Why the Power of Technology Rarely Goes to the People

In a new book, economists Daron Acemoglu and Simon Johnson provide a sweeping historical overview of just how unevenly the spoils and costs of technological change have been distributed. Power and Progress: Our 1,000-Year Struggle Over Technology and Prosperity reminds us that technology is not itself a force but rather a tool that is developed to support the agendas of the people and institutions who hold power in society. Claiming a fair share of technology’s benefits for the rest of society — that is, for most of humanity — requires that that power be challenged. Acemoglu and Johnson chatted with features editor Kaushik Viswanath about what lessons the past holds for how we should develop and implement technology today and in the future. This conversation has been edited for length and clarity.

Kaushik Viswanath: What’s the central argument you’re making in Power and Progress, and what motivated you to write it?

Daron Acemoglu: This is a critical time to be thinking about the future of technology. A lot of decisions of great import are being hampered by the fact that there is “techno-optimism” in academia, the tech world, and the policy world. Techno-optimism is the notion that impressive technological change will automatically lead to better outcomes for society, especially for workers via the labor market, even if there are some transition costs.

Our understanding of the relevant economic theory and history has led us to believe this isn’t right. Throughout history, deliberate decisions have had a bearing on who gained and lost from a particular technology, whether it brought anything approaching shared prosperity, or even whether it helped or destroyed democracy. So our purpose in writing Power and Progress was to dispel the notion that in the history of technology, everything has always worked out OK. There are similar choices and struggles over technology today as we’ve had in the past.

One of the key concepts you discuss is the productivity bandwagon. What is this, and how does it create winners and losers whenever we have technological change?

Simon Johnson: The productivity bandwagon is the notion that when technology improves, you get higher wages, more opportunity, and better health, and everybody gains from it eventually. Our key problem with that notion is the “eventually.” “Eventually,” from the beginning of the Industrial Revolution, was 120 years. The 1720s to the 1840s saw a lot of new technology, but we know that in the 1840s, children as young as 6 were still pushing coal carts deep underground with their heads. Conditions improved for more people in the second half of the 19th century but as a result of a lot of effort, not through any kind of automatic economic or political process.

Acemoglu: The perspective that Simon and I bring to the British Industrial Revolution is that it was really a revolution of vision. A new class of ambitious people emerged who wanted to apply technology to improve how people control their environment and the production process. They weren’t doing it out of altruism; they were preoccupied with making money, wanted to rise within the British hierarchy, and didn’t have much sympathy for the people who were below them in that hierarchy, whether in Britain or the rest of the world.

This is an illustration of what ambition does unless it is countered by institutions and other groups that have alternative visions of how society should be organized. It also illustrates the weaknesses of the productivity bandwagon. People were left behind in the early phases of the Industrial Revolution for two reasons. First, most of the technology was used for automation, not increasing workers’ productivity contributions. When technology displaces workers, it doesn’t increase their contribution to production or create a powerful reason for employers to go out and pay higher wages to workers. Second, this was all embedded in an institutional setup, both because of the vision of the entrepreneurs and because trade unions were banned and heavily prosecuted, and Britain was very far from a democracy at the time.

The working class did not have any rights or protections. That’s why, even as many people made fabulous amounts of money, workers’ real incomes stagnated or even declined. Sharing the gains of technology required a complete change in the institutional fabric of British society, which the elites and upper middle classes resisted. It required a change in the direction of technology, too — for example, it was necessary to invest in urban infrastructure to improve sanitation and bring infectious diseases under control. Until then, urban life was horrible for working people.

Fast-forwarding to post-World War II in the U.S., you describe how this period saw a more equitable distribution of the productivity gains from technology. How did this happen?

Acemoglu: That episode illustrates how the factors that worked against shared prosperity during the Industrial Revolution were turned in favor of shared prosperity in the 20th century, especially in the decades that followed World War II.

Its origins can be traced to the American system of manufacturing, because this was a key part of a general effort to make unskilled labor more productive using machinery. That, in turn, was critical for less-skilled workers to earn a high and rising wage. In this period, workers’ contributions to the production process could be bolstered by training. This was facilitated by a combination of technologies that didn’t simply automate work but created new and more technical tasks, more maintenance tasks, and more advanced machining tasks for workers. And it was in the context of institutions that provided countervailing powers to the most powerful firms — in particular, a secure democracy by historical standards, a labor movement that had become much stronger after the New Deal and during World War II, and a supportive regulatory environment by the U.S. government that encouraged technological change but also brought limits to what the largest companies could do, for example, through antitrust enforcement.

You also write that the U.S. labor movement during this period actually encouraged the mechanization of the industries in which they worked. Why did they do this?

Johnson: The key was in their insistence that their workers get trained to use the machines. They realized that mechanization was coming whether they liked it or not. They couldn’t simply ask for higher wages, because that would lead to more automation. So [labor unions] asked for their workers to acquire the necessary skills and be compensated appropriately. Unions are much weaker today, so that kind of countervailing power is missing, which means the benefits of automation will go to whoever has social power — which means relatively few people.

