Reorganizing work, not just getting AI to making decisions: interview with Aaron Benanav

Aaron Benanav, a post-doctoral researcher at Humboldt Universität zu Berlin, published one of the best books of 2020: Automation and the Future of Work. In the book, Benanav argues that it is not the rise of robots and automation that are taking our jobs, but the worsening of economic stagnation and deindustrialization.

In an interview published in Notes from Below, Benanav says: “What I’m particularly interested in is how this hype around workplace automation has given rise to a broader social theory, which argues that the tendency towards automation is driving social conflicts in advanced capitalist societies and across the world. They say: look around you, workers are being replaced more and more by intelligent machines; in the future, most labor will be obsolete.”

See below our interview with Aaron Benanav. Some of the subjects: automation theorists, UBI, informality “non-standard employment” in Global South and North, gig economy, the unemployment and underemployment, algorithms and protocols.

DIGILABOUR: Your argument against automation theorists is very clear – and, for me, very consistent. But I would like you to detail your critique of left-wing authors.

AARON BENANAV: Before discussing my critique, I want to say that I find the utopianism of the left-wing automation theorists inspiring. They are trying to rescue the emancipatory project from its ruins, to dream and to fight for a better tomorrow in which (1) humanity is liberated from economic insecurity, (2) working becomes much less central to social life, and (3) people are free to pursue their passions as they please. Left-wing automation theorists do look to newly emergent technologies for both the explanation and the solution to capitalist societies’ unemployment issues. However, they are not naive about the ways that capitalism has shaped the development of technologies. That is, compared to automation theorists of the center and the right, those on the left are more critical of the way technologies are currently being deployed. They also see the dangers associated with neoliberal variants of UBI.

My criticism of the left-wing automation theorists is that, like the other automation theorists, they give a poor explanation for why capitalist economies are generating high levels of underemployment and inequality today. They identify these problems as evidence of that qualitative breakthrough in technology has taken place. In short, they believe that industrial machines and artificial intelligence are rendering human labor obsolete across more and more economic sectors. But there is no evidence of that this is happening from an economic perspective. Of course, technological innovations are being implemented in production all the time. But rates of labor productivity growth are slowing down, not speeding up. This is the exact opposite of what one would expect to see were there a breakthrough to a new age of automation. That’s why I say that the main problem in the economy today is not rapid job destruction due to the deployment of new automating technologies. The main problem is that jobs aren’t being created as quickly as they were in earlier eras because economies are growing at a slower and slower pace. Economies are stagnating, and that in itself retards the development of technologies, since it is associated with persistent under-investment.

This critique of the left-wing automation theorists regarding questions of causal explanations also has implications for how we think about questions of strategy. Freeing people from economic insecurity, making work less central to life, and encouraging everyone think of their lives in terms of the pursuit of passions should remain central goals of the emancipatory project. But automation technologies aren’t going to get us there. That means that we are going to have to find social rather than technological pathways to the post-scarcity world. We are going to have to think about not just eliminating work but also reorganizing it, and not just getting artificial intelligences to make decision for us, but figuring out how to organize our institutions so we can plan and coordinate our activity for ourselves.

DIGILABOUR: Do you think the adoption of UBI is a mistake?

BENANAV: Whether the adoption of UBI is a mistake or not is a strategic question that one can only evaluate in the context of specific, political conjunctures. My critique is aimed at the idea that UBI is the “silver bullet” that will resolve the fundamental problems of a low-labor-demand economy. Whenever I lay out my criticisms of UBI, I always start by explaining why I think UBI appeals to so many people. In many countries, including the USA and Brazil, welfare benefits are conditional or means-tested. States create punitive bureaucracies, which treat the poor with suspicion if not contempt: the poor have to prove that they are deserving of assistance (as if being poor weren’t hard enough). Given that this is how welfare states are organized, it makes sense to advocate for unconditional or universal benefits instead.

The idea that these universal benefits should be paid in cash also makes sense, at least at first, insofar as it implies a greater freedom for the poor to manage their own affairs. But should we really rely on the market to meet our most basic needs? I find this position untenable. Markets will always underproduce collective goods, such as education and healthcare, and overproduce collective “bads,” such as pollution and traffic. Turning to markets to meet our needs also subjects our economic and social security to the decisions of highly undemocratic, profit-seeking firms. That, I think, turns out to be a gigantic mistake. Especially in periods of crisis, the investment decisions of private corporations would determine what a UBI was worth (by determining which of the things we need are produced and in what quantities).

