×

Expand

Damian Dovarganes/AP Photo

An employee uses an automated packaging machine to create a made-to-measure bag for small items at the Amazon OXR1 fulfillment center in Oxnard, California, August 21, 2024.

A new paper on Amazon’s anti-union efforts at its Bessemer, Alabama, warehouse is the first academic study to examine the ways Amazon has leveraged algorithms to crush pro-union movements inside one of its sprawling warehouses. The study, published by researcher Teke Wiggin of Northwestern University, was compiled using dozens of worker interviews and FOIA requests to the National Labor Relations Board.

Ultimately, the 2021 union vote in Bessemer ended with workers voting against joining the Retail, Wholesale and Department Store Union (RWDSU) by a count of 1,798 to 738 out of 5,867 workers eligible to vote, proving that the sprawling surveillance and discipline system is highly effective.

Since the 2021 election, Amazon has been ordered to hold two more elections at the plant. However, the company has proved exceptionally deft at thrashing worker organizing with both old-school and cutting-edge technologies. The modern tactics detailed by Wiggin include leveraging algorithmic discipline on worker devices, algorithmic slack-cutting or “electronic whipping,” in which algorithms rapidly alter quotas and rules, mobile device spamming, and social media monitoring.

More from Daniel Boguslaw

Amazon weaponized workplace devices in use at warehouses that algorithmically direct and discipline workers. It exploited these machines to send anti-union messages, ask questions that workers say were designed to gauge union sympathies, and make “captive audience” meetings even more intimidating. During management-led meetings where workers were fed anti-union talking points, employees at the Bessemer plant would scan workers to monitor their efficiency and disciplinary record, reminding them about the digital eye surveilling them throughout the workday.

The second tactic, dubbed “algorithmic slack-cutting” by Wiggin, was used to rapidly humanize working conditions to make workers feel better about their jobs. Amazon did this in two ways, first by softening enforcement of algorithmic quotas and rules and by removing automatic termination initiated for missed quotas, thus dampening the so-called “electronic whip.” The second part of this process was improving an HR experience that had been nearly fully automated. During the election, Amazon flooded the warehouse floor with HR managers trying to soothe concerns raised by the union about an unresponsive and robotic HR process.

But slack-cutting loosened during the union election time period could always be tightened after. And while Amazon attempted to soften parts of the disciplinary process to address worker demands, it simultaneously upped its algorithmic harassment on employees’ personal devices. Amazon leveraged the app relied upon for overtime offers, shift changes, and pay records to disseminate anti-union messages that workers couldn’t ignore, given the importance of their other notifications.

The final tactic Amazon utilized was exploiting social media activity. Wiggin writes that the extreme stress cultivated in the Bessemer warehouse from AI and algorithmic surveillance led workers to cluster into social media groups to discuss grievances and figure out how to circumvent the specific contours of Amazon’s digital surveillance.

According to the study, Amazon “ran a social media surveillance program that monitored more than 43 Facebook groups, most of which were nominally private, as well as numerous Web sites [and] subreddits.” It also found that “[t]he program’s described aim was to ‘capture’ and categorize posts of interest for potential investigation, including those mentioning complaints from warehouse workers and planned strikes or protests.”

Wiggin details employee accounts of being terminated for social media posts about working conditions. He also highlights an “ambassador program,” which in 2018 paid warehouse workers to counter criticism about working conditions at Amazon. “[W]orkers were recruited and trained to counter ‘all posts and comments from customers, influencers (including policymakers), and media questioning the FC [fulfillment center] associate experience … leaving no lie unchallenged and showing that people who actually know what it’s like to work in our FCs love their jobs.’”

AI and algorithmic technology was not deployed solely to monitor workers, but also the unions fighting to represent them. A 2020 memo obtained by Vox details an artificial-intelligence platform used to map organized labor across site locations: “geoSPatial Operating Console, or SPOC—would help the company analyze and visualize at least around 40 different data sets,” the document says. “Among them are many related to unions, including ‘Whole Foods Market Activism/Unionization Efforts,’ ‘union grant money flow patterns,’ and ‘Presence of Local Union Chapters and Alt Labor Groups.’ Additionally, one of the potential use cases for the tool is described in the memo as ‘The Union Relationship Map,’ though no other details are provided.”

Wiggin says that his paper shows that Amazon is not just tweaking pre-existing AI systems to make unionization harder to achieve for workers, but actually converting and weaponizing sprawling systems into new tools for quashing dissent. “We knew employers could leverage algorithmic management by generating real-time unionization risk maps, but this study reveals that these managers are acting as transformers when they convert the devices and software to have dehumanizing effects and engage in more oppressive forms of coercion,” Wiggin said, adding that “another way of thinking of it is employers have an arsenal of weapons to fire at unions whenever they want, and these algorithmic tools aren’t just whips, they’re also bazookas.”

As Amazon CEO Jeff Bezos sidles up to Donald Trump with a seven-figure inaugural committee donation, a revamped and MAGA-curious Washington Post, and the rebroadcast of The Apprentice on Amazon Prime, the movement for a unionized Bessemer warehouse enters its fifth year.

After National Labor Relations Board officials called for a second Bessemer election over claims that Amazon had illegally influenced the vote, the results of that 2022 vote were also contested by both sides for years, leading to the scheduling of a third election. That vote now seems improbable, given Trump’s recent actions to decimate the NLRB by illegally firing board members, though a federal judge reinstated Democrat Gwynne Wilcox and she returned to work on Monday. Eventually, Trump can nominate new members to the board and obtain a MAGA majority.

In the conclusion of his study, Wiggin writes, “Amazon’s warehouse regimes merit the modifier despotism because they clearly have an overall coercive character. This character is the product of overtly coercive control techniques, such as algorithmic management. But coercion is also implicit in the firm’s consent-oriented techniques because the techniques are premised on underlying coercive conditions, including a despotic labor market, algorithmic discipline, and in the case of Bessemer at least, ‘plantation-style management.’”