When Chris Smalls was fired by Amazon for protesting unsafe conditions at a warehouse during the COVID-19 pandemic, he made apparent a usually hidden human infrastructure: the front-line workers that make our quarantined world function. (Update, April 9, 2020, 4:45 p.m.: Amazon disputes this characterization of events. It says that Smalls was fired for violating social-distancing rules and for not remaining at home after having contact with someone diagnosed with COVID-19.) Warehouse workers, delivery drivers, hospital staff, and grocery checkers all face near-constant exposure from customers, patients, and one another.
It may be tempting to believe that the best way to protect such workers would be to have drone delivery, grocery stores without checkers, and increased automated decision-making across fields as diverse as content moderation and medical diagnosis. But automation will not keep front-line workers safe for the simple reason that automation requires people to fix, repair, and work alongside machines.
During this pandemic, when we are expected to keep 6 feet of space between us, it is easy to imagine a world with grocery shopping entirely by self-checkout and everything else delivered to our doorsteps with a click of our trackpads.
Globally, robotics and tech industry professionals appear to be claiming that such a world would be safer for us all. In “Calling All Robots,” the Wall Street Journal’s CIO Journal section describes an acceleration toward automation, reporting that 50 percent of businesses in an industry survey of large companies are using automation to “help front-line workers cope with the pandemic.” An editorial published in the journal Science Robotics suggests that during the COVID-19 outbreak, robots could be used for “disinfection, delivering medications and food, measuring vital signs, and assisting border controls,” taking over from humans these and other dirty and dangerous jobs. Ran Poliakine, founder of a robotics firm called SixAI, told the publication IndustryWeek that, in light of the pandemic, it was “more important than ever” for companies to use automation to keep workers safe. Content moderators at Facebook and other social media platforms who were working in close quarters have already been sent home to keep them safe while their decisions on the appropriateness of posts, images, and videos are made with the help of an A.I. that automates content removal. Meanwhile, the management consulting firm Bain & Company advises banks to protect their front-line employees by shifting to working from home and increasing digital banking solutions to customers who have been resistant (even while they recommend scaling up customer service support operations to make these changes effective). Drone delivery companies in Canada and China are experimenting with medicine delivery, and researchers at the Indian Institute of Technology Guwahati are developing robots to carry food to and remove waste from hospital isolation wards.*
Certainly, venture capitalists and Wall Street firms will put more money behind online, distributed, and robotic applications going forward. While most companies’ stocks are tanking, those supplying automated delivery and checkout, like Walmart and Amazon, are trending upward. Amazon stock is up nearly 20 percent since the beginning of the COVID-19 pandemic in March, and Walmart stock is at an all-time record high. Some automation efforts may yield protections for workers, especially in hospital settings. Robots can, for instance, take patient temperature readings and help with disinfecting surfaces. Tech companies often promise to eliminate the dirty, dull, and dangerous jobs through machine learning and automation. But they will not protect workers by themselves, because it’s workers who will provision, fix, and train humans to use machines properly. The best way to describe what automation means today might be: machines upfront, people in the back.
Self-checkout at grocery stores, for instance, requires cashiers to stand by to help customers and recalibrate the machines. Self-checkout machines reliably misrecognize bar codes, and people just as equally become frustrated when they don’t quite know how to give the machine the information it’s asking for. As Alexandra Mateescu and Madeleine Clare Elish show, at each of these moments, a self-checkout attendant steps in to help. If drone delivery were to become a reality, it would still require human workers to pack and load the delivery boxes. For each box delivered to the doorstep and each online meeting taken from home, there will babysitters and house cleaners who make the home a presentable office and keep children occupied. There will be customer service workers sitting in cubicles ready to answer product questions. There will be content moderators working under similar conditions in the United States, the Philippines, and India, ensuring that our socially distanced worlds run smoothly and without interruption.
Rather than creating safe barriers around all people, such a drive toward automation would simply hide the people who are at risk. Susan Leigh Star, a scholar of technological infrastructures, has made the important point that in times of crisis, the true natures of infrastructures reveal themselves.* During this pandemic, it has become abundantly clear that the workers who labor alongside the screen are some of the most vulnerable to contracting and spreading the virus. They cannot afford to stay home, and their employers classify them as independent contractors rather than employees who would have access to benefits. But the social distancing the rest of us can engage in depends on them continuing to work at packing boxes and reviewing what we see online. This human infrastructure was already in place; COVID-19 made its extent and the risks associated with this type of work incontrovertible. Taking in the international scope of this labor force, it is also clear that the effects of the pandemic will disproportionately fall on Black, Asian, and Latinx workers, many of whom are immigrants. In New York City, 75 percent of front-line workers—in grocery stores, warehouses, cleaning, transit, child care, food, health care, and more—are people of color, and 50 percent were born in a different country.
Perhaps one day we will see more machine learning applications for diagnosing sick patients, allowing doctors to spend time making decisions about proper treatment without unnecessary exposure. Maybe artificial intelligence will be used to help track and stem pandemics. But even as we pursue these options, we should not believe that reduced exposure for all will be achieved through these means. Instead, there will be people in all of these spaces, part of an automation infrastructure but at the same time partially hidden.
What is needed now and on the other side of this pandemic is increased protections for these workers. There are several initiatives currently underway to achieve this. In New York state, Gov. Andrew Cuomo has passed an emergency sick leave bill, and the Department of Labor waived the seven-day waiting period for unemployment benefits. The system that takes unemployment insurance applications, however, is so overloaded that a de facto waiting period seems to be in effect anyway. Some unions are also negotiating relief packages for their workers. While these measures are important, unemployment benefits extended to independent contractors and freelancers, many of whom provide food delivery and other services for those social distancing, are stopgap measures. In general, these workers have to rely on the largesse of their employers to provide sick leave benefits, a precarious proposition when these companies are not normally bound by employment law. Meanwhile, workers in fulfillment centers and food distribution warehouses may have to continue working in unsafe conditions.
Even the best of these crisis measures are only temporary. They will disappear once this pandemic is behind us. Future visions of automation need to recognize humans as partners up in front with the machines, rather than hiding the people who make automation work at the back. There are many ways to do this. Two of the most important are including freelance and gig workers in state protections and preventing tech companies from presenting themselves as exceptions to corporate regulations in the U.S. and abroad. Beyond these practical regulatory levers, we should not be lulled into believing that introducing more automation will keep all of us safe. We need a different story about automation, one that subverts the narrative that robots operate in the absence of people and instead gives equal weight to the people who make what seems automatic work.
Correction, April 9, 2020: This piece originally misspelled Guwahati and Susan Leigh Star’s last name.