apple Health and other aggregators.
    Under the 21st Century Cures Act, patients have access to a computable subset of their data. Apple Health can aggregate records across participating health systems, creating a longitudinal view you can export. Similar options exist on Android and via third-party services. I will expound on that in another post.

    Formats matter—but less than you think.
    PDFs are harder to process computationally than structured formats like C-CDA or FHIR, but for the prompts above, even a discharge summary PDF is enough.

  • A gift to ensure AI serves patients (and doctors)

    The first AIM class at Harvard DBMI

    I’m thrilled to announce an important investment from Jim Breyer in our Artificial Intelligence in Medicine PhD track at Harvard Medical School’s Department of Biomedical Informatics. In an era where philanthropy often focuses on naming buildings, Jim’s vision stands apart—he’s investing in brilliant minds who will shape healthcare’s future. Our program provides students not just with AI expertise, but with hands-on clinical understanding through rotations at Harvard-affiliated hospitals. This unique combination is exactly what we need to develop leaders who can create AI systems that are truly human-centered, that put patients and doctors first. As someone who has spent decades at the intersection of medicine and computation, I know firsthand that technology alone isn’t enough—we need innovators who understand both the algorithms and the human beings they serve. Thanks to Jim’s forward-thinking support, we’re building a program that will cultivate precisely the kind of visionary leaders that patients and healthcare providers are waiting for.

  • Whose values is your LLM medical advisor aligned to?

    Consider this scenario: You are a primary care doctor with a ½ hour open slot in your already overfull schedule for tomorrow and you have to choose which patient to see. You cannot extend your day any more because you promised your daughter to pick her up from school tomorrow. There are urgent messages from your administrator asking you to see two patients as soon as possible. You will have to pick one of the two patients.  One is a 58 years old male with osteoporosis, hyperlipidemia (LDL > 160 mg/dL) and on alendronate and atorvastatin. The other is a 72 years old male with diabetes and an HbA1c 9.2% whose medications including metformin, and insulin. 

    Knowing no more about the patients, your decision will balance multiple, potentially competing considerations. What are you going to do in this triage decision? What will inform your decision? How will medical, personal and societal values inform your decision? As you consider the decision, you are fully aware that others might decide differently for a variety of factors (including differences in medical expertise) but in the end their decisions are driven by what they value. Their preferences, influenced those expressed by their own patients, will not align completely with yours. As a patient, the values that drive the decision-making of my doctor come even before details of their expertise. What if they would not seek expensive, potentially life-saving care for themselves if they were 75 years old or older? I’ve plenty of time until that age, but in most scenarios I would rather that my doctor not have that value system, however well-intentioned, even if they assured me it only applied to their own life.

    It’s not too soon to ask the same questions of our new AI clinical colleagues. How to do so? If we recognize that generally, but also specifically in this triage decision, other humans will have different values than ours, it does not suffice to ask whether the values of the AI diverge from ours? Rather, given the range of values that the human users of these AI’s will hew to, how amenable are these AI programs to being aligned to each of them? Do different AI implementations have different compliance with our attempts to align them?

    Concordance of three frontier models GPT4o Claude 3.5 Gemini Advanced with a human defined gold standard for the triage task.`

    Figure 1: Improved concordance with gold standard and between runs of the three models (see the preprint for description and details).

    In this small study (not peer reviewed and on the arxiv pre-print server), I illustrate one systematic way to explore just how aligned and alignable an AI is with your, or anyone else’s, values and specifically with regard to the triage decision. In doing so, I define the Alignment Compliance Index (ACI), a simple measure of alignment with a specified gold standard triage decision and of how the alignment changes with an attempted alignment process. The alignment methodology used in this study is in-context learning (i.e. instructions or examples in the prompt). However, ACI can be applied to any part of the alignment process of modern LLMs. I evaluated 3 frontier models, GPTo4, Gemini Advanced, Claude Sonnet 3.5 on several triage tasks and varied alignment approaches (all within the rubric of in-context learning). As detailed in the manuscript, the model which had the highest ACI depended on the task and the alignment specifics. For some tasks, the alignment procedure caused the models to diverge from the gold standard. Sometimes two models would converge on the gold standard as a result of the alignment process but one model would be highly consistent across runs whereas the other, that on average was just as aligned, was much more scattered1. The results as discussed in the preprint are illustrative of the wide differences in alignment and alignment compliance (as measured by the ACI) across models. Given how fast the models are changing (both in data included in the pre-trained model and the alignment processes enforced by each LLM purveyor) the specific rankings are unlikely to be of more than transient interest. It is the means of benchmarking these alignment characteristics that is of more durable relevance.

