Labor-Displacing Technology Is Good for Labor

The Brookings Institution published a report this year titled “Apple Pie and Motherhood Are Good” by three MIT economists (two of whom are Nobel laureates). Oh no, wait. It was actually titled “Building pro-worker AI.” Damn, talk about pushing the Overton window. After all, what jerk wants anti-worker AI?

I guess I’m that jerk because I disagree with nearly everything in their report.

For the report’s authors, a pro-worker technology is one that makes human skills and expertise more valuable, while a pro-worker policy is one that steers innovation away from displacing workers or devaluing their existing skills. This is because they hate pain and because, with AI, “the pain will be magnified if this automation occurs rapidly.”

Under this logic, tractors, assembly lines, electronic telephone switches, automatic bowling pinsetters, and e-commerce are all pain-inducing technologies. Tech becomes suspect whenever it eliminates particular jobs or devalues specific skills.

And, of course, the authors repeat the standard bromides about technology and productivity: Income inequality has grown, wages have stagnated, and labor’s share of income has fallen. Data continue to show that treating automation as the principal cause of these developments is incorrect or exaggerated. Yet the authors warn that “AI will accelerate these worrisome trends.”

To be sure, no one ever said that losing one’s job is painless. I know it’s not: I lost mine when the Office of Technology Assessment (OTA) closed, and it was not high on my list of pleasurable experiences.

But as they say in training, no pain, no gain. The authors want the equivalent of getting stronger—in this case, richer—without the exertion of an actual workout.

The paper groups technologies into five broad categories: (1) labor-augmenting, (2) capital-augmenting, (3) automating, (4) expertise-leveling, and (5) new task-creating.

But surprise, surprise, these AI Luddites effectively reject four of the five groups, deeming the technologies they include potentially harmful to workers. Apparently, new task-creating tech is the only pure category, which the authors describe as “unambiguously pro-worker, meaning that it creates only winners.”

What? Where do they think productivity growth and higher living standards come from? A great deal comes not from creating new tasks but from making old tasks more labor-efficient. Their standard suggests we should not have developed tractors, which their framework would classify as a labor-augmenting technology, but instead should have given farmers a better tool for planting seeds by hand.

Of course, they would shudder at being accused of Luddism, so they go out of their way to say that the other four “bad” uses of technology are not “necessarily socially undesirable, but they pose additional labor market trade-offs.” You mean some tasks shifting from workers to machines, workers moving into other work, and higher per capita incomes as a result?

It’s kind of amazing that three scholars with PhDs could get it so wrong. But that is what maternalist groupthink produces: No one should ever lose a job. Worse, their framework largely treats displacement as the end of the story. Once a worker is laid off, the poor sap is evidently finished and headed straight for the UBI dole to survive.

The fact is, most displaced workers get new jobs. The most recent Displaced Workers Summary from the Bureau of Labor Statistics found that roughly 66 percent of workers displaced from jobs they had held for at least 3 years were reemployed. That rate rose to 74.5 percent for workers ages 25 to 54. And among those reemployed in full-time jobs, the majority earned as much or more than they had before. Sure, those numbers may not erase the costs of transition, but they demolish the idea that displacement means permanent exclusion from work.

But in the authors’ world, if an enterprise employs fewer workers, it is treated as a permanent disaster rather than transitional friction. In reality, technology-driven displacement is the economic equivalent of DOMS (delayed onset muscle soreness) after leg day: painful for workers and communities in the short term, but a temporary symptom of progress as the economy adapts and grows stronger.

So, the authors want AI technologies that “collaborate with workers, enabling them to (a) perform more-sophisticated tasks related to their jobs, (b) tackle new tasks, and (c) acquire new expertise.”

Then comes the wage argument. The report states: “Even if demand for cable installation soars as robotic automation lowers prices, electricians will not be needed to do this work anymore.” And so, the argument goes, cable installers’ expertise will be devalued and their wages may fall.

