I was recently quoted in an article in MUNCHIES on the incoming US Secretary of Labor, Andrew Puzder. Puzder has (notoriously) argued that government regulations (to protect the welfare and rights of workers) drive the cost of labor up, which forces employers to automate their businesses by replacing human workers with machines. He has quipped of machines, “They’re always polite, they always upsell, they never take a vacation, they never show up late, there’s never a slip-and-fall, or an age, sex, or race discrimination case.”
In the article, I responded by pointing out how “Puzder is implying that deregulation will slow or halt automation by keeping labor cheap. Notably, the only two options Puzder presents for workers are (1) low-paying, uncertain, and exploitative employment or (2) no employment.”
This damned-if-you-do, damned-if-you-don’t analysis of mechanization, automation, and robotization has plagued workers facing technological developments since the industrial revolution. And it is still a major problem today. Especially with the increasing prevalence of Big Data and advances in artificial intelligence, we’re looking at a future in which not just manual jobs—such as picking heads of lettuce or riveting widgets—but also desk jobs like legal work will be in danger of replacement by computerized machines. President Obama even went so far as to warn of the dangers of automation in his farewell address: “The next wave of economic dislocations won’t come from overseas,” he said. “It will come from the relentless pace of automation that makes a lot of good, middle-class jobs obsolete.” The question is, of course, what can we do about it? Are we doomed to be stuck in Puzder’s de-humanizing catch-22 scenario?
It just so happens that this is a problem I’ve recently been researching. Together with Prof. Alastair Iles and several undergraduate student researchers, I’ve been examining cases of machine labor replacing, or threatening to replace, human labor in agriculture. In particular, we’re looking at the rise of mechanical harvesters for vegetable, fruit and nut crops during the post-War years. While we anticipate the research to continue for several more years, a few important points on automation are already clear. I’ll outline them in the rest of this blog post.
1. Machine labor is not the same as human labor
Historically, machines have outperformed humans when it comes to performing the same exact task over and over again. This is an area in which people do not excel. As I alluded to in my comment on Puzder, trying to compete with machines dehumanizes human workers. Americans have a national fable about a heroic individual, John Henry, out-working a steam-powered drilling machine. However, the feat costs John Henry his life, whereas the machine presumably went on to a long “career” driving steel for the railroad magnates.
The problem highlighted in the fable is the industrial standardization of production. Machines only “work” in a standardized, scripted environment; humans, by contrast, are flexible and adaptive workers, capable of performing many complex tasks depending on the situation. One deception of the catch-22 is that people and machines must inevitably compete with one another in a zero-sum game, even though the actual labor is very different.
2. Machines change the nature of production and the nature of what is produced
It’s not just that machines are “naturally” better suited to perform many tasks than are people, but rather that machines actually shape the nature of work, the working environment, and the product itself to better suit their standard requirements.
To take an example from our research, when California lettuce farmers in the 1960s were considering switching to mechanical harvesters, they ran into a problem: their fields and their plants were too variable and heterogeneous for the machines. Whereas human workers could differentiate between healthy, mature heads of lettuce and damaged or immature plants, a machine could not. Human harvesters likewise could handle humps and dips or other quirks of the field and furrows, but a machine needed perfectly level, uniform, and even rows to cut the heads.
In an effort to accommodate the envisioned mechanical harvesters, farmers, breeders, farm advisors, and agronomists set out to craft a new form of farming, what they called “precision agriculture.” The idea would be to create a perfectly level field of perfectly uniform rows populated by perfectly uniform crops that would all mature at precisely the same time. There would be no weeds, no irregularities, nothing to disrupt the machine’s ability to do exactly the same task over and over again. The field was made to suit the machine.
3. Mechanization is not automatic
New technologies are always embedded in an existing matrix of other technologies, behaviors, and organizational structures—what might better be referred to as a technological system. For a technological development to be truly transformative, to propagate and redefine a given production process, the rest of the system has to transform as well.
In the case of lettuce, the mechanical harvester was accompanied by a host of other technological, behavioral, and organizational changes. Breeders had to develop new seeds that farmers could use to shift toward a “plant-to-stand”—one seed, one plant—approach that would remove the entire production stage of thinning, that is, removing the unhealthy plants from the stand; a highly selective and irregular task not conducive to machine labor. At the same time, chemical herbicides were ushered in to eliminate the need for weeding, a form of chemically automating another form of selective labor. Lastly, farmers had to adapt to a once-over harvest. Whereas harvest crews comprising human laborers with hand knives could sweep a field multiple times over the course of several days to harvest lettuce that matured at different rates, a machine can only really harvest a field once, cutting most everything in its path in one pass.
The point is that mechanization is never “automatic”—ironically, it takes a lot of work to accommodate land, resources, businesses, and even consumers to the strict requirements of machine labor. Automation is a process that must be designed, guided, and even coerced into being. Importantly, this means that people determine how mechanization happens, and we have the power to do it differently.
