"Machines…quell the revolt of specialized labor."
In my previous post, I wrote about alternative ways of viewing the encroaching effects of automation on employment. I suggested that, instead of viewing it as a zero-sum game, with industry hell-bent on automating everyone's jobs out of existence, that it is rather a phenomenon driven by firms' needs to maintain profitability and market share. In this sense, automation – and technology more generally – is an optimization function, but only in a ‘local' sense. The character of employment required by a firm is only commensurate to the needs that it can foresee in the near future. So for all the talk of a ‘post-work' future, we won't get there any time soon.
Nevertheless, this leaves open an important succeeding question: What does the technological substitution of labor actually look like, and what, if anything, can be done about it? The first thing that ought to be made clear is that the process of substitution is neither neat nor obvious. Introducing a single robot into the workplace does not necessarily displace a single human being. Indeed, in the case of industrial manufacturing, it may be more: a factory making cell phone parts in Dongguan, China, recently automated much of its operations and saw its headcount plummet from 650 to 60 workers. In a further blow against humanity, the output of the factory increased nearly threefold, and product defect rates declined from 25% to less than 5%.
It's worth noting that a factory making cellphone parts is an ideal subject for automation. A fully automated factory floor is the final reductio that, one might argue, began with Adam Smith's exposition of the power of the division of labor. But regardless of the factory's output – whether it's Smith's pins or components for mobiles – the fact is that we are making the same thing, thousands of times over. However, while significant, this kind of specialized manufacturing is but a fraction of global economic output.
As one leaves the carefully controlled confines of a plant, technological substitution becomes less effective. Consider the phenomenon of driverless cars, another favorite bogeyman of automation's Cassandras. To continue with the example of the above factory, let's look at the problem of distribution. The firm may opt to replace its drivers with autonomous vehicles, but at the moment there is an enormous difference between designing a self-driving car that will handle the predictability of long stretches of open road, versus the intricacies of city driving, where potholes, unpredictable pedestrians and other phenomena create much riskier scenarios. And in the case of firms whose very business model is distribution (such as UPS), a human being is still needed to perform the final handoff of the package to its recipient.
These fairly obvious remarks nevertheless intend to illustrate a larger point: automation is never not human-assisted. The question then becomes what proportion of a job is automated: as in the case of the truck driver, a journey may be 80% long haul, which is handled by the automated system, but the remaining 20% of navigating urban areas, or delivering the product to its recipient, is still the driver's ‘job'. For the firm, this is a decidedly awkward position. Whereas the factory is an ideal type – automate everything, fire 90% of your staff, and watch productivity and quality soar – interaction with the supply chain, or customers, or just the world itself, still requires people, and people expect to be paid.
Interestingly, this process is not just manifest in the stubbornly physical world, but also services that we might at first blush consider to be the exclusive domain of code. I am thinking about such things as chatbots, content monitoring of social media, and training of artificial intelligences on data sets of one sort or another. Writing in the Harvard Business Review, Mary Gray and Siddharth Suri note that
The truth is, AI is as "fully-automated" as the Great and Powerful Oz was in that famous scene from the classic film, where Dorothy and friends realize that the great wizard is simply a man manically pulling levers from behind a curtain. This blend of AI and humans, who follow through when the AI falls short, isn't going away anytime soon.
Shreeharsh Kelkar puts it another way: technology in the workplace is not apart from labor, and the interaction between labor and technology should be seen as
…an assemblage that embodies a reconfigured version of human-machine relations where humans are constructed, through digital interfaces, as flexible inputs and/or supervisors of software programs that in turn perform a wide-variety of small-bore high-intensity computational tasks (involving primarily the processing of large amounts of data and computing statistical similarities). It is this reconfigured assemblage that promises to change our workplaces, rather than any specific technological advance. The [research] agenda has been to concentrate on the human labor that makes this assemblage function, and to argue that it is precisely the invisibility of this labor that allows the technology to seem autonomous.
Crucially, the fact is that this incomplete technological substitution is furthermore a dynamic process, and one that sees employees' contributions as ever-receding, where work is never stable, but rather occupying a margin that is not unlike piecework. If your job is really about "doing the things that automation can't do…yet" then all sorts of other things break down. The idea of mastery of a profession is eroded, and the prospects of a stable career are diminished. Riffing off of Marx, the laborer is not only alienated from the product of their labor, but is further alienated by the processes of capital that allow ever less consequential input into the creation of that product. On a larger scale, we can speculate that the ever-elaborating infusion of technology into what were human-only tasks leads to, as Gray and Suri put it, "the rapid creation and destruction of temporary labor markets for new types of humans-in-the-loop tasks."
