Skip to content

Iatrogenics

Definition

Iatrogenics — literally "caused by the healer" — is the usually hidden or delayed net harm produced when an intervention's costs exceed its benefits. Taleb generalizes the medical term into a rule for any domain: "anything in which there is naive interventionism, nay, even just intervention, will have iatrogenics." The concept demands a break-even calculation — probabilistic benefit minus probabilistic cost — that "naive interventionism" instinctively skips in favor of the urge to simply "do something."

In the Book

Chapter 7, "Naive Intervention," opens with a documented case: of 389 children presented to New York doctors in the 1930s for tonsillectomies, successive panels of independent doctors each recommended surgery for roughly 45 percent of the remaining pool, despite a real mortality risk per operation — illustrating "probabilistic homicide" at scale. Taleb traces iatrogenics historically through the death of George Washington (hastened by his doctors' bloodletting) and the story of Ignaz Semmelweis, mocked and institutionalized for observing that hospital births killed more mothers than street births, later dying of the same hospital-borne fever he had warned against. He notes medical iatrogenics rose with "scientific progress" itself — the birth of the clinic concentrated patients and increased "hospital fever" deaths — and connects the concept to the principal-agent problem, where a doctor's or stockbroker's own incentives diverge from their client's welfare. Chapter 21 extends the idea into a general rule for medical decision-making under asymmetric payoffs: intervene aggressively on the very sick (convex payoff, little to lose) and minimally on the healthy (concave exposure, much to lose from side effects).

Why It Matters

Iatrogenics supplies a general test for when NOT to act: an intervention is only justified once its harm, including hidden and delayed harm, is shown to be smaller than its benefit — a calculation people instinctively skip because the "doing something" is visible and celebrated while the harm is diffuse, delayed, and attributed elsewhere. This generalizes far beyond medicine to any system where an agent's intervention (policy, management, therapy, "fixing") can quietly cost more than the problem it was meant to solve, and where the intervener bears little personal downside for that cost.