4. A job is not an end in itself
In an era of rapid advances in artificial intelligence, machine learning, robotics, and other technologies that can perform human tasks, the fixation on “jobs” as the only metric of well-being is problematic. A lot of those jobs will no longer require a human to perform them, as McKinsey & Company’s new report on automation argues. The question we should be asking ourselves as a society is, do Americans really need to be doing tedious, repetitive tasks that can be left to a robot?
As long as life, liberty, and the pursuit of happiness—plus housing, food, healthcare, and so forth—is tied to whether a person has a “job” or not, the answer has to be “yes”. And as more and more machines come online that can perform human tasks, the more and more human workers will have to compete against the robot, hypothetical or real, that threatens to take their job. This can only drive the value of labor down, meaning that Americans will continually be asked to work more for less. That’s not a sustainable trend, and will only exacerbate our already high levels of socioeconomic inequality.
The United States is desperately in need of new public policies that can deal with this fundamental trend of working more for less. Basically, we need ways to ensure that the productivity gains generated by new technologies improve the lives of all Americans, not just the small percentage who happen to own a business and can afford the capital investment of the latest robots. Trying to preserve jobs as the sole route for people to improve their lives without changing the underlying pattern I’ve described above is a downward spiral that will harm many and benefit a few.
5. The problem with automation is one of distribution of wealth
Which leads to my penultimate point in this post. The problem is not that new technologies replace menial or repetitive jobs. It’s that they supplant peoples’ livelihoods, a concept which we should think of as a way of life rather than merely an occupation. That technologies can destroy people’s ways of life has been understood for centuries—it’s what spurred the Luddite movement during the early 19th century in England and also underpinned Karl Marx’s critique of capitalism some fifty years later. In many ways, communism was based on the idea that technological development is a good thing if the productivity gains are shared equally among everyone. While the practical implementation of that idea turned out to be oversimplified, the basic point—that new technologies raise questions about the distribution of wealth as much as they do about productivity gains—still applies today.
As farm worker movements in the 1960s and 1970s attested, the debate cannot focus on the simple ratio of input/output efficiency or even on jobs saved versus jobs lost. Those farm workers protested against mechanized harvesting and processing technologies in agriculture because they realized that all of the productivity gains would go to the farm owners, the patent holders, and the manufacturing companies. None would go to the people who had done the sweaty, back-breaking work in the fields that underpinned the entire agricultural industry and put the owners in the position to consider mechanization in the first place.
This is one of the reasons why people are starting to talk about a universal basic income, an idea which Robert Reich (himself a former Secretary of Labor under Bill Clinton) nicely outlined in a short video last fall. Basically, the idea is that if machines can do all the work to meet peoples’ needs, then people will not have jobs, and therefore will not have the money to buy the basic goods those machines produce. The universal basic income circumvents the problem by separating purchasing power from jobs, which has the effect of more equally distributing the productivity gains of automation such that people will actually be able to enjoy their lives.
6. Democratizing technology
I see the universal basic income as a solution to a symptom, the maldistribution of wealth. What we should be thinking about, if we want to address root-causes, is the maldistribution of power.
I had the opportunity to visit Immokalee, Florida a couple of years ago. Immokalee grows a large share of the nation’s fresh tomatoes, a delicate crop that is generally harvested by hand because machines tend to destroy or damage the fruit. Immokalee is also the location of some of the most recent cases of modern-day slavery—workers held in captivity and forced to work the fields. But it’s also home to an empowered workers movement, the Coalition of Immokalee Workers, which has fought for, and won, improvements in pay and working conditions for tomato harvesters.
On the visit, one Coalition spokesperson demonstrated how field workers harvest tomatoes. They walk through the fields filling up large buckets, which weigh around thirty pounds when full. It was a big deal for the Coalition when they bargained a penny-per-pound increase in the piece-rate compensation for pickers. As the spokesperson explained, the Coalition also achieved success in setting a new bucket-filling standard, under which workers would no longer be forced to fill their buckets over the rim, a practice that used to be required, but for which workers were not paid extra (see p. 32 of the 2015 Annual Report for a picture).
As the presentation continued, I was struck by the impressive intensity of the struggle around field workers’ rights, but also by the fact that the bucket itself was never questioned. Ergonomically, the buckets are terrible. They are plain plastic, with no handles or straps to ease lifting them onto the shoulder or help keep the bucket stable and distribute its weight once loaded. Why, I wondered at the time, does all the negotiating around the tomato picking seem to take these back-breaking buckets for granted? Is there no way to design a better tool for collecting the tomatoes and hauling them out of the field?
Of course there is, but that’s not how we tend to think of technological development. Which is why I argue that it’s not just about where the money goes after the fact, but about who has a say in how technologies are developed before the fact. The direction of technological development is open and changeable: technological development can be democratized. Spreading out the power to drive technological development will be the route to designing machines first to improve the conditions of labor and ways of life, and only second to increase productivity.
Our ongoing research into the history agricultural mechanization is motivated by a desire to understand how and why power over technological development was consolidated in the first place, in order that we might understand how to spread that power out moving forward.