Wow. Let’s break that down:

1. Technology adoption must be evaluated from society’s perspective, not just from a single firm’s or worker’s perspective. If society can produce the same output with less labor, productivity rises. If we need fewer electricians to install the same amount of cable, those workers can do other work, increasing output and, over time, per capita GDP.
2. Labor markets and skills are not fixed. When demand for an occupation declines, a strange thing happens: Workers shift occupations and find jobs that use related skills.
3. Many occupational skills transfer relatively easily across industries.

Again, that does not mean technological disruption is painless. It means the proper response is to enhance workforce training and adjustment policies to help workers transition successfully—not to suppress productivity-enhancing technology.

But rather than accept the pain of layoffs (or even downsizing through attrition), the authors favor “machine capabilities [as] a force-multiplier for human expertise” over “using machines to automate human capabilities and hence make human expertise superfluous.”

That sounds reasonable until you spend a moment thinking about it. Their argument makes me wonder whether any of the three authors has ever worked a manual-labor job or one that is mind-numbingly boring. If they had, as I have, they would know how hard that work is, even harder than writing a dissertation.

And they would know how much better off many of us grunts would be if there were fewer such jobs. A University of Chicago analysis found particularly low levels of job satisfaction among waiters, laborers, hand packers and packagers, cashiers, food preparers, customer service representatives, and butchers, as well as low levels of general happiness among garage attendants, pressing machine operators, kitchen workers, and amusement and recreation staff.

Other studies have found low levels of satisfaction in similar manual, service, and clerical occupations, such as hotel and motel desk staff, mail clerks, insurance policy processing clerks, telemarketers, and aircraft cargo handling personnel. We do not need technology merely to collaborate with workers in these jobs. We need technology to do more of this work, allowing workers to move into work that offers greater autonomy, meaning, and satisfaction.

To be clear, that is entirely possible. Governments and companies around the world are implementing advanced technologies and automation tools across many of the industries listed above to raise productivity and living standards. For example, Japan’s Henn-na Hotel chain was the first in the world to staff hotels primarily with robots, and some restaurants in South Korea use robots to deliver food to tables.

Here is one candidate for automation: taxi and rideshare driving. A recent study of taxi drivers in Cameroon found that just over 73 percent had experienced low back pain—the most prevalent musculoskeletal disorder globally—during the previous year and nearly 66 percent reported job dissatisfaction. I know, you are supposed to say those jobs have dignity and provide opportunity. Of course they do. All jobs have dignity, and elites should not look down their Ivy League noses at people who do “mundane” jobs. But that does not mean that driving eight hours a day is pleasurable. I, for one, don’t know how Uber, Lyft, and taxi drivers do it.

What about the opportunities automation could create? If governments stopped prohibiting taxi and rideshare companies from deploying autonomous vehicles more broadly, fares could fall and capital investment could rise. That would create greater demand for workers who build and maintain autonomous systems, while consumers could spend their savings from lower fares on other goods and services that employ people. A win-win, unless you work at MIT.

So, let’s end the nonsense that says the economy’s goal is to prevent existing workers from ever losing their jobs. And let’s stop treating “pro-worker” as the decisive test. What about pro-growth, pro-consumer, or pro-American? A nation that suppresses labor-displacing technology will not protect its workers; it will simply lose investment, production, and jobs to countries that do not.

But the report comes right out and says it: A robot that installs cable is “not pro-worker” because fewer tasks are done by workers and more by capital. The authors should just go full Luddite, get rid of shovels, and require workers to use spoons, to paraphrase Milton Friedman.

To no one’s surprise, their policy agenda, if followed, would be a disaster: subjecting tech firms to stricter antitrust scrutiny, reshaping the tax code to limit incentives to invest in automation, and imposing new legal restrictions and compensation requirements on companies that use worker expertise to train AI systems. You get the idea. Protect the proletariat from the greedy, robot-deploying capitalists.

Maybe life in the rarefied world of job security through academic tenure makes universal job security seem plausible. Tenure for the masses! Here’s an idea: Let’s abolish academic tenure and promote automation so we can be sure our kids will have a much better life than ours.