No, ‘Transparency’ Won’t Stop Amazon Being a Crappy Employer

It’s official: Amazon’s far-reaching infrastructure of surveillance, digital monitoring and algorithmically-generated productivity measures are to blame for the “mental toll” faced by those who work for the company and a poor health and safety culture. Or so says the largest independent survey of Amazon workers yet, according to UNI Global Union which has published a new report into working conditions at the retail giant.

According to the research, conducted by Jarrow Insights, over 2,000 survey respondents (all self-described Amazon workers) across eight countries consistently expressed that safety comes second to productivity at Amazon, with 52% of workers agreeing Amazon’s approach to monitoring work performance has had a negative impact on their own physical health, whilst 57% recorded a negative impact on their mental health. Other responses noted indifference to variations in individual needs according to age or medical condition.

Methods of ensuring the productivity of warehouse workers ranged from hand scanners, swipeable ID badges and workstation screens, while activity tracking software measured office workers’ down-time. Workers reported fears of being punished for small errors, as well as being treated like a “slave”, “number” or “robot” by the digital system that governs their work.

We’ve long known about Amazon’s far-reaching technological infrastructure, as well as the use of data tracking that allows workers to be managed by algorithms rather than humans. It’s a central tenet of how Amazon works as a logistical operation, requiring both workers and goods to be conceived of as like-for-like components within a larger machine capable of providing world-beating e-commerce to its customers (which number in the tens of millions each month in the UK alone). Moreover, its ability to treat workers in this way is fundamental to the future Amazon hopes to achieve.

Yet unions and other campaigners have historically lacked strategy when it comes to technology, digital monitoring and algorithmic management. Instead, they’ve preferred to advocate for Amazon workers on the more familiar grounds of pay, job security, or simple union recognition itself. Of course, these issues are important, but none – even if they were all won tomorrow – threaten to protect workers’ health, safety and dignity within Amazon’s workplaces and overturn the “toll” of working for the company.

The distance unions across the world have to cover to catch up with a company like Amazon on digital monitoring, management and surveillance practices broadly depends on their historical attitude to matters arising from the organisation of work. In the Anglophone world, but especially in Britain, the organisation of work – including its corresponding technology – has typically been considered the fair business of the company rather than unions, along with companies’ interest in turning a profit. Accordingly, although technology and surveillance make for media-friendly slogans (‘Workers Not Robots’ was the initial banner of the GMB’s efforts to unionise Amazon domestically), it has seldom been the subject of actual demands. 

Since 2020, when the pandemic put a renewed spotlight onto Amazon’s working practices and the #MakeAmazonPay campaign was launched in conjunction with Black Friday, unions have begun a welcome foray into the regime of data tracking and AI surveillance that exists at Amazon and other algorithmically-enabled employers.

Initially, this was confined to concerns about workplace health and safety – a well-trodden, if narrow, avenue for unions to assert themselves in terms of the work process. It is a reasonable terrain of struggle at Amazon given the plainly injurious nature of its digital infrastructure upon workers. But more recently, union researchers are hoping to break ground in the realm of digital rights, data processing and the application of artificial intelligence.

In Britain, official efforts around Amazon’s use of algorithmic management are being guided by the TUC’s AI Manifesto. Produced in 2021, the document called for new legal protections in a context where employment law has fallen behind the curve on the use of digital technology at work. Its recommendations include a new right to “human review”, the establishment of “employment-focused ethical principles” and a “duty to consult” unions in relation to the deployment of AI in workplaces deemed at “high-risk” of automated decision making.

In common with the conclusions of UNI Global’s recent report, the TUC’s recommendations rest on an imperative to make algorithmic decisions more ‘transparent’ or ‘explainable’. But the idea that algorithmic transparency will make work more equitable, less stressful and safer is based on a misplaced sense that exposing the least visible part of a digital system – that which has been ‘black-boxed’ from view – will reveal to us the hidden elements that make working for Amazon so miserable.

Without doubt, it’s good that unions are thinking about digital technology. But by pinning hopes of political advancement on data transparency – a preoccupation gifted to us by liberal academics in the field – they risk missing the point: it’s a technical solution to a political problem. Transparency itself doesn’t mean what we find beneath the digital hood will help us better navigate (and change) the power differentials between worker and company. There’s no algorithm or section of code we could hope to isolate that would contain the explanatory power to account for the modes of organisation and politics we see enacted at Amazon, which is surely the point, rather than transparency for its own sake.

The emergent working assumption is that knowledge of the inner workings of a computer system will enable unions to argue for benchmarks and assurances about the parameters of algorithmic calculation. Yet as the UNI Global report rightly reports, managers themselves aren’t clued into the functioning of the systems they serve. This is because they don’t need to know; in other words, the secret to managerial power in Amazon isn’t to be found inside the algorithm itself. What needs to change is the parameters of the company itself.

The immiseration of work under algorithmically-enabled management regimes is a feature, not a bug. There’s no harmonious way for Amazon to function as a profitable next-day delivery specialist of the widest range of consumer goods available under one roof and at the same time be equitable for workers. Assertions to the contrary, whether by Amazon or by unions, are a fiction; we’re talking about politics, not a design flaw.