Acemoglu: We are not against automation. Blocking automation would not just be infeasible, but to the extent that it’s tried, would be hugely costly. In its best moments, the labor movement, both in the United States and in Europe, encouraged the introduction of advanced automated machinery but at the same time negotiated the creation of better, more advanced tasks for workers to operate and inspect these machines. Where workers didn’t have those skills, employers would have to train them. So it was the combination of new tasks and training that unions advocated for. Today the question is, can we still encourage the right type of automation?

What is the role of business leaders in determining the direction of technological advancement and distributing its gains?

Acemoglu: The future of technology is inseparable from the vision of powerful actors. It’s not something we can all democratically vote on. The same is true of how CEOs decide to split profits between different stakeholders. Do they see labor as one of those stakeholders? That is a question that is entangled with the future of technology.

Over time, business leaders have shifted toward just serving the interests of the shareholders. Labor is viewed as troublesome and costly, so they try to eliminate it as much as possible. And that has synergized with the vision of the tech community to develop machines that can automate as much as possible.

But nothing in the laws of capitalism makes that necessary. During other periods, in other contexts, businesses have prioritized increasing worker productivity. They have found ways of rewarding their shareholders while giving raises to their workers when the company is doing well, and investing in technologies that increase worker productivity. A new vision among business leaders would be feasible and highly useful for the kinds of futures of work that we’re talking about. But that won’t emerge by itself. It will require pressure from institutions, civil society, and the media, as well as some amount of organized labor.

You describe how the doctrine of maximizing shareholder value became consensus in management schools and then management consultancies, ending an era of widely shared gains from technology. Do you see that changing?

Acemoglu: I have a paper with Alex He and Daniel le Maire where we find that CEOs with business degrees from the top MBA programs in the United States don’t increase productivity, exports, or investment, but they reduce wage growth and labor share. But the CEOs in our sample are all from the 1970s, ’80s, and ’90s. Today, the same schools have a somewhat different air. Students seem to care much more about broader aspects of business. Faculty don’t just talk about increasing shareholder value and creating lean corporations by eliminating labor. So I already sense some change in that direction. How effective it is, we don’t know yet.

Johnson: There’s a lot more progress to be made. The curriculum and the core ideas that are imprinted on students still lean a lot more toward Milton Friedman than toward Acemoglu-Johnson or any other view.

If you consider the pressure from financial markets and look at the language used by analysts, it reinforces that narrow view, which is, I think, not ultimately good for business.

Turning to the tech that’s on everyone’s minds these days: Where do you think AI — and generative AI, specifically — is headed?

Acemoglu: These are phenomenally interesting and impressive technologies. That only raises the stakes of getting the direction of this technology right and setting up the right regulatory structure.

But the two polar views that are most loudly heard in the media are both unhelpful: On one end are techno-optimists, who say, “Everybody will benefit. Yes, a few people might lose their jobs. But you’ll get more massage therapists, even if you don’t have enough white-collar workers.” On the other end is the view that killer robots are coming and we have to worry about existential risk.

Neither of these views addresses the right concerns. AI can do a lot to help workers and society. It could go along the lines of the platforms that Taiwan introduced, for example, to facilitate more democratic participation; those have worked reasonably well. Or it can go in the direction of automation that deepens inequalities, delivers more misinformation, disinformation, manipulation of users — what we’ve seen with social media, especially platforms like Facebook.

We really worry about that direction, and that’s where our leaders are asleep at the wheel. Society is not worrying enough about these things. There isn’t even the right set of aspirations that have been articulated about what we should want from this technology.

Johnson: I’ve heard the view that people are complaining now because it’s cognitive tasks that are being replaced by a machine, whereas before it was manual work. What we say in our book is that what’s actually vulnerable here are all routine cognitive tasks. Wendy’s, for example, has said it’s going to use chatbots to take orders at drive-throughs. They’ll still use humans to flip the burgers. Is your ordering of a burger going to be any better with this machine? Are they going to be paying the burger flipper any more money? No, they’re just doing this so they can have fewer workers.

We call it so-so automation. It’s a way to tilt power against the workers. You’re replacing people who are quirky and sometimes difficult to manage with machines that are designed for mediocrity. Where’s the productivity breakthrough? Where’s the big positive benefit?

Acemoglu: In productivity revolutions of the past, like at the Ford Motor Company, automation was critical, but only when combined with new products, new tasks, new ways of using machinery, new creativity. The Ford factory would not have done anything of note if it took exactly the cars that other companies were producing and made them with a bit more automation.

This is why we prefer to emphasize machine usefulness rather than machine intelligence. We should be using machines to make humans better. Generative AI is so promising because it has that capability. It could help with the retrieval and filtering of information so that human decision makers make better decisions. But that’s very different from automating a few more McDonald’s kiosks.