Automation and the Future of Work. Aaron Benanav
rhetorics of collective self-sacrifice and anti-consumerist austerity.
Recurrent Fears
These futurist visions, arising from all points along the political spectrum, depend upon a shared prediction about the trajectory of technological change. If anything, the confidence that is characteristic of the automation discourse has only increased in the midst of the pandemic recession. Although technological change was not itself the cause of job loss—at least this time around—automation theorists argue that the spread of the pandemic will hasten the transition to a more automated future. Lost jobs will never return since, unlike their human counterparts, cooking, cleaning, recycling, grocery-bagging, and caretaking robots can neither catch COVID-19 nor transmit it to others.15 Have the automation theorists got this story right?
To answer this question, it is helpful to have a couple of working definitions. Automation may be distinguished from other forms of labor-saving technical innovation in that automation technologies fully substitute for human labor, rather than merely augmenting human productive capacities. With labor-augmenting technologies, a given job category will continue to exist, but each worker in that category will be more productive. For example, the addition of new machines to a car assembly line will make line work more efficient without abolishing line work as such; fewer line workers will be needed in total to produce any given number of automobiles. Whether such technical change results in job destruction depends on the relative speeds of productivity and output growth in the automotive industry: if output grows more slowly than labor productivity—a common case, as we will see below—then the number of jobs will decline. This is true even without automation entering the picture. By contrast, true automation takes place, as Kurt Vonnegut suggested in his novel Player Piano, whenever an entire “job classification has been eliminated. Poof.”16 No matter how much production increases, there will never be another telephone switchboard operator or hand manipulator of rolled steel. Here, machines have fully substituted for human labor.
Much of the debate around the future of workplace automation turns, unhelpfully, on an evaluation of the degree to which present or near-future technologies are labor-substituting or labor-augmenting in character. Distinguishing between these two types of technical change is more difficult than one might suppose. When a retailer installs four self-checkout machines, watched over and periodically adjusted by a single employee, has cashiering ended as an occupation, or is each cashier now operating three additional registers? Taking an extreme view on such issues, one famous study from the Oxford Martin School suggested that 47 percent of jobs in the United States are at high risk of automation; a more recent study from the Organisation for Economic Co-operation and Development (OECD) predicts that 14 percent of jobs are at high risk, with another 32 percent at risk of significant change in the way they are carried out—due to innovations that augment labor rather than substitute for it.17
In fact, both types of technical change can be expected to leave many workers without jobs. It is unclear, however, whether even the highest of these estimates suggests a qualitative break with the past has taken place. By one count, “57 per cent of the jobs workers did in the 1960s no longer exist today.”18 Alongside other forms of technical change, automation has been a persistent source of job loss over time. The question I address here is not whether new automation technologies will destroy additional jobs in the future (the answer is certainly yes). It is whether these technologies—advanced robotics, artificial intelligence, and machine learning—have so accelerated the rate of job destruction and so diminished the rate of new job creation that increasing numbers of people are already finding themselves permanently unemployed.
If so, that would completely upend the normal functioning of capitalist economies. This insight, on which the automation theory is based, was stated most succinctly by Nobel Prize–winning economist Wassily Leontief in 1983. The “effective operation of the automatic price mechanism,” he explained, “depends critically” on a peculiar feature of modern technology, namely that in spite of bringing about “an unprecedented rise in total output,” it nevertheless “strengthened the dominant role of human labour in most kinds of productive processes.”19 In other words, technology has made workers more productive without making work itself unnecessary. Since workers continue to earn wages, their demand for goods is effective. At any time, a technological breakthrough could destroy this fragile pin holding capitalist societies together. Artificial general intelligence, for example, might eliminate many occupations in a single stroke, rendering large quantities of labor unsalable at any price. At that point, information about the preferences of large sections of the population would vanish from the market, rendering it inoperable. Drawing on this insight—and adding that such a breakthrough now exists—automation theorists frequently argue that capitalism must be a transitory mode of production, which will give way to a new form of life that does not organize itself around wage work and monetary exchange.20
Automation may be a constant feature of capitalist societies; the same is not true of the theory of a coming age of automation, which extrapolates from instances of technological change to a broader account of social transformation. On the contrary, its recurrence in modern history has been periodic. Excitement about a coming age of automation can be traced back to at least the mid nineteenth century, with the publication of Charles Babbage’s On the Economy of Machinery and Manufactures in 1832, John Adolphus Etzler’s The Paradise within the Reach of All Men, without Labour in 1833, and Andrew Ure’s The Philosophy of Manufactures in 1835. These books presaged the imminent emergence of largely or fully automated factories, run with minimal or merely supervisory human labor. Their vision was a major influence on Marx, whose Capital argued that a complex world of interacting machines was in the process of displacing human labor from the center of economic life.21
Visions of automated factories appeared again in the 1930s, 1950s, and 1980s, before reemerging in the 2010s. Each time, they were accompanied or shortly followed by predictions of a coming age of “catastrophic unemployment and social breakdown,” which could be prevented only if society were reorganized.22 To point out the periodicity of this discourse is not to say that its accompanying social visions should be dismissed. For one thing, the technological breakthroughs presaged by the automation discourse could still be achieved at any time. Just because they were wrong in the past does not necessarily mean that they will always be wrong in the future. More than that, these visions of automation have clearly been generative in social terms: they point to certain utopian possibilities latent within capitalist societies. Indeed, some of the most visionary socialists of the twentieth century either were automation theorists or were inspired by them, including Herbert Marcuse, James Boggs, and André Gorz.
Taking its periodicity into account, automation theory may be described as a spontaneous discourse of capitalist societies that, for a mixture of structural and contingent reasons, reappears in those societies time and again as a way of thinking through their limits. What summons the automation discourse periodically into being is a deep anxiety about the functioning of the labor market: there are simply too few jobs for too many people. Why is the market unable to provide jobs for so many of the workers who need them? Proponents of the automation discourse explain this problem of a low demand for labor in terms of runaway technological change.23
Too Few Jobs
If the automation discourse appeals so widely again today, it is because the ascribed consequences of automation are all around us: global capitalism is failing to provide jobs for many of the people who need them. There has been, in other words, a persistently low demand for labor, one which is no longer adequately registered in unemployment statistics.24 Labor underdemand is reflected in higher spikes of unemployment during recessions, as in the 2020 pandemic recession, and in increasingly jobless recoveries, a phenomenon likely to be repeated in the pandemic recession’s aftermath.25 Low labor demand has been evident, as well, in a trend with more generic consequences for working people: a decline in the share of all income earned in a given year that is distributed as wages rather than profits.26 Mainstream economists long held the steadiness of the labor share to be a stylized fact of economic growth, which was supposed to ensure that the gains of economic development were widely distributed. In spite of massive accumulations