Economists make sense of the world by building models to capture the messy and pervasive reality of modern economies, but these models are intentionally simplified. The goal is to illustrate fundamental choices and tradeoffs that shape the economy. In the process, these models often help define what policymakers pay attention to.
As economists update their automation models, they are simultaneously changing the field’s understanding of what the technology does to workers and shifting the debate about how politicians and regulators should respond.
Also read:
Continues after advertising
Why do some new inventions appear to raise wages broadly—at least over time—while others leave large swaths of workers worse off? Over the past decade, economists have answered this question by distinguishing between technologies that create new types of work and those that merely automate jobs that already exist.
The journey toward these newer models began in the mid-2000s, when economists took advantage of richer data and began breaking down work into individual tasks. For example, a researcher’s job may include collecting data, performing analysis, and writing reports.
In the beginning, the three tasks are done by one person. But over time, technology can take over the task of data collection, leaving the researcher responsible for analysis and report writing.
Task-based models have allowed a more detailed look at the impact of technology on work and have helped to better explain rising inequality in the United States and much of the world.
Starting in the 1980s, digital technology began to take over tasks associated with middle-income jobs, such as accounting or administrative work.
It has made many high-skilled tasks — such as data analysis and reporting — more productive and better paid. But as middle-class workers were displaced, many migrated to lower-paying jobs—and the abundance of available workers often caused wages to fall in some of these already low-paying occupations.
Continues after advertising
From 1980 until the beginning of the 21st century, job growth bifurcated between high-paying knowledge jobs and low-paying service jobs.
The task-based view also clarified the importance of expertise—it matters which tasks computers take on. From the worker’s point of view, it is better for machines to take on repetitive, low-value work — as long as it is possible to continue using one’s own expertise to perform higher-value tasks.
A limitation of the task-based view, at least initially, was the assumption that the list of potential tasks was static. But as researchers cataloged the evolution of roles and requirements, they discovered how many people work in jobs that, until recently, didn’t exist.
Continues after advertising
In 1980, the Census Bureau added remotely piloted vehicle controllers to its list of occupations; in 2000, it added sommeliers. These examples highlight two related ways in which technology can create work.
In the first case, a new technology directly creates a new type of job that requires new skills. In the second, a wealthier society — full of computers and remotely piloted vehicles — means consumers can spend money on new extravagances, like the services of a sommelier.
This “new work” is key to understanding how technology affects the job market, according to some economists.
Continues after advertising
In their view, whether technology will be beneficial to workers depends on whether society invents new things at which they can excel — like piloting remote vehicles. If the economy is rapidly adding new occupations that use human skills, then it can absorb some of the displaced workers.
Daron Acemoglu of MIT and Pascual Restrepo of Boston University formalized this idea in 2018 in a model in which automation races against the creation of new tasks. New technologies displace workers and create new things for them to do; when commuting comes before new work, wages can fall.
As economists reformulated their theories, they also revised their recommendations. In the age of the race between education and technology, they often recommended that more people go to university or otherwise raise their qualifications.
Continues after advertising
Today, they tend to emphasize more the importance of creating new work and supporting policies and institutions.
Major technologies create entirely new categories of human activity. This means both new jobs and new demand as society becomes wealthier.
This is similar to an old management idea: “reengineering.” In 1990, Michael Hammer wrote a famous article in the Harvard Business Review urging managers to “stop paving cow trails.” Old processes should not just be automated, he argued, but reinvented from scratch.
The implication of Acemoglu and others’ “new task” models is similar. Instead of simply automating the tasks we currently perform, we should invent entirely new ways for AI to make our lives better — and new ways for humans to develop and use expertise.
The tasks AI takes on will depend, in part, on who is making the decisions — and how much leverage workers have. Unions have an ambivalent relationship with technology and are often skeptical of automation.
Here again, economists’ thinking has evolved. In the 1980s, the most prominent view was that companies with unionized employees had less incentive to invest in innovation and new technologies.
Since unions would ensure that workers received most of the benefits, the reasoning was that investors would have little incentive to spend on research and development.
But there are several other ways to think about this, says John Van Reenen, an economist at the London School of Economics.
Companies that make good use of new technologies generally pay more because they are more productive and profitable. Van Reenen says that in the right circumstances, unions can help ensure that workers have the power to claim some of those profits in the form of higher wages.
Worker participation — which unions often facilitate — can also guide companies toward more productive (and worker-friendly) uses of AI.
“There is an emerging view that bottom-up innovation will be the best way to discover the best uses of AI,” says Kinder. “So there is a business case for keeping employees in the loop.”
And worker participation can protect against “so-so technology.” The idea is that companies sometimes automate just enough to replace workers, but without creating big productivity gains.
Acemoglu uses the example of self-service checkouts: they work well enough to take work away from tellers, but not so well as to provide a big boost to the economy that could fuel demand in other sectors.
c.2026 Harvard Business Review. Distribuído pela New York Times Licensing
