Performing Automation

Interior, pizza arcade.

Sioux Falls, 1995 or thereabouts. Family groups and clusters of post-soccer-practice tweens sit around napkin-strewn tables, while younger kids play whack-a-mole and skeeball in the arcade corner. A high-pitched electronic wheeze becomes audible above the din, heralding the arrival of Wilbur the Coyote. Wilbur—a four-foot-tall robotic coyote—rides a tricycle attached to an electronic track, and carries a tray laden with pizza and french fries. He arrives at tables, greets the customers by name, and announces their order to bemused adults and slightly terrified children. You must lift the pizza from his tray yourself, but if you’re lucky, Wilbur will stick around for some banter or to sing Happy Birthday to anyone celebrating.

Robots of all shapes and sizes are often called on to perform for an audience—dancing, wiggling their eyebrows, entertaining children on state television, or just rolling along silently with their glowing eyes. In 2018, the humanoid Robot Boris appears on stage at a televised youth robotics forum in Russia showing an impressive range of motion and leading the audience of teens in a simple movement exercise. In 2021, Elon Musk announced his forthcoming Tesla Bot project by ushering a robotic figure onto the stage, its initial jerky movements giving way to a joyful vaudevillian dance as the beat drops.

Behind the one-way mirror at the pizza parlor, a very different scene takes place. A teenage employee sits at a desk, controlling Wilbur’s movement around the track and speaking lines into a mic so they emerge from the coyote’s fluffy mouth across the room. Wilbur’s controller hams it up to amuse his friends working over at the counter, making Wilbur bust out in-jokes and verging-on-inappropriate comments to annoying customers. From this perspective, Wilbur isn’t a robot; he’s a remote-controlled puppet, and this kid is a puppeteer. The same is true of the other performing robots. Beneath the carapace of Robot Boris and the sleek suit of the dancing Tesla Bot, human actors glide and twerk with impressive muscle control, a plastic mask keeping them at one level of remove from the critical gaze of the audience, which stays fixed on their shiny surface.

Where, in all of this dancing and play acting, are the real robots? The promise of robots is that they aren’t humans, but instead “a physical machine that’s usually programmable by a computer that can execute tasks autonomously or automatically by itself.”1 From The Jetsons’ Rosie the robot maid to the precision-engineered dream robot Ava in Ex Machina, the robots we tell stories about are almost invariably mobile service robots: humanlike and hyper-capable.2 By this standard, most of history’s famous robots aren’t robots at all, but are actors wearing robot suits or hiding behind one-way mirrors. However, robots also have immense symbolic value, and whether or not they’re “real” or effective is beside the point. Companies put robots on display in workplaces and public environment to signal their commitment to technological progress and future automation.3 While general purpose humanoid robots don’t exist yet (and they may never exist), we’ve long been dreaming of and play-acting as them. Pretending that no human labor is taking place. Performing automation for an audience that wants to believe.

Workers have always called upon to pretend to be inhuman, automated, in order to serve our bosses without disturbing their peace or sense of order. But the robot actor is a specific kind of performance. Despite the uneven power dynamic inherent in robot-audience interactions, robot play has a specific set of pleasures. The ability to unsettle. The potential to cause chaos, to break the fourth wall. The raw enjoyment of putting on a performance. Of course it’s not all peachy, being a robot. And it’s not like me to look on the bright side of wage labor, let alone the most poorly paid and least respected kinds. All I can say is this: I’ve done robot jobs, hidden my labor behind the pretense of automation, and there was plenty of enjoyment to be had.

The spatial dynamics of robot performance are changing, with globalization and advances in computing capability creating new architectures within which the robot workers of the world must operate. Robot play now often takes place at a distance—as with drone operators or remote human assistants to “self-driving” cars—leaving the actors further alienated from the audience or customer for whom they perform. As with all working environments, the structures within which robot labor is performed defines what is possible for workers, what can be demanded by bosses, and what small pleasures or opportunities for resistance can be eked out along the way.

The Transference Machine

Much has been written in recent years about the purposeful overhyping of automation technology. Astra Taylor writes about “fauxtomation,” overblown claims of automation that obscure or devalue human labor, and Jathan Sadowski uses the term “Potemkin AI” to describe systems designed wow a credulous public with the appearance of AI, all the while relying on workers to make them run. 4 These coinages point to the grim yet funny reality that new “automated” services and products often depend on wholly un-automated human labor to perform essential tasks beside or behind the machines. However, to call robot play “fauxtomation” would be to overstate how much fakery is going on. Wilbur the Coyote, the dancing Teslabot, or Robot Boris are artificial automation in the same way George Clooney on ER was an artificial doctor. They’re people playing a role to an audience that is content to suspend its disbelief for the sake of entertainment or convenience.