Automation theorists turn to UBI for yet another reason. Their argument is that technologies like machine learning, artificial intelligence, and advanced robotics are creating a world of abundance, in which goods and services are produced in large volumes without the need for human labor. If that were true, then the problem we would face would mainly be one of distribution: workers would no longer be able to sell their labor, so they would lack the means to purchase society’s ever more abundant goods and services. A UBI would fix that. But this is a misdiagnosis. Economic growth rates aren’t speeding up, due to high rates of productivity growth. On the contrary, economic growth rates are slowing down, likewise productivity growth rates. In ever more stagnant economies, UBI programs would face the same issue as other welfare programs, especially given how expensive a truly liberating UBI program would be. Firms would blame UBI for workers’ low productivity, and would put a lot of pressure on governments to keep benefit levels low, in order to “improve business morale” and speed up economic growth rates. Instead of pushing for higher UBI levels, advocates would have to defend their programs against austerity and budget cuts.

The problem we face is today not one of distribution but mainly of production. We have to reorganize production, wresting away control over investment decisions from the tiny class of wealthy investors, reorganizing and redistributing work, and guaranteeing people access to what they need. We should do that partly by way of a basic income, but mostly by providing people with non-market access to goods and services, produced according to plan.

DIGILABOUR: You comment, on the one hand, OECD employment protection index in which Brazilian permanent workers on standard employment contracts have had stronger employment protections than equivalent workers in the UK. On the other hand, you point to “non-standard employment” as a misnomer. And the very history of the Brazilian economy – like most countries in the world – is based on informality and gig as the norm of the working class. How significant is that? How do you position Brazil in your argument?

BENANAV: That requires a clarification. I say “non-standard” work is a misnomer to mark out the evident conceptual and empirical contradictions at the heart of this concept. How could this work be “non-standard” if it accounts for the majority of work worldwide? Informality is more common than formality. What is surprising and requires explanation is how formal employment with strong protections was ever seen as a norm towards which capitalist economies were trending. In reality, “standard” employment became a norm only in wealthy countries—which account for a small portion of the world’s workforce—and only for a brief period of time. Even in those countries, it was never fully generalized across the workforce. Large sections of the population, particularly women, remained precariously employed even at the height of the postwar boom. Still, the fact that these groups were not fully included in the standard-employment regime negates neither that regime’s past existence nor the way that the division between secure and insecure forms of work continues to shape labor markets.

After WWII—in response to both workers’ demands and the threat of Soviet expansion—governments in advanced capitalist countries extended welfare state benefits and offered new legal protections for workers. Those protections were copied by many post-colonial states in the era of decolonization and development. At the time, it seemed like it would be possible to generate worldwide full employment based on Keynesian principles of macroeconomic management, which also informed development economics. The goal of global full employment was enshrined in the Article 55 of the Charter of the United Nations, but it never came close to approaching realization. Getting to that goal would have required that countries all over the world sustain rapid rates of economic growth and job creation for a long time. Those conditions were not sustained—or even temporarily achieved in many countries.

Indeed, international observers were already noting the lack of job creation and the expansion of urban informal settlements and casual labor in Brazil in the 1950s and 60s. Those observers argued that this problem was specific to the developing world, having to do with high levels of inequality and poorly chosen development strategies. Many in Latin America attributed it, instead, to patterns of international trade, which disadvantaged producers of primary commodities. In any case, in the 1950s and 60s, this issue seemed to be confined to the global South. Countries in the global North were still experiencing full employment.  Yet by the 1970s, the same problems of low labor demand were spreading to the global North.

Rates of unemployment rose and remained high. With the neoliberal turn of the 1980s, governments everywhere began to respond punitively to workers who had trouble finding stable jobs: states reduced workers’ legal and welfare-state protections and pushed them to take whatever jobs were available. Under conditions of worsening economic stagnation, that meant workers had to take jobs that underemployed them in terms of wages, hours of work, or the use of their skills. Not finding anyone willing to hire them, many workers made jobs for themselves in the informal sector, selling directly to the market. These trends had already been unfolding across global South countries for a long time (since at least the 1920s); but starting in the 1980s, governments everywhere began to encouraged these trends. The shift towards neoliberal policy—towards what they called “flexible labor market policies”—was accompanied by high or rising levels of inequality, particularly in the global South, where states rescinded labor protections as part of IMF-led structural adjustment programs.

Of course, there are still workers in all countries who have strong job protections. These workers are relatively more protected from the consequences of low labor demand. They are sometimes able to win wage increases even when unemployment and underemployment are high for everyone else. By contrast, workers who lack such protections are subject to the ups and downs of the labor market to a much greater degree. When labor demand is weak—which is the normal condition these days—precarious workers see their bargaining positions weaken significantly vis-a-vis employers. They cannot demand higher wages or better working conditions; they have to give up on any hope for autonomy at work. The gig economy does not create these conditions. It exploits them, as do many other, traditional businesses. To understand the gig economy better, we should to compare gig work to other forms of insecure work, rather than to secure jobs that most workers have never had.