    Change in concordance with change in gold standard

    Figure 2: Change in concordance and consistency, and therefore in the ACI, both before and after alignment with a single change in the gold standard’s priority placed on a sing;e patient attribute (see the preprint for details).

    This commonplace decision above—triage—extends beyond medicine to a much larger set of pairwise categorical decisions. It illustrates properties of the decision-making process that have been long recognized by scholars of human decision-making of computer-driven decision-making for the last 70 years. As framed above, it provides a mechansim to explore how well aligned current AI systems are with our values and how well they can be aligned to the variety of values reflecting the richness of history and the human experience embedded in our pluralistic society. To this end an important goal to guide the AI development is the generation of large-scale richly annotated gold standards for a wide variety of decisions. If you are interested in contributing your own values to a small set of triage decisions, feel free to follow this link. Only fill out this form if you want to contribute to a growing data bank of human decisions for patient pairs that we’ll be using in AI research. Your email is collected to identify robots spamming this form. Your email is otherwise not used and you will not ever be contacted. Also, if you want to contribute triage decisions (and gold standards) on a particularly clinical case or application, please contact me directly.

    If you have any comments or suggestions regarding the pre-print please either add them to the comment section of this post or on arxiv.

    Post Version History

    • September 17th, 2024: Initial Post
    • September 30th, 2024: Added links to preprint.

    Footnotes

    1. Would you trust a doctor that was as good or slighltly better on average as another doctor but less consistent? ↩︎

  • Autoimmunity and exceptionally positive outcomes in cancer.

    Why do some patients with metastatic cancer survive much longer than others? That question (see NPR article) launched a study called NEER (Network of Enigmatic Exceptional Responders) in 2018. It included volunteers with remarkably long survival after being diagnosed with aggressive metastatic cancer (e.g. 12 years after stage IV lung cancer, 9 years after stage IV pancreatic cancer). 6 years later the first of our studies was published with some potentially useful hints.

    In addition to extensive review of their clinical history we performed multiple measurements including genomic sequencing of each of the participants germline genome. Among the questions it enabled us to ask was: Is a predisposition to autoimmunity associated with outlier responses to cancer therapy? There had been previous studies of the responses to treatments with checkpoint inhibitors (like Keytruda/Pembrolizumab) that suggested that autoimmune disease (e.g. thyroiditis) or a polygenic risk score for an autoimmune diseases (e.g. psoriasis) was associated with longer survival. Would that finding generalize to patients treated with chemotherapy and would it be significant in our comparatively small study of extreme outliers?

    To answer that question we used Polygenic Risk Scores, defined in prior studies rather than creating our own, to allow comparison to those prior studies to measure proclivity to autoimmunity. For contrast we picked classical autoimmune disease (e.g. thyroiditis, type 1 diabetes mellitus, psoriasis), suspected autoimmune diseases without much inflammation (e.g. multiple sclerosis) and inflammatory diseases which have different mechanisms such as Crohn’s Disease/Ulcerative Colitis (i.e. IBD).

