Parents arrive in clinic already knowing a good deal. It used to be from online forums; now they search, and they ask an AI, before they come. The pattern is familiar by now — hearing a diagnosis for the first time, going home and reading through the night, and returning to the next appointment with questions written down. Over the years I have watched the way parents come by their information change.
For a long stretch it was disease-specific patient forums and online communities. Word accumulated there: this professor is good, that hospital finished it in one operation, this one had to redo it. In the process, what came to decide the choice of surgeon was not research output or an accumulated surgical record or technical skill, but the experience and advice of a handful of parents — and that stayed on my mind for a long time. Word of mouth built the authority first, and that authority then became the basis of the choice.
With time, though, my view softened somewhat. The reason parents asked one another was that there was nowhere else to ask. Consultations are short, and a diagnosis and an operation heard for the first time are hard to take in at once. The questions arrive after you get home, and by then there is no one to put them to. The forums filled that gap. Looking back, the gap was not made by the parents but left by us.
What is within reach, and what is not
So I have welcomed the arrival of AI. Put yourself in the position of a parent whose son has just been diagnosed with hypospadias and search: what the condition is, when it is operated on, what complications can occur — all of it comes back well explained. Information that once meant asking several people and digging through old posts now arrives organised, in seconds.
Yet the same search still returns nothing on certain things. Which techniques a given centre uses and what strategy it follows; whether it has the resources a staged repair or a revision would require; and therefore how it would approach this particular child — the substantive information a choice actually rests on was nowhere to be found. What filled the top of the results was each hospital's account of itself, and those are not the kind of thing that can be set against one another. The shape of the gap has changed; the gap has not gone.
And more information does not by itself lead to a better decision. There are three distinct steps here.
Finding the papers and the guidelines is a matter of information.
Understanding which patients a study was done in, and how far it applies to this child, is a matter of knowledge.
And deciding which of several options to apply to this child now is a matter of judgement.
AI already does the first very well. The second will improve quickly. But the hardest part of actual practice usually lies in the third.
Before questions that have no answer
Medicine holds many questions that no amount of searching resolves into a single answer. In hypospadias alone: how many degrees of curvature must be corrected, when the urethral plate should be preserved and when divided, what ought to be called a good result — the literature does not give one answer. Feeding an AI several hundred papers changes nothing. Not because the AI falls short, but because medicine itself does not yet hold a single answer to these questions.
And yet, in front of a patient, a decision has to be made. With the same degree of curvature, the choice can differ with the child's age and anatomy, whether he has been operated on before, the growth still ahead of him, and what the family holds important. To this is added what a surgeon has learned from seeing similar patients over many years.
Interpreting those factors against one person's situation, setting out the advantages and drawbacks of each option, and then thinking through together which direction suits better — this is not the same as looking something up. That process is judgement, and judgement carries responsibility.
What is a doctor for?
It will become steadily harder to describe a doctor's role purely as telling people which drug to use, which test to order, which operations exist. That kind of information is likely to be delivered faster and more systematically by AI.
What will move to the centre instead, I think, is judgement: distinguishing what is established from what remains uncertain, weighing evidence that points in different directions, and setting out the options with the patient's wishes, their circumstances and one's own experience all in view.
The other thing that matters is telling people what to ask. A parent can ask what the success rate is, but has little way of knowing what that figure counted as success. They can ask whether their son's case is severe, without knowing which factors actually determine severity. It is not that they cannot ask — they have had no occasion to learn which questions matter.
So this site does not try to answer everything. What I want to leave here, as accurately as I can, is what is known and what is still uncertain — and what is worth asking in the consulting room.
However much better AI becomes at supplying information and at organising it, the work of interpreting what that information means for one particular person, and of thinking through the options with them, will remain.
A doctor will end up less the person who holds information and more the person who gives it meaning, and who works out that meaning together with the patient.