Pessimism on pilot programs

The context is health care reform:

Here’s the problem. Both in health care and in education we have lots of examples of low-cost, high quality service. As in other bureaucratic systems around the world, excellence exists, and it’s often known about, acknowledged and even studied. It also tends to have three characteristics: (1) islands of excellence spring up in a sea of mediocrity and they tend to be distributed randomly—they’re not correlated with anything; (2) they almost always exist because of the effort, ingenuity, enthusiasm, energy, and vision of a few people involved in actual production, and almost never are the result of anything that’s happening on the demand side of the market; and (3) (most importantly) they tend not to have any objective characteristics that anyone else can copy.

I agree, and think of the issue in terms of Baumol's cost disease. To a first-order approximation, artisanal industries never increase their labor productivity until they are industrialized. The population-to-doctor ratio is about 390:1 in the US. It's about 300:1 in Europe, whose health care systems are admired by many in America. (On the other hand, doctors in Europe aren't paid as much.) Somalia has a 25,000:1 ratio; can they teach us about efficiency? The population-to-doctor ratio was about 700:1 in colonial America. In contrast, we went from most colonial Americans working on farms to less than 2% of Americans even living on farms today.

You might argue that a doctor that saved your life by prescribing you antibiotics was a lot more productive than a doctor that prescribed leeches to your distant ancestor. However, the modern doctor's value-added was merely his diagnosis of bacterial disease. Most literate people with an internet connection could diagnose most instances of dangerous bacterial disease, so the value-added was probably close to value of the patient's time saved by going to the doctor instead of the internet: small or negative. Some medical diagnoses are much more valuable, but keep in mind that doctors' wages, nurses' wages, and medical lab costs have increased a lot since colonial times.

And doctors still aren't that great at diagnosis. Like Alex Tabarrok, I would prefer to be diagnosed by a computer. Until I am diagnosed by a computer, I don't expect health care costs to get under control. Until my students are taught by a computer, I don't expect education costs to get under control. There are ways to achieve cost control without AI, but they are not going to happen.