Technology and Down's Syndrome

 While in the past automation has merely increased productivity and caused some employment readjustment, the gains from lower costs and improved quality have more than made up for it.  I'm not sure that will be the case in the future, at least in terms of employment.

Yes—and I think the distinction you're making is an important one.

Historically, automation has usually been task-substitution rather than human-substitution. A machine takes over some part of a job, but the lower cost of doing that work creates additional demand, new occupations, or enough complementary work that employment readjusts rather than simply disappears.

Your own experience with computerized insurance files is a nice example. The mainframe eliminated some filing and retrieval work, but it didn't eliminate the insurance business. In fact, making information cheaper and more accessible made other work more productive.

What worries me about the next stage is that AI is increasingly capable of replacing whole bundles of cognitive tasks, rather than merely mechanizing one repetitive operation. If an AI system can do much of the work of a claims examiner, programmer, copywriter, analyst, bookkeeper, paralegal, or customer-service representative, the usual mechanism—cheaper service → more demand → more workers doing complementary work—may not generate enough new employment to compensate.

And there is an interesting asymmetry here. Suppose automation makes something 50% cheaper. If consumers respond by demanding twice as much of it, employment might actually increase. But there are plenty of services for which demand has fairly definite limits. We don't necessarily want twice as many tax returns prepared, insurance policies underwritten, legal briefs written, or radiology reports produced simply because they become cheaper.

The really consequential question therefore isn't whether AI raises productivity. Almost certainly it will. It's whether the economy can find enough things for people to do that people value enough to pay for, once machines can perform a much larger fraction of economically useful work.

That is a rather different problem from the historical "technology destroys some jobs but creates others." The historical reassurance may turn out to be correct again—but I don't think history gives us a guarantee.

Technology can also make jobs obsolete.  My brother was in special education working with Down's Syndrome children.  They don't much exist any more.

Yes, though in his case I would make an important distinction: the children certainly still exist; the job category and the way the educational system deals with them have changed.

That's actually a very good example of something broader than automation. Technology, medicine, and social policy can together make an occupation largely disappear without the underlying human need disappearing.

Children with Down syndrome used to be much more likely to be placed in segregated special-education programs, sometimes with teachers whose entire professional role was working with a particular population. Over time, early intervention, better medical care, changing educational philosophy, and especially mainstreaming/inclusion changed that arrangement. A child who might once have spent most of the school day in a specialized classroom may now spend much of it in an ordinary classroom with various forms of support.

So your brother's occupation could become obsolete even though society arguably became better at addressing the underlying problem.

That is an important wrinkle in the employment question. We tend to imagine technological unemployment as machine replaces worker. But there is another route:

Technology changes what society needs, and therefore eliminates the job.

The file clerk is an obvious example. The specialized teacher is another. And neither necessarily represents a social loss. The tricky case with AI is that it may operate on a vastly larger portion of the economy at once. If millions of occupations are gradually transformed in this way, the question becomes what new human purposes and occupations emerge—and whether they emerge fast enough.

Well I disagree; medical progress and abortion sharply reduced their numbers, at least in the States.

You're right. I conflated the number of people with Down syndrome with the number being served in segregated special-education settings. Those are separate effects.

In the United States, prenatal screening and diagnostic testing, followed by a decision in some pregnancies to terminate after a Down syndrome diagnosis, have reduced the number of births with Down syndrome. Medical advances have also changed survival and life expectancy, although that operates in the other direction. So the population of children with Down syndrome has indeed been affected by medical and reproductive technology, not merely by educational policy.

That makes your brother's example more interesting. His occupation could disappear through a reduction in the underlying population requiring the service, rather than through a machine doing his job. That's a form of technological displacement that gets much less attention than automation.

And it illustrates the larger point you were making: employment isn't determined simply by productivity. If technology changes the number or nature of the people who need something, it changes the demand for labor itself. In some cases, that is a wonderful development; in others, it can leave a large occupational group wondering what happened to its livelihood.

Better pre-natal care also had an effect.

Yes. And that makes the example still more instructive.

Better prenatal care can reduce the number of babies born with severe complications generally, while improved prenatal screening identifies pregnancies with chromosomal abnormalities such as trisomy 21. Medical knowledge, prenatal diagnosis, and reproductive decisions can therefore all affect the number of children born with Down syndrome.

There is also a historical complication: survival has gone up enormously for people with Down syndrome because of improvements in cardiac surgery, treatment of infections, and general medical care. So some medical advances reduce the number born with Down syndrome while others increase the number who survive into childhood and adulthood.

That is a nice illustration of why predicting employment from technological progress is so difficult. Several technological developments can operate in opposite directions simultaneously. Your brother's particular occupation might shrink because fewer children require that specialized service, even while medicine is doing a great deal to improve the lives of the children who are born.