Workers often end up in the gig economy because it appears to offer something better than the other insecure jobs: compared to micro-management by vindictive human beings, management by algorithms has a certain allure. In many non-gig based jobs, workers are forced to work on “flexible schedules” at firms that save money by remaining understaffed. Workers have to come to work on short notice, or stay later than they expected. Instead of being laid off, they have their hours cut to almost zero, so they are forced to quit and receive no unemployment benefits. In the gig economy, workers are told that they can set their own schedules. These workers only learn how insidious algorithmic management is after they have been working in the gig economy for a while and have invested themselves in it.

DIGILABOUR: Other authors such as Antonio Casilli (on taskification of labor) and Janine Berg (from ILO, on job quality) have also argued about persistent underemployment. How does your diagnosis differ, for example, from the job quality approach?

BENANAV: One of the main issues faced by labor statisticians today is that unemployment doesn’t work very well as a category for analyzing labor market conditions, that is, the supply and demand for labor. Facing poor prospects, many workers either stop looking for waged work, or take poor quality jobs while looking for something better. In response, labor statisticians have tried to develop a range of other categories—informal work, non-standard work, vulnerable work, discouraged workers—to capture the true extent of the labor surplus. Particularly since the 1970s, when the category of informality emerged, one common strategy has been to try to identify particular qualities of bad jobs and then to count the number of jobs with these qualities. How many people are in involuntary part-time work? Or in temporary jobs? Or lack formal contracts? Or work in micro-enterprises of five of fewer workers? This statistical work is very important because we need to publicize the true extent of labor insecurity today.

A major issue is that no single, summary statistic has ever been found that could really replace the unemployment rate. There isn’t even one standard way to report “non-standard” work. It is usually divided up into a range of subcategories. Increasingly, we have been forced to recognize that unemployment is not a self-evident economic experience. It is shaped by policy. Implementing labor protections, health and safety regulations, minimum wages, and unemployment insurance programs was a way for governments to standardize work. Those who needed jobs were thereby encouraged to reveal themselves to the state, as the unemployed, so the government could count them and manage their affairs. In countries that never implemented the full range of welfare and labor market policies, like Brazil, it was impossible to produce a reliable unemployment rate. Too many workers continued to work on their own account, or as unpaid family labor, in what was later called the “informal sector” of the economy. These workers experienced a lack of work as a loss of income rather than as the loss of a job and hence could not be counted on the terms of unemployment statistics.

When the state gave up on the project of standardizing labor market experiences—in the face of falling rates of economic growth and rising economic insecurity—experiences of worklessness increasingly shifted. Statisticians increasingly recognized that, instead of one primary form of worklessness called “unemployment,” a variety of different forms of “underemployment” were proliferating across society. Statisticians tried to collect information about these different forms, but it is difficult to summarize them neatly.

The issue is that collecting statistics in this way can sometimes give one the wrong impression—that improving workers’ lives is simply a matter of improving job quality, piece by piece, and hence of shifting workers from “bad jobs” into “good jobs” over time. This perspective is, I believe, too static. Proponents of the Latin American “underground approach” already said as much in the 1980s. Alejandro Portes and his colleagues were correct to criticize two-sector models, like that of the ILO, which made it seem as if fixing the problem was simply a matter of formalizing the informal economy. In reality, the informal and formal sectors are dynamically interrelated in ways that depend on the specific legal-institutional protections operative in each country. The ILO was forced to recognize this in the 1990s, when changes in labor law made it easier for formal enterprises in many countries to hire workers without formal contracts. The ILO then expanded the concept of the “informal sector” to include a variety of forms of so-called “informal employment.”

Facing a persistently low demand for their labor, workers do whatever they can to survive. At the same time, businesses are always looking for new ways to exploit insecure workers. When these workers win labor protections, businesses find ways to get around them. Since politicians need to try to find ways to improve business confidence, to get private investment going again, they are often willing to pass laws making it easier to for businesses to exploit workers’ insecurity. Or else governments ignore their own regulations, turning a blind eye to unsavory business practices. Once governments become complicit in making workers more insecure, the project of measuring underemployment becomes much more difficult.