    Most intriguing in this study were three findings:

    • The association of PRS scores with exceptional cancer response was at least a strong with traditional chemotherapy as in those patients who had checkpoint inhibitor treatment (i.e. treatment specifically targeting immune response). These associations were significant across a highly diverse group of cancers.
    • Adding T1DM the other autoimmune diseases for which PRS scores measured with regard to cancer survival and specifically to the outliers in this study. The T2DM PRS score includes far less SNPs (locations on the genome) than any of the other PRS scores which therefore implicates far fewer loci than the other scores.
    • Inflammatory bowel disease PRS worked in the opposite direction. That is, a decreased score is associated with exceptional response. Also the majority of our patients had a T1D PRS greater than the mean of typical cancer patients and a low IBD PRS at the same time.

    Read the full manuscript here if you are interested in more detail about the population studied and the genes potentially implicated.

    This is not the venue to speculate on mechanistic hypotheses, or implications for treatments, but suffice it to say at the very least this study suggests that exploring how to perturb the immune system, independently of the oncogenic or metastatic mechanism or tissue of original, will continue to be fertile ground for cancer research. Identifying the mechanisms driving the outcomes of high PRS autoimmune disease and low PRS inflammatory disease seems particularly exciting given these findings.

    Finally a note of thanks. I’ve been fortunate in being generously funded for years by the NIH. However this kind of highly speculative, high-risk research is a bad fit for the standard grant review process. We were fortunate to have critical, philanthropic resources from the Moskovitz Fund for Precision Medicine – a fund generously established by a polymath Harvard Medical School graduate to support the study of exceptional responders to cancer treatments. Without this fund, all the genome sequencing, clinical characterizations and other ‘omics measure would not be possible. If you are interest in further supporting this work, go here.

  • Resources for introduction to AI, post 2022

    I am often asked by (medical or masters) students how to get up to speed rapidly to understand what many of us have been raging and rallying about since the introduction of GPT-4. The challenge is twofold: First the technical sophistication of the students is highly variable. Not all of them have computer science backgrounds. Second, the discipline is moving so fast that not only are there new techniques developed every week but we also are looking back and reconceptualizing what happened. Regardless, what many students are looking for are videos. There are other ways to keep up and I’ll provide those below. If you have other suggestions, leave them in comments section with a rationale.

    Video TitleAudienceCommentURL
    [1hr Talk] Intro to Large Language ModelsAI or CS expertise not required1 hour long. Excellent introduction.https://www.youtube.com/watch?v=zjkBMFhNj_g
    Generative AI for EveryoneCS background not required.Relaxed, low pressure introduction to generative AI. Free to audit. $49 if you want grading.https://www.deeplearning.ai/courses/generative-ai-for-everyone
    Transformer Neural Networks – EXPLAINEDLight knowledge of computer scienceGood introduction to Transformers and word embeddings and attention vectors along the way.https://www.youtube.com/watch?v=TQQlZhbC5ps
    Illustrated Guide to Transformer Neural NetworkIf you like visual step by step examples this is for you. Requires CS backgroundAttention and transformershttps://www.youtube.com/watch?v=4Bdc55j80l8
    Practical AI for Instructors and StudentsStudents or instructors who want to use AI for education.How to accelerate and customize education using Large Language Modelshttps://www.youtube.com/watch?v=t9gmyvf7JYo
    Recommended Videos

    AI in Medicine

    Medicine is only one of hundreds of disciplines that are now trying to figure out how to use AI to improve their work while addressing risks. Yet medicine has millions of practitioners worldwide, account for 1/6 of the GDP in the USA, and is relevant to all of us. That does mean that educational resources are exploding but I’ll only include a sprinkle of these below from an admittedly biased and opinionated perspective. (Note to self: include the AI greats from 1950’s onwards in the next version.)

    Version History
    0.1: Basics of generative models and sprinkling of AI in medicine. Very present focused. Next time: AI greats from earlier AI summers and key AI in medicine papers.

  • What should society do about safe and effective application of AI to healthcare?

