The End of Information Arbitrage: Why the Best Doctor Is No Longer the One Who Knows Most
For a century, medicine rested on a quiet asymmetry: the doctor held the facts, the patient did not. That asymmetry has dissolved. A patient with a phone now reaches the same survival curves as the professor. So what is the doctor for?
Dr. Sven Jungmann
CEO
A patient sits down across from you and slides a printout across the desk. She has the five-year survival figures, the Kaplan-Meier curves, the protocol number of a trial running in Boston. She found all of it before the appointment, on her phone, in an evening. A generation ago this knowledge lived almost entirely inside the physician's head, and the consultation was, in part, a transaction for access to it. That arbitrage has quietly ended.
For most of the profession's history, the doctor's standing rested on an asymmetry of information. We held the facts; the patient did not. We took a certain pride in being the walking reference — the colleague who could recite the figure without looking it up. Medical knowledge is no longer scarce, though. It is ambient, and it is nearly free.
If a physician still draws their professional pride from knowing the statistic, they are in a quiet kind of trouble. They are competing with a utility that has more memory than any human and costs almost nothing to run. That is not a contest worth entering.
“The value of the doctor lies not in knowing the statistic, but in explaining what that statistic means for this particular patient.”
Probability is not meaning
The pivot turns on a distinction we tend to blur: between probability and meaning. A machine can tell you, instantly and accurately, that a given chemotherapy regimen carries roughly a fifteen percent risk of severe nerve damage in the hands. That is a mathematical fact, and it is the same fact for everyone.
What it means is not the same for everyone. To a software engineer in their twenties, a fifteen percent risk of numb fingertips is an acceptable price for the years on the other side of treatment. To a concert pianist of the same age, it is the end of the only life they have trained for, and it may render the treatment a non-starter. The number is identical. The meaning is inverted. The machine recites the fifteen percent; the physician's work is to find out whose hands these are.
The part the machines keep failing at
Technology has made real progress on the first part of medicine: aggregating the data, reading the images, retrieving the literature. Where it consistently stumbles is the last step — carrying a population average down to a single person. An algorithm is trained on the aggregate. It optimises for a patient who does not exist: the average one.
The living patient is messier than any average. The doctor's enduring role is to stand at that join, to understand a particular life well enough that the model's output can be placed correctly within it. Not to compute the probability faster, but to know which probability matters here, and which can be set aside.
What this asks of the people we hire
This shifts, slowly, what we should mean by a high performer. We long equated it with recall — the most knowledgeable person in the room, measured by the speed of retrieval. Raw retrieval is now something we can buy by the server. The scarce thing is judgement under uncertainty, and the harder thing still is the conversation in which that judgement is shared with a frightened family.
A doctor who can recite the guideline but cannot sit with a family and work out what it means for them is not quite practising medicine. They are reading the internet aloud, and the patient already has the internet. The arbitrage is over. What remains is the part that was always the point.
