Image: Joe Piette via Wikimedia Commons (CC BY-SA-2.0)
Published jointly for Power at Work and Work in Progress
Employers are increasingly using algorithms and digital devices to control workers. As a new Human Rights Watch report puts it, “Workers around the world are increasingly hired, compensated, disciplined, and fired by algorithms that can be opaque, error-prone, and discriminatory; their faces, office badge swipes, email exchanges, browsing histories, keystrokes, driving patterns, and rest times are scanned to monitor performance and productivity.”
My research shows how this “algorithmic management” not only affects working conditions, but also expands the capacity of employers to subvert the efforts of workers who wish to organize unions for better treatment and to increase their power in the workplace.
It details just how they can do so by studying an early battle in the current labor upsurge: the first large union election at an Amazon warehouse, which took place in Bessemer, Alabama in early 2021. Drawing on interviews with 42 Amazon workers and court records obtained through a Freedom of Information Act (FOIA) request, my investigation reveals that Amazon weaponized aspects of algorithmic management against the union drive by: A) repurposing tracking devices, workstation displays and a mobile app to stoke fear and doubt; B) engaging in “algorithmic slack-cutting” to curry favor; and C) exploiting patterns of social media activity encouraged by algorithmic management.
These findings highlight a vital but underexplored consequence of algorithmic management: its potential as a sophisticated tool for union busting. They also underline the value of exploring how control techniques can shape counter-organizing opportunities for employers, not just organizing opportunities for workers.
Five ways employers weaponize algorithmic management to weaken worker power
1. Weaponizing tracking devices: Algorithmic management often involves the use of digital devices to monitor and discipline workers. For example, mobile ride-hailing apps direct drivers and automatically terminate them if their ratings fall below a certain level. In Amazon’s case, scanners, badges and computers play a similar role. Beyond tracking productivity and administering write-ups, these tracking tools also aid employers’ anti-union tactics. Amazon theatrically deployed tracking devices in mandatory anti-union meetings where supervisors and human resource officers scanned attendees and watched workers with open laptops as the meetings unfolded, according to my interviews and court testimony by workers. Workers said that Amazon even publicly singled out employees whose questions or comments appeared to indicate support for the union by scanning their badges as other workers looked on.
2. Repurposing workstation displays: Employers have long polled workers during anti-union campaigns, but algorithmic management greatly expands the capacity of employers to identify union partisans and effectively campaign against them. Amazon used workstation displays (monitors that instruct workers on how to do their jobs) to beam workers with anti-union messages, such as “vote ASAP and vote No,” and to ask questions to gauge their union sympathies. Several interviewees said they abstained from voting because they believed Amazon could find out if they voted for the union and then fire them.
Image: Socius (CC BY-NC 4.0). (This is from a display in the Bessemer warehouse a few months after the election.)
3. Algorithmic slack-cutting: Employers habitually improve working conditions to peel off votes during anti-union campaigns. But thanks to the flexibility and dehumanizing effects of algorithmic management, employers can turbocharge this tactic through what I call “algorithmic slack-cutting.” To explain this technique further, I need to specify two components of Amazon’s algorithmic management system.
The first component is what Jamie McCallum calls the “electronic whip.” This is a drive system that uses algorithms and digital devices to intensify the pace, monitoring and discipline of work, often resulting in high stress and injury rates among workers. Amazon’s electronic whip tracks the minutes workers are not working, enforces algorithmically-generated quotas, and automatically fires workers if they deplete their allotted unpaid time off. The second component is automated HR management. This typically involves using a mobile app to handle HR-related issues and questions, resulting in the elimination of the “interpersonal and empathetic aspects of people management.”
Ironically, the degrading effects of these aspects of Amazon’s algorithmic management also gave it the capacity to engage in slack-cutting to quickly gain favor with some workers. Confronted with the union drive, Amazon softened the “electronic whip,” such that Time Off Task, quotas and automatic termination for unpaid time off depletion went largely unenforced during the campaign. Simultaneously, it temporarily reintroduced the human touch missing from its automated HR management. It did so by flooding the floor with human resource officers, out-of-town managers and consultants to solicit grievances, express empathy and offer to help solve problems.
4. Leveraging a mobile app: Employers often deliver anti-union messages through letters, phone calls and text messages, but mobile apps significantly expand the capacity of employers to reach into the private lives of workers. Amazon workers are all but captive to “A to Z.” Workers are supposed to use the app to correct punches, request time off and file HR reports. They also receive important notifications through A to Z that they can ill afford to ignore, including schedule-change alerts and offers of overtime or “voluntary time off.” During the campaign, Amazon leveraged workers’ dependency on the app to reinforce its warnings to workers through captive audience meetings and other communication channels.
