Abstract
In this short and engaging book, Luke Munn challenges the view in the automation literature that the technological displacement of human labor is a universal, inevitable phenomenon. Recent improvements in artificial intelligence and robotics have induced labor economists to produce forecasts about future employment effects of automation and infer the power of new labor-displacing technologies from the logic of Moore’s law. Moore’s law states that the number of transistors on microchips would double every year. Presumably, the compounding effects of improving algorithms would allow for the ever-quicker replacement of workers, slowly at the beginning but much faster in later periods. Munn argues that these economists’ predictions of fast and universal automation are just as wrong as the prediction by left-wing thinkers about a “postcapitalist world without work” and “fully automated luxury communism” (p. 15).
The argument of Automation Is a Myth is that automation is a limited, localized, and socially specific phenomenon, and it echoes Gray and Suri’s (2019) point that automation discourse neglects “ghost workers”—that is, the many invisible workers who are needed to train the algorithm or to fix the kinks and flaws in technology deployment (p. 30). Quoting Tesla founder Elon Musk’s tweet “humans are underrated” (p. 17), he contends that the main problem plaguing the modern workplace is not too much automation, but insufficient automation, given its imperfections. Tesla was incapable of fixing the inconsistencies in assembly tasks that require human judgment, which resulted in less productivity. Amazon warehouses are filled with shelf-moving robots, which may have reduced the amount of walking among warehouse workers but has also increased physical injuries based on monotonous but fast-paced body movements (p. 94). Furthermore, the more robots Amazon is introducing, the more reliant they are on skilled workers who can fix and maintain the robots. For technological systems to function, workers must internalize the logic of the system and perform their activities such that the algorithms recognize them (p. 25). Rather than producing a world without work, the new work in algorithmically controlled environments could be low quality: social media content moderators must identify violent or pornographic content that the algorithms cannot detect on their own, resulting in psychological trauma (p. 38).
Munn also argues that technologies are adopted in a cultural context, noting that East Asian cultures are more likely to trust automation than Middle Easterners (p. 56). However, the dark side in the Chinese context is the use of surveillance technology to control daily lives and forced labor in cotton picking among the Uyghur Muslims, a repressed minority group that the Chinese government forces to assimilate to Han majority culture. In line with the main thesis, the surveillance technology, focused on facial recognition algorithms and cameras, is not fully automated but requires heavy police, lab work, and neighborhood watch presence (p. 69). Cotton picking is very labor intensive despite commercials about smart cotton harvesters (p. 74).
Munn also argues that automation has a racial and gender dimension: American warehouse workers tend to be poor and black, which exposes them to the indignity of stressful work while it exists and precarity if parts of the work became automated. Women tend to do housework, which has historically been uncompensated, and their work is unlikely to be automated because men, who dominate high-tech sectors, focus their innovation on paid work for which there is a market. Female coders are unfortunately pushed out of the field due to sexual harassment and bullying in the industry. In the book’s conclusion, Munn advocates for desirable, non-alienating forms of automation that are also ecologically desirable (p. 127). He lists initiatives like Data for Black Lives and Maori Data Sovereignty Network as examples of social organizations that focus attention on the pitfalls of predictive policing and algorithmic decision-making and the importance of data ownership rights. Technological development has to move toward a more egalitarian “data commons” and away from capitalist money-making.
There are some points of critique. Munn focuses too much on automation failures rather than on where it has succeeded. Surely, there are other works that have covered that thesis, including some the author has cited (e.g., Ford 2016; Susskind 2020). Horse taxis have been replaced by cars, and elevator operators have been made redundant. Factory workers still exist today in higher-skilled form, but with automation there are fewer of them. The same logic applies to warehouse workers, whose employment growth has been more than offset by losses in the brick-and-mortar retail sector. On the other hand, I agree with the author that there are limits and difficulties to automation. Munn is also right to show that the inevitability narrative sidesteps the importance of improving working conditions for existing workers. That narrative tends to increase the legitimacy of the capitalist order and the owners of technology companies that create these automation products.
This book has not rigorously created a link to the capitalist order. Marx’s contention is accurate that capitalism, while having unleashed technological progress, is its own biggest barrier to full automation, in part because there is an endless demand for new consumer products and because capitalists still derive surplus value from workers rather than robots. Capitalism requires growth in aggregate demand, investment, and the labor force, and more technology has not changed that logic. Could declining fertility and a shrinking labor force make automation more feasible, or do its limits, raised in the book, create too many bottlenecks?
I like Munn’s view of female reproductive labor as labor, even though it is unpaid and outside the labor market. But the author’s cautious skepticism of universal basic income is inconsistent with that position (p. 100). Surely, the worst-case scenario is a radical cut in welfare spending and tech billionaires being undisturbed robber barons while the underemployed masses get a pittance. But UBI could also be used as an income floor for low-wage workers and would finally societally subsidize and recognize unpaid household work. The technological wealth generated over the last two centuries should make the financing of a UBI scheme quite simple.
Last, the discussion of solutions and desirable futures is somewhat thin. What does a non-alienating, communal, and ecologically sustainable form of automation look like? Sustainability is a fraught point given that technology is so electricity-dependent. The more technologies we have, the more carbon we likely have to emit. This suggests that ecological solutions would imply less automation and a lower standard of living (“degrowth”), which is contrary to how contemporary capitalism is set up. Identifying solutions is a much harder task than what social scientists are trained to do, namely, to analyze contemporary social problems. So that last critique is certainly not a fatal one. Scholars of labor, technology studies, and economic sociology will gain great insights from this book.