A short video has been circulating in our circles: two people stood before a machine that renders the human body in shimmering cross-section. Waveforms, a rotating skull, the nervous system laid bare. It is genuinely impressive technology. It is also, I think, a good place to pause and ask a question that medicine has not answered well in two hundred years. What is a patient?
The worry I want to put on the table is not that this technology is bad. It is that it quietly trains us to mistake a part for the whole.
Reductionism is medicine’s great strength, and its blind spot
Modern medicine works because it reduces. It breaks a sick person into systems, organs, tissues, cells, and finally biomarkers, numbers that can be measured, compared, and acted upon. This is not a flaw. The reduction is precisely what gave us antibiotics, imaging, vaccines, and the ability to catch a tumour before a patient feels a thing. Anyone who has done public health knows that a population only becomes tractable once you can count it.
But a method that succeeds by abstraction carries a permanent temptation, which is to forget that the abstraction was ever a person. A biomarker is not a patient. An organ is not a patient. A scan is not a patient. Each is a true part of the patient, and a part mistaken for the whole is how good medicine quietly becomes incomplete medicine.
AI does not create this problem. It inherits it, and then accelerates it. A machine that diagnoses from images and labs is doing, faster and at scale, exactly what reductionist medicine already does. The danger is not that the machine is wrong. Often it will be more accurate than the tired clinician beside it. The danger is that its very fluency makes the reduction feel complete, as if, once the cells and the curves have been read, nothing of medical importance remains.
The part of the argument that does not survive contact
Let me be honest about where the easy version of this worry breaks down, because a worry worth holding should be able to withstand its own strongest rebuttal.
It is tempting to say that the machine only sees cells and organs, never the whole being. That was true of yesterday’s tools and is becoming less true every year. Modern systems already fold in family history, longitudinal records, medication patterns, and increasingly the social conditions a person lives in. If the entire complaint is that the AI cannot see enough, then the complaint dissolves with the next model, and we will have built our ethics on sand.
So the durable objection cannot be about what the machine can see. It must be about what the machine can be.
What a machine cannot do is be responsible
Medicine is not, at its root, an information-processing task. It is a relationship of responsibility. A doctor is answerable: to the patient in front of them, to that patient’s family, to the community that sends its sick to be cared for and expects them back. The clinical encounter is a covenant, not a calculation. When something goes wrong, a person bears it.
A model can correlate. It cannot be accountable. It does not sit with fear, hold a hand, weigh a frightened family’s hopes against a hard prognosis, or carry the moral weight of a decision afterward. These are not gaps in its training data. They are not problems a larger model fixes. They are simply not the kind of thing a model is. The whole-person dimension of medicine, the patient as someone embedded in family and community, with a life that the disease is only one thread of, lives precisely in this relational and moral space that no amount of computation reaches.
This is the point Prof. Aasim Padela has spent a career pressing, and it is worth noting who makes it: a practising emergency physician with a background in biomedical engineering and in classical Islamic scholarship. He is not a romantic standing outside the technology shaking his fist. He understands the machine, and still insists that a human being is not reducible to what the machine can measure.
A caution against the opposite error
There is a lazy version of this argument I want to refuse, the one where the human doctor is holistic and wise, and only the machine is cold and reductive. That is not true, and pretending it is weakens the case.
A seven-minute consultation, a clinician who never looks up from the screen, a referral that treats a person as a throughput to be cleared: these are reductionism too, committed by humans, every day, in every health system including ours. The contrast that matters is not human versus machine. It is whether the system of care, whoever or whatever staffs it, still treats the patient as an end in themselves or as a problem to be processed.
AI could, in fact, make us more holistic, by absorbing the pattern-matching that exhausts clinicians and freeing them to do the irreducibly human work of presence, judgement, and care. Or it could do the opposite, making the reduction so efficient that the human encounter is optimised away as a costly inefficiency. Which future we get is not a technical question. It is a question of what we believe medicine is for.
Integrating, not surrendering
In the Islamic tradition, the human being is not a sum of organs but an integrated whole. Body, mind, and spirit, held within family and community, owed dignity for what they are and not merely for what their biomarkers say. That is not nostalgia. It is a standard against which to measure any tool we adopt.
So the task is not to reject the machine. It is to keep it in its place, a powerful servant of care, never its substitute. We should let it read the cells better than we ever could, and refuse to let it convince us that reading the cells is the same as knowing the patient.
The technology is not the threat. Forgetting the whole person is. We would do well to understand that clearly, and to build our medicine, and our use of AI within it, around it.
Wallahu a’lam.