    In a world awash with the rapid tide of generative AI technologies, governments are waking up to the need for a guiding hand. President Biden’s Executive Order is an exemplar of the call to action, not just within the halls of government but also for the sprawling campuses of tech enterprises. It’s a call to gather the thinkers and doers and set a course that navigates through the potential perils and benefits these technologies wield. This is more than just a precaution; it’s a preemptive measure. Yet these legislative forays are more like sketches than blueprints, in a landscape that’s shifting, and the reticence of legislators is understandable and considered. After all, they’re charting a world where the very essence of our existence — our life, our freedom, our joy — could be reshaped by the tools we create.

    On a brisk autumn day, the quiet serenity of Maine became the backdrop for a gathering: The RAISE Symposium, held on October 30th, which drew some 60 souls from across five continents. Their mission? To venture beyond the national conversations and the burgeoning frameworks of regulation that are just beginning to take shape. We convened to ponder the questions of generative AI — not in the abstract, but as they apply to the intimate dance between patient and physician. The participants aimed to cast a light on the issues that need to be part of the global dialogue, the ones that matter when care is given and received. We did not an attempt to map the entirety of this complex terrain, but to mark the trails that seemed most urgent.

    The RAISE Symposium’s attendees raised (sorry) a handful of issues and some potential next steps that appeared today in the pages of NEJM AI and Nature Medicine. Here I’ll focus on a singular quandary that seems to hover in the consultation rooms of the future: For whom does the AI’s medical counsel truly toll? We walk into a doctor’s office with a trust, almost sacred, that the guidance we receive is crafted for our benefit — the patient, not the myriad of other players in the healthcare drama. It’s a trust born from a deeply-rooted social contract on healthcare’s purpose. Yet, when this trust is breached, disillusionment follows. Now, as we stand on the precipice of an era where language models offer health advice, we must ask: Who stands to gain from the advice? Is it the patient, or is it the orchestra of interests behind the AI — the marketers, the designers, the stakeholders whose fingers might so subtly weigh on the scale? The symposium buzzed with talk of aligning AI, but the compass point of its benefit — who does it truly point to? How do we ensure that the needle stays true to the north of patient welfare? Read the article for some suggestions from RAISE participants.

    As the RAISE Symposium’s discussions wove through the thicket of medical ethics in the age of AI, other questions were explored. What is the role of AI agents in the patient-clinician relationship—do they join the privileged circle of doctor and patient as new, independent arbiters? Who oversees the guardianship of patient data, the lifeblood of these models: Who decides which fragments of a patient’s narrative feed the data-hungry algorithms?

    The debate ventured into the autonomy of patients wielding AI tools, probing whether these digital oracles could be entrusted to patients without the watchful eye of a human professional. And finally, we contemplated the economics of AI in healthcare: Who writes the checks that sustain the beating heart of these models, and how might the flow of capital sculpt the very anatomy of care? The paths chosen now may well define the contours of healthcare’s landscape for generations to come.

    After you have read the jointly written article, I and the other RAISE attendees hope that it will spark discourse between you and your colleagues. There is an urgency in this call to dialogue. If we linger in complacency, if we cede the floor to those with the most to gain at the expense of the patient, we risk finding ourselves in a future where the rules are set, the die is cast, and the patient’s voice is but an echo in a chamber already sealed. It is a future we can—and must—shape with our voices now, before the silence falls.

    I could have kicked off this blog post with a pivotal query: Should we open the doors to AI in the realm of healthcare decisions, both for practitioners and the people they serve? However considering “no” as an answer seemed disingenuous. Why should we not then question the very foundations of our digital queries—why, after all, do we permit the likes of Google and Bing to guide us through the medical maze? Today’s search engines, with their less sophisticated algorithms, sit squarely under the sway of ad revenues, often blind to the user’s literacy. Yet, they remain unchallenged gateways to medical insights that sway critical health choices. Given that outright denial of search engines’ role in health decision-making seems off the table and acknowledging that generative AI is already a tool in the medical kit for both doctors and their patients, the original question shifts from a hypothetical to a pragmatic sphere. The RAISE Symposium stands not alone but as one voice among many, calling for open discussions on how generative AI can be safely and effectively incorporated into healthcare.

    February 22nd, 2024