The speed of this creation and destruction is a defining feature of the current situation, and questions the effectiveness of older, established programs that were designed to aid worker retraining, such as the Trade Adjustment Assistance Program, first funded in 1974. When one takes into account that, according to more than 90% of jobs created between 2005 and 2015 were contract gigs (ie, not full-time), "Retraining for what?" becomes a legitimate question. Obviously, these part-time arrangements do not all belong to the ‘human-machine assemblage' postulated above, but at the same time this does not lend comfort to the recevied wisdom that technology will continue to create new opportunities for labor that are equal to or better than what came before. In fact, that position has come increasingly under attack.
Just as technology cannot be viewed as monolithic, neither can the workforce. It's worth asking who will bear the brunt of these changes, and what recourse there might be. In software there is the concept of LIFO, or ‘last in, first out', used to describe the order in which items can be added to and removed from a data structure. This idea may be applied just as easily to the workforce – those who are the most recent arrivals tend to have the most tenuous hold. A recent piece in Foreign Policy speculated on the implications of automation on women, who entered the workforce substantially only during the mid-20th century:
Women are projected to take the biggest hits to jobs in the near future, according to a World Economic Forum report predicting that 5.1 million positions worldwide will be lost by 2020… Men will see nearly 4 million job losses and 1.4 million gains (approximately one new job created for every three lost). In comparison, women will face 3 million job losses and only 0.55 million gains (more than five jobs lost for every one gained).
One could make similar arguments for other segments of the labor force that have faced structural challenges, for example, minorities, immigrants and those possessing only a high school education. Moreover, the rapidity with which temporary labor markets will continue to evolve privileges the agile, who can not only adapt to new work, but can also physically relocate to new markets. Unfortunately, there is strong evidence that labor mobility has been declining across the United States since the 1970s. (This is yet another strong argument for universal healthcare.)
As far as recourse goes, labor is vastly ill-prepared. As Brishen Rogers noted in a thoughtful review of automation, unemployment and the prospects for universal basic income written for the Boston Review,
Our labor and employment laws still envision the economy of the 1930s, which was dominated by massive industrial firms with hundreds of thousands of direct employees. Those laws rarely touch modern "fissured" work relationships such as Uber's relationship with its drivers, Walmart's relationship with its suppliers' workers, or McDonald's relationship with its franchisees' workers. Those laws also limit workers' ability to unionize or bargain effectively since they encourage bargaining at the firm or even plant level whereas today's modal workplace is growing ever smaller. Workers have fewer and fewer means to exert power on their own behalf.
In fact, this idea of power, or rather powerlessness, is perhaps the single greatest indicator of the difficulties in store for the labor force. With unions in disarray, the technologically-driven deskilling of the workforce continues apace. For example, a New York Times profile of Travis Kalanick, the CEO of ridesharing service Uber, noted that "roughly a quarter of [Uber's] drivers turn over on average every three months. According to an internal slide deck on driver income levels…Uber considered Lyft and McDonald's its main competition for attracting new drivers." Uber's technology platform is so easy to use (and its recruitment process so tolerant) that, rudely put, if you can flip burgers you can also be an Uber driver, and vice versa.
The point about technology as a form of power over labor cannot be overstated. This is, in fact, its primary consequence. As Brishen Rogers notes, "technology is not a substitute for menial labor in this story but rather one among many tools to keep labor costs down by exerting power over workers." In order to effectively interface with encroaching automation, it is necessary that human interactions with technology be measured, evaluated, stored and recalled whenever needed. Thus the Guardian observes that
In the logistics sector, companies are using technology not to replace warehouse staff and couriers, but to put them under increasing surveillance to control their working patterns, reducing employee autonomy, skill and dignity. Wrist-based technology allows bosses to monitor activity minute-by-minute, including bathroom breaks.
If labor is to formulate new and effective means of dissenting from the emergent status quo, it is here that the battle must be met. The fully automated factory, while most plainly visible, is nevertheless a red herring compared to the myriad ways in which labor has already submitted to a status that grows ever more fraught with contingency. When he wrote "Machines were the weapon employed by the capitalists to quell the revolt of specialized labor," Marx was not thinking of the wristbands that Amazon workers wear, but I can't imagine he would have been that surprised, either.