Image: Socius (CC BY-NC 4.0). (This image includes push notifications sent to a Bessemer warehouse worker a few months after the campaign ended.)
5. Exploiting social media: Workers create and cluster in social media groups to decipher, gripe and learn tricks for coping with the opaque processes and dehumanizing effects of algorithmic management. In Amazon Facebook groups, members commonly complain and ask questions about issues like quotas, unpaid time off (UPT) and automated terminations, with discussion sometimes turning to unionization as a partial solution to their grievances.
To deter unionization, Amazon used three techniques to exploit workers’ social media activity. First, it infiltrated such groups to “capture” and categorize posts of interest for potential investigation. Second, it paid employees through an “ambassador program” to counter “all posts and comments” questioning the “FC [fulfillment center] associate experience” and leave “no lie unchallenged,” so the public can learn that Amazon workers “love their jobs,” according to a leaked internal document. Third, it used a high volume of social media ads and communications to campaign directly against the union. The incentive structure and fear these techniques generated appeared to bias the social media discussion in an anti-union direction and cause anti-union rumors to spread like wildfire.
The counter-organizing opportunities arising from algorithmic management described above are just the tip of the iceberg.
Among other tools that can be used to suppress worker organizing are automatic censorship of union-related terms on internal communication platforms; polling software and data analysis to generate real-time unionization scores to guide prevention efforts; role-playing AI bots to train supervisors how to question, persuade and surveil workers to great effect; and AI analysis of “feedback, Slack messages, and even subtle trends in workplace chatter to figure out what workers are feeling and what issues are about to pop off
Another avenue through which counter-organizing efforts have been expanded is the weaponization of diversity, equity and inclusion (DEI) management. This includes the rebranding of anti-union consultants as diversity experts, the use of employee resource groups and DEI officers to detect and campaign against organizing, and the manipulation of social justice and wellness tropes to counter unions. These efforts alert us to the tensions and synergies between identity politics and class politics. Indeed, concerns around diversity, inclusion and equity animate organizing campaigns, such as the Starbucks union drive, and anti-union consultants are defending DEI initiatives on the grounds that their potential removal could encourage unionization.
Understanding these evolving anti-union strategies is crucial for labor advocates, policymakers, and workers themselves to develop effective countermeasures and protect organizing rights in an increasingly digitized workplace.
Organizing takeaways?
Are there any organizing takeaways from my research? Here are five potential lessons.
First, while real-life organizing conversations between workers remain most important, some pro-union workers should be ready to campaign for the union and refute comments by antiunion activists in the relevant social media groups. Some Amazon workers decided not to vote for the union due to the antiunion comments from coworkers that they encountered on social media groups (the many pro-union comments from outsider supporters mattered little to them). As I detail in my study, social media seems to have a structural anti-union bias, so conscious efforts to counteract this may be well advised.
Second, conceptualizing the “electronic whip” (a term originally coined by laundry workers) and automated HR as two encompassing forms of algorithmic management may help more workers clarify and target some of the major sources of their grievances — perhaps serving as analytical and rhetorical resources during more organizing campaigns.
Third, pointing out the superior communication and surveillance capacities available to employers for anti-union campaigning due to algorithmic management can further highlight just how rigged the election game is against workers. Activists might publicly ask not just for as much time as management to present their case during company meetings, but also for the ability to campaign through the company app and workstation displays. If the employer can, why can’t workers? The likely denial of such requests will further discredit any purported commitment to a free and fair election that employers and consultants might have expressed.
Fourth, this research underscores the continued need to find high-level programmers, HR professionals and even employee relations team members with guilty consciences. Unions need to persuade more specialists, managers and even executives – some of whom are also facing dehumanizing AI-related pressures – to disclose other ways in which employers are mobilizing technologies against workers behind the scenes. What are Amazon’s “intelligence analysts” up to these days?
Fifth, the paper offers more fodder to attorneys and organizers grappling with how algorithmic management impacts the organizing rights of workers. And its findings might be helpful to organizers looking to pick up on new forms of employer activity that could in some instances potentially constitute unfair labor practice charges. Relatedly, workers at an Amazon warehouse in St Louis, Missouri recently filed an unfair labor practice charge against Amazon alleging that the firm’s use of “intrusive algorithms” interfere with workers right to engage in protected concerted activity. Practitioners and academics interested in learning more about the legal implications of algorithmic management should consider consulting a new legal brief on the topic by Melanie Miginucci and Brishen Rogers’s Data and Democracy at Work.
Editor’s note: This article was authored jointly for Power at Work and Work in Progress, where a shorter version appeared.
