By Mike Koetting April 4, 2018
I am resuming my series of posts suggesting that the biggest problems facing the world are mitigating inequality, protecting the environment, and reconceptualizing work. In the last post I made before my fall break, I argued that the historic capitalist view of jobs was outdated. https://betweenhellandhighwater.com/2017/09/08/reconceptualizing-work/ This post returns to that discussion.
My first formal empirical activity was in 1956 when I clipped a series of articles in the St. Louis Post Dispatch about what life would be like in 1980 with the intent of keeping them for evaluation in 1980. I carried out my plan. This led to a more general interest in evaluating the accuracy of various futurecasts. Over the years I have developed several generalizations about futurecasts.
- If they imagine a technology, it will probably happen.
- It will almost always take longer than predicted.
- There will still be unanticipated developments that have very big impacts.
- There is very little ability to see how society will react to technological changes, largely because authors are stuck in their own social frames.
This is the context in which I look at claims about the future impact of automation.
Not All in the Future
In earlier posts, I addressed the extent to which automation has already impacted the work force. The most obvious is in manufacturing. As a share of the work force, manufacturing jobs have declined relentlessly for the last 50 years. Automation is estimated to account for 85% of the manufacturing job losses. But automation has also had major impacts in retail, clerical, and other sectors. Numerous recent articles have quotes from banks and others saying they were considering new tools like artificial and intelligence and robotics to reduce workforces.
The job losses from automation are not “one-for-one.” Retail stores go out of business, but technical support and order fulfillment jobs increase. Still, there is net job loss. And, it is very important to keep in mind that even relatively small changes can have outsized impacts at the margin. Imagine that 100 people out of 1000 employees lose jobs because of automation but 50 are hired into other jobs. The net loss is 5%. While on the one hand, that is a small percentage; on the other hand, at a population level differences of that magnitude can make huge differences in, for instance, elections. Additionally, these job shifts may also be accompanied by diminution of wages, benefits or security and the impacts may flow to others not directly affected. Researchers in Boston have shown that extensive use of robots not only reduces jobs directly, but depresses wages and reduces jobs in the surrounding community. Again, the losses may be marginal, but for people without much margin, these losses can compound the social impact.
There Will Be More
Virtually every student of the future of labor anticipates major disruption as the capabilities of artificial intelligence become more and more embedded into the machines around us, ranging from robots to automobiles to appliances.
Estimates of potential job loss from automation range from 9% to 25%. Most estimates for the next 10 to 20 years are very much in the lower end of that range, but that is still significant. Moreover, it seems inevitable more of the work that will be available will require more technical skills and more cognitive capability than today’s jobs.
Will that happen? Obviously we don’t know. Based on my “rules” for evaluating futurecasting, my guess is that it doesn’t happen on this schedule, but that sooner or later something like this will. (Rodney Brooks, former director of the Artificial Intelligence Lab at MIT has a thoughtful article on why AI advances will be much slower than anticipated. But he doesn’t dispute there will be advances, as supported by a recent New Yorker article.) Keep in mind, also, that this will be happening gradually throughout the period, not all at once at the end. So even if a certain percentage of net job loss is 30 years out (as opposed to 10), there will be impacts all through the period. Indeed, we are seeing them now. These smaller changes will continue to impact the social fabric: consider the plausible impacts of further net job loss of “just” 5% over the next ten years.
What do we do?
One possibility is to assume that something will bail us out—the “unexpected” event in my generalizations about futurecasting. It is possible. Unanticipated developments have been postponing this moment of reckoning since the time of Malthus, who predicted we would have reached this point a long time ago. (Ironically, the developments that have kept starvation at bay for large parts of the world have created new threats by fostering circumstances that could lead to environmental catastrophe.)
Simply assuming something will turn up, however, carries significant risks. We know that at least some fraction of the predictions are likely to come true and the longer we wait, the less time we will have to adapt. The irony of hoping for some deus ex machina is that the machina is at the root of the trouble in question.
Unfortunately, conceding that simply waiting for a surprise solution isn’t a great idea doesn’t tell us what we should do.
Although my study of futurecasting does not give me special skills at forecasting the future, it does seem to me that there are at least two broad measures we should consider.
First, we should recognize that work force demands are likely to continue changing. It is much easier to change education and training now in anticipation than bank on retraining at a later date. Since we don’t know exactly how the future will play out, we can’t be overly specific. But we understand generally that people need to develop skills compatible with a highly automated work force. This starts out by insuring students master reading, math and critical-thinking, which will all be more important than ever. Understanding how technology works—at both the macro and micro levels—will be crucial. Today’s students particularly need practice problem-solving, including the development of capabilities for self-direction. As a society, we will also need a flexible, needs-driven, affordable, system of career education. While we think of this as primarily a junior/technical college system, some of this should be part of secondary education.
Nothing in the above is new or controversial. People have been saying these things for a number of years and, while some progress has been made, it is not commensurate with the pace of technological change in the economy. The lack of progress has multiple roots: reluctance to divert resources from traditional educational approaches; relative overall reductions in education funding; reflection of how hard it is, for reasons good and bad, to change large social institutions; and because these are tough skills that are hard to impart. But we are missing an important chance if we don’t actually accomplish these changes.
I think there is an important sub-question about what are the proper roles for children who simply can’t master the relevant curriculum. There will be some jobs that still don’t require much education, so that it is one possibility. But often the issues are more complicated than lack of academic aptitude. Handling these cases will require some thought to minimize social disruption. But it is important to not derail the overall approach by focusing on the problems.
Which brings me to my second point. As a society we need a national jobs plan. For all the talk of the importance of jobs—and it’s something politicians have been talking about for a very long time—there has never been an across the board commitment to treating jobs as a fundamental right (although Franklin Roosevelt talked about it), let alone a serious plan to actually deliver on something like that. Moreover, Americans have historically been suspicious of government planning and the rhetorical contrast of “letting the market decide” versus “Communists and their planned economy” made this kind of planning even more remote.
But maybe a plan for jobs is an issue we should reconsider. That will be the subject of the next post.