My use of the concept of underemployment is, for all these reasons, not a precise statistical category. For me, underemployment is a theoretical category that is meant to capture a dynamic process. We cannot measure the extent of underemployment so easily. I do not believe it is equivalent to the number of workers who have part-time work but seek full time work. Defining it in that way, as the ILO now does, is to admit defeat with respect to statistical construction. In fact, many of the ways in which workers find themselves underemployed are not captured by this statistic (or by any other). For example, workers with college degrees who work in jobs that do not require more than a high school degree are not counted as underemployed. This “educated underemployment” makes up a large portion of underemployment in many countries, especially in global South regions such as North Africa.

In the end, I argue, the best measure of the degree to which workers are underemployed is the trend in the labor share of income. We should look at gaps between average productivity growth, average real wage growth, and median real wage growth. This measure captures the effects, rather than topology, of labor underdemand. Wherever workers’ incomes fail to rise in line with labor productivity, that is an indication that they have weak bargaining power. They are not only underpaid; they are likely experiencing deteriorating working conditions.

Of course, we should still fight to improve job quality for insecure workers in whatever ways we can. But in the end, improving workers’ conditions will depend more on transforming the broader economic and political environment than on lobbying to improve labor laws. Given the worsening stagnation of economies, making workers’ lives better will require that we wrest control over investment decisions away from the small class of private investors. Investment decisions needs to be socialized, made democratic, disconnected from profitability, and used to improve work conditions as much as labor productivity.

DIGILABOUR: You say that people can design algorithms and protocols from digital technologies as a possibility to the current scenario. What projects do you understand are interesting?

BENANAV: Businesses today are using digital technologies to supervise workers in sneaky ways, to manage them at a distance, and to break apart production processes and distribute them across the world, as parts of complex global supply chains. What would it mean to use these technologies to benefit workers themselves, in a really free and emancipated world? Left-wing automation theorists talk about how to use new digital technologies to make work obsolete, by automating it. Some theorists also talk about using digital tools to plan the economy, algorithmically, with supercomputers solving vast systems of equations.

That is the wrong approach. Of course, we should use new technologies to get rid of work we don’t want to do, where possible (it may be much less possible to do that than people think, with our current technical capacities). We should also use digital technologies to help us make decisions—for example, about the optimal location to build a battery warehouse to store renewable energy. The point is that human beings should be the ones making the decisions about these issues. We need to think more creatively about how digital technologies can help us coordinate our activities in new ways, without the need for endless meetings. I call the structured forms of communication that digital technologies make possible “protocols” and contrast these to algorithms, which are (complex) automated calculations.

I am particularly interested in how digital protocols might make it possible for workers to coordinate across many different workplaces, to solve the “socialist calculation” problem without recourse to markets or centrally planning. But here I will give you a simpler example. In the Netherlands, there is a firm called Buurtzorg, or Neighborhood Care, which provides home healthcare services to people who cannot fully care for themselves. The nurses who work at Buurtzorg have no human managers, but unlike Uber drivers, they are not managed by algorithms either. Instead, they work in small, self-managed teams, taking care of their own affairs and organizing their work together within a given geographic area. The nurses use a specially developed intranet to ask advice of other teams of nurses, for example, to find an expert among the nurses who knows how to deal with a specific disease.

Again, if we think creatively, we can find many more ways to use these technologies to facilitate coordination among autonomous groups of workers. But these efforts will face strong limits as long as society remains capitalist. Most firms will refuse to implement technologies that make it easier for workers to coordinate with each other. Owners and managers fear that workers will be empowered to demand higher wages. Generally speaking, firms refuse to invest in the development of such technologies; they are never made. Instead, firms focus on finding ways to use technologies to separate and alienate workers from one another, to make it easier to exploit them. We won’t really be able to change that firm logic until we change society more broadly. But we should nonetheless think about how to use digital technologies for emancipatory ends and begin to develop the protocols of the future.

DIGILABOUR: I really like your final phrase and I have to do the hardest question: what are the first steps to build a massive social struggle?

BENANAV: Unfortunately, there is no way to call massive social struggles into being. The truth is that they are not built, but rather swell. People get swept up in them. Still, we live in an era in which such struggles are occurring more frequently—and in more countries. There are more and more social struggles taking place. We can prepare for them, and when they occur, we can immerse ourselves in them, trying to push them in broadly emancipatory directions. Doing so is especially important today, since so many movements seem to be inchoate in their beginnings, turning either to the right or the left over time. Much has already been written on these questions. I don’t feel I have much to add myself. What I will say is that I do not see how we will be able to really transform the economy and society without those struggles getting larger. In my book, I have simply tried to explain why I think these struggles are growing—in a sense to justify their occurrence. Right now, I am focused on doing what I can to help clarify the goals towards which I believe such struggles should aim.