Certainly, robotic technology exists and is in use: industrial robots weld metal and shift boxes the world over, and robotic-assisted surgeries allow doctors perform complex procedures through keyhole incisions. With the exception of Boston Dynamics’ commitment to producing the ultimate in broad-shouldered parkour war-bots, most actually-existing robots are incredibly prosaic: a programmable electronic arm, a remote operated vehicle. To bundle these varied machines—factory robot arms, Amazon’s homebot Astro, Star Wars’ C-3PO—under the single term “robot” is to conjure a suite of images and attendant fears from Skynet-style superintelligence to the threat of technological unemployment. As with “artificial intelligence,” where varied computational processes are packaged together with an aspirational and ultimately misleading name, the “robot” is a language trick that hides human workers and abstracts labor.

This container called robot is the current manifestation of an age-old fantasy. From ancient Greece to third-century China, accounts flourish of inventors and mathematicians attempting to manifest self-operating machines and automata in order to entertain, amaze, and assist patrons and emperors. In Jewish folklore, stories of golems—anthropomorphic beings created from mud or clay—have proliferated since at least the sixteenth century, often conjured to serve and perform drudgery for their maker. The specific word robot (from the Czech “robota,” meaning “forced labor”) wasn’t introduced to English until 1920, when Karel Čapek’s play R.U.R. (Rossum’s Universal Robots) updated the old automata myth for industrial modernity. As with the golems and automata before them, one of the essential qualities of Čapek’s robots is that they might perform repetitive labor without tire or complaint. As R.U.R.’s engineer character states, “One Robot can replace two and a half workmen. The human machine … was terribly imperfect. It had to be removed sooner or later.”5

Enduring but unstable, the archetype of the robot is ripe for transference onto human bodies, mobilized by slaveholders and hucksters alike to dehumanize and delegitimize individuals and whole categories of people. In “The Uncanny History of Minstrels and Machines, 1835–1923,” Louis Chude-Sokei recounts the story of notorious showman P. T. Barnum, who built much of his career around the display of enslaved woman Joice Heth.6 To gin up audience intrigue, Barnum encouraged the speculation that Heth may in fact be an automaton, “made up of whalebone, India-rubber, and numberless springs ingeniously put together, and made to move at the slightest touch, according to the will of the operator.”7 For white audiences in the 1830s (and others long afterwards), to insist on or imagine a Black person as an inhuman automaton, rather than a fellow human, was a maneuver that served to assuage what Chude-Sokei identifies as “a fear among whites of their own transformation into slaves by machines.”8 Regardless of whether they’re dancing, working, or at rest, the robot or robot-actor’s serves as a kind of animate morality play for viewers, turning labor into a performance and shoring up viewers’ sense of their own humanity in contrast to the robot’s perceived mechanical lack. The robot is a storyteller, whether they intend it or not.

In the workplaces and on the stages where today’s pretend robots ply their trade, the relationship between the robot and their audience involves a shared negotiation of suspended disbelief. The robot-actor performs subservience and inhumanity, and the audience-master gets off on being served, both knowing the other could break the illusion at any time. It’s a relationship mediated through a capitalist social relation as well as through the skin of the robot. That skin—the shiny plastic carapace of the Tesla Bot or the one-way mirror of Wilbur’s operator—is a small affordance for the robot worker, a barrier that allows them to carve out a half-step of space between themselves and the audience. Viewers can project what they like onto the robot, but all they’ll see is their own face reflected back at them from the gleaming surface.

Michel De Certeau describes a small act of labor refusal or recuperation by the worker as la perruque, or “the wig” when “the worker’s own work disguised as work for his employer.”9 In the robot actors’ case, we might describe their withdrawal from the surveilling gaze of the boss or customer and the small sliver of freedom it opens up as a twist on la perruque: “the mask” (le masque if we insist on remaining en français). The mask offers refuge as an affordance to the performer-worker. It allows the worker to keep a piece of themselves aside, withholding their full self from the boss’s gaze by throwing up a shield of shiny perspex or brushed steel.

Remote Control

Unsurprisingly, changes are afoot in the robot landscape. The intimate architecture of the robot-actor’s encounter with the audience eventually becomes a risk to those who seek to make a profit from the illusion of automation, or those who want access to surveil the worker’s whole self. In the past two decades, though, advances in computation and lightning-fast internet speeds have rearranged spatial dynamics of robot work and expanded the terrain across which robot play can be staged. This transition has been partly obfuscated by the recent omnipresent conversations about rising automation the threat of technological unemployment, promulgated by everyone from McKinsey & Company to Barack Obama. From these conversations, one would assume that if a robot is present, human labor has been removed or made redundant. Aaron Benanav describes this as “automation theory… a spontaneous discourse of capitalist societies, which, for a mixture of structural and contingent reasons, reappears in those societies time and again as a way of thinking through their limits.”10 This discourse belies a different technologically-enabled reality. Today, worker-performers still control the movement of robots that surprise and delight customers, but these workers are no longer behind a mask or one-way mirror. Cleaved from the body of the robot and removed from physical proximity with their customers, robot workers now labor from their homes or offices in cities and countries around the world.

As with manufacturing and customer service before it, robot work can now be displaced and distributed through the same colonial pathways that fulfilled the quest for exploitable resources and cheap globalized labor for centuries. In Medellín, Colombia, workers remotely pilot small “autonomous” food delivery robots called Kiwibots around college campuses in the US, delivering food to students (the last yard problem remains unsolved; like the Wilburs before them, the Kiwibots require customers to manually remove their food from the robot’s clutch).11 Ensconced in a secluded room at a Nevada airbase, US Air Force employees control Predator drones circling over villages in Pakistan and Afghanistan, ambivalently aware that their bombs’ targets, the “black blobs on a screen,” are human beings.12 Employees in San Mateo California guide large surveillance robots around far-off tech offices and manufacturing facilities, operating multiple bots at once to reduce labor costs for companies that would otherwise employ a full-time human guard.13 The distribution and decentralization of robot work is a boon to employers, especially as workers are increasingly managed not by an on-site supervisor but by the demands of a computerized management system. These algorithmically optimized systems direct the workers’ activity—whether via opaque platform apps or heavily surveilled workplace technology—and cut off many of the usual avenues of dissent and worker solidarity. Although for now, the fantasy of the fully autonomous, general purpose robot remains elusive, it’s no longer a problem for today’s “autonomous” robot companies. Indeed, it’s possible that the truly autonomous humanoid robot was always a red herring: the real promise of today’s robot technology is not the fully-automated worker, but the fully-automated boss.

“How hard will the robots make us work?” asks Josh Dzieza in his deep dive into automated management for Verge.14 Dzieza runs through the broad range of workplaces now controlled by “robot” bosses, from insurance call centers to hotel maintenance crews to factories. Even promotion decisions are now ceded to automated processes and predictive analytics that make decisions based on guesses at a worker’s future performance, as in the case of IBM’s Watson Analytics.15 Workers at some Uniqlo and 7-Eleven stores have their sales data tracked by Percolata, a Silicon Valley company that claims to calculate the productivity of an individual retail worker and allow human managers to optimize staffing levels. Microsoft’s Workplace Analytics tools can track the time workers spend on different tasks, from meetings to “deep, focused work,” and managers can use these insights to “nudge” workers or discipline teams. Across all these platforms and contexts, automated systems are used to wring more productivity from workers and endlessly optimize processes to benefit the bottom line.

In this move towards the distanced worker and the automated boss, the dehumanizing gesture of imagining the worker as mere robot is no longer necessary. The conceptual transformation of the worker, the enslaved person, or the colonized person into a laboring machine is long complete, and capital has reorganized itself around this framework. Instead, by using automation for what it’s best at—tasks like optimizing processes and monitoring behaviors—corporations can achieve the goal of a ceaselessly toiling workforce while externalizing the risk of unruly encounters or noncompliant labor that comes with in-person robot-workers. For the worker, this means an ever-intensifying pace of work and fewer opportunities for dissent. An Amazon warehouse worker Dzieza interviews tells of “a new device that shines a spotlight on the item he’s supposed to pick, allowing Amazon to further accelerate the rate and get rid of what the worker described as ‘micro rests’ stolen in the moment it took to look for the next item on the shelf.”16 The constant attention of the automated management system extinguishes the possibility that the worker might stealing back time for their own purposes. Le masque, c’est fini.

Testimonials from Dziera and other exhausted subjects of automated management systems echo the accounts from workers of an earlier era, after Taylorist principles of scientific management changed factory work by increasingly breaking down production into specialized repetitive tasks in the pursuit of economic efficiency. Gary Bryner, a General Motors union representative interviewed by Studs Terkel in his 1974 book Working, describes the reality of scientific management in the GM factory: “Instead of having the guy bend over to pick something up, it’s right at his waist level… [They try] to take every movement out of the guy’s day, so he could conserve seconds in time, to make him more efficient, more productive.”17 The repetition and lack of autonomy in these jobs wears on the worker. In “Prerequisite to Dignity of Labor,” Simone Weil’s reflections on manual labor in the wake of her time working in auto and electronics factories, she writes “The inferior kind of attention required by taylorized (conveyor-belt) work is incompatible with any other kind of attention since it drains the soul of all save a preoccupation with speed.”18 When Weil complains to her coworker that they don’t even know what machines or vehicles their labor is contributing to, her coworker replies bitterly “Others are here to think for us.”19

In today’s distributed work environment, where precarious platform workers are assigned jobs by an algorithmic interface, and where computerized management systems push warehouse workers to perform a set number of tasks per hour, the hardships are similar though the boss is more abstract than ever. New factory the same as the old factory. Perhaps the true job of the humanoid robot was never to perform labor effectively, but to entertain, to put on a pantomime of servitude while the real work took place elsewhere. This dual purpose—shoring up viewers’ sense of their own humanity, sometimes at the expense of a robotized other, and signaling the robot-owner’s technophilia and commitment to progress—helps smooth the contradictions of a long-term decline in labor demand and the as-yet unrealized promises of general purpose robots and AI. The affect and excess of robot performance—the dancing, the sassy backchat, the gliding through hospital hallways and wiggling of digital eyebrows—that’s the robot’s essential role. In public, the robot boogies and shows off on stage, serving its traditional job as a proxy, an entertainer, and a screen on which to project fantasies during moments of social anxiety. In workplaces and warehouses, however automated management means that roles are reversed